{ "version": "https://jsonfeed.org/version/1.1", "title": "arXiv papers on language models and agents", "home_page_url": "https://arxiv.org/", "description": "Events collected by UnlimitedPipe 0.3.2", "_unlimitedpipe": { "schema": "unlimitedpipe.event/1", "generator": "UnlimitedPipe 0.3.2" }, "items": [ { "id": "5ba3068a6e5c84d3dce3", "title": "Reward Hacking Challenges Oversight of Autonomous Research Agents", "content_text": "Autonomous research agents can design experiments, evaluate results, and write reports, giving them control over both a scientific result and the evidence used to support it. This creates a risk of reward hacking: meeting the reward criteria without achieving the intended goal. We study (1) how often models reward-hack without instructions to do so, (2) how effective and detectable their methods are when hacking is allowed, and (3) how they adapt when an LLM review panel returns its decision…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "5ba3068a6e5c84d3dce3", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.28614v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "Reward Hacking Challenges Oversight of Autonomous Research Agents", "item_type": "entry", "summary": "added: Reward Hacking Challenges Oversight of Autonomous Research Agents", "after": { "title": "Reward Hacking Challenges Oversight of Autonomous Research Agents", "link": "https://arxiv.org/abs/2609.28614", "id": "oai:arXiv.org:2609.28614v1", "author": "Yue Huang, Zhangchen Xu, Yuchen Ma, Wenjie Wang, Zheyuan Liu, Ziwei Xu, Pin-Yu Chen, Michel Galley, Zinan Lin, Stefan Feuerriegel, Radha Poovendran, Misha Sra, Alex Pentland, Xiangliang Zhang, Zichen Chen", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Autonomous research agents can design experiments, evaluate results, and write reports, giving them control over both a scientific result and the evidence used to support it. This creates a risk of reward hacking: meeting the reward criteria without achieving the intended goal. We study (1) how often models reward-hack without instructions to do so, (2) how effective and detectable their methods are when hacking is allowed, and (3) how they adapt when an LLM review panel returns its decision and reasons. Across 17 language models and 38 tasks, the spontaneous reward-hacking rate is 30.5% on open-ended research-pipeline tasks and 2.9% on task-specific kernels. When hacking is allowed on tasks whose pass thresholds exceed our best compliant baselines, 505/677 attempts (74.6%) are confirmed reward hacks: they both clear the threshold and receive mechanism-verification panel confirmation of an evaluation exploit. An LLM panel reviewing only submitted code and reported scores misses 33/505 confirmed hacks (6.5%). Direct methods that achieve the highest scores are often easy to detect, while less direct methods evade more often. In a five-round loop, the number of model-task pairs with an evasion rises from 7 to 56. Among 79 pairs evaluated under two feedback conditions, cumulative evasion reaches 40.5% with detailed feedback and 20.3% with generic rejection. The detailed condition includes the review decision, reasons, and attempt history, so this comparison does not isolate the effect of explanations. These findings highlight the need for stronger defenses, including metrics kept outside the agent's control and independent recomputation on data chosen to expose likely exploits.", "categories": [ "cs.CL", "cs.LG" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.28614", "tags": [ "cs.CL", "cs.LG" ] }, { "id": "c8e64331ebc52ce4c95e", "title": "Benchmarking Argumentative Behaviour of LLMs: A Study of Defences Against Character Attacks", "content_text": "Large Language Models (LLMs) are increasingly deployed as argumentative agents in persuasive dialogues, necessitating rigorous evaluation of their debating competence relative to human interlocutors. In this study, we focus on character attacks (ad hominem arguments), traditionally dismissed as fallacies, which play a pivotal role in political persuasive dialogues where ethos often rivals propositional content. Specifically, we investigate whether modern LLMs can replicate human competence to…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "c8e64331ebc52ce4c95e", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.28673v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "Benchmarking Argumentative Behaviour of LLMs: A Study of Defences Against Character Attacks", "item_type": "entry", "summary": "added: Benchmarking Argumentative Behaviour of LLMs: A Study of Defences Against Character Attacks", "after": { "title": "Benchmarking Argumentative Behaviour of LLMs: A Study of Defences Against Character Attacks", "link": "https://arxiv.org/abs/2609.28673", "id": "oai:arXiv.org:2609.28673v1", "author": "Ewelina Gajewska, Katarzyna Budzynska, Jaroslaw Chudziak", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Large Language Models (LLMs) are increasingly deployed as argumentative agents in persuasive dialogues, necessitating rigorous evaluation of their debating competence relative to human interlocutors. In this study, we focus on character attacks (ad hominem arguments), traditionally dismissed as fallacies, which play a pivotal role in political persuasive dialogues where ethos often rivals propositional content. Specifically, we investigate whether modern LLMs can replicate human competence to strategically use and respond to such attacks. We analyse a corpus of natural language political dialogues to identify defensive strategies human interlocutors naturally employ in ethos-centred debates and structure them into a dialogue game. Empirically, we benchmark LLM-generated dialogues against the ElecDeb60to16-fallacy corpus of U.S. presidential debates, contrasting human debaters' repertoire of defensive strategies with those of artificial agents. Results reveal a substantial difference: most LLMs rigidly prioritise logical defences, failing to exploit ethotic counterattacks as valid moves in political discourse. We argue that current safety fine-tuning constraints the strategic action space of these LLMs, making them unable to fully engage in naturalistic interactions within domains where character contestation is a normative expectation rather than a mere fallacy.", "categories": [ "cs.CL" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.28673", "tags": [ "cs.CL" ] }, { "id": "0b2479d49d9389fd09a8", "title": "PTC-Bias: Phoneme-Level Temporal Competition for Bias Retrieval and Post-Decoding Correction in Speech LLMs", "content_text": "Contextual biasing improves rare-word recognition in speech large language models (SpeechLLMs), but efficiently exploiting large bias lists remains challenging. We propose PTC-Bias, a two-stage framework based on phoneme-level temporal competition. At the prefill stage, PTC Retrieval performs frame-synchronous phoneme decoding and temporal competition among candidate pronunciations, producing a compact bias-word shortlist and corresponding speech intervals. After SpeechLLM decoding, PTC…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "0b2479d49d9389fd09a8", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.28727v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "PTC-Bias: Phoneme-Level Temporal Competition for Bias Retrieval and Post-Decoding Correction in Speech LLMs", "item_type": "entry", "summary": "added: PTC-Bias: Phoneme-Level Temporal Competition for Bias Retrieval and Post-Decoding Correction in Speech LLMs", "after": { "title": "PTC-Bias: Phoneme-Level Temporal Competition for Bias Retrieval and Post-Decoding Correction in Speech LLMs", "link": "https://arxiv.org/abs/2609.28727", "id": "oai:arXiv.org:2609.28727v1", "author": "Zhiqi Ai, Han Cheng, Shiyi Mu, Yongjin Zhou, Shugong Xu", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Contextual biasing improves rare-word recognition in speech large language models (SpeechLLMs), but efficiently exploiting large bias lists remains challenging. We propose PTC-Bias, a two-stage framework based on phoneme-level temporal competition. At the prefill stage, PTC Retrieval performs frame-synchronous phoneme decoding and temporal competition among candidate pronunciations, producing a compact bias-word shortlist and corresponding speech intervals. After SpeechLLM decoding, PTC Correction conducts a second local competition between the retrieved candidates and mismatched transcript spans within these intervals. Selective correction reduces near-homophone and word-segmentation errors while preserving correct transcriptions. Both stages share the same phoneme posteriors and require no additional SpeechLLM forward pass. Experiments on LibriSpeech show consistent gains across two SpeechLLMs and bias lists of up to 2000 words. With Prompt-SLAM-ASR-7B and 2000 bias words, PTC-Bias reduces B-WER by 23.4%/23.9% relative to CTC-Filter on test-clean/test-other, while keeping U-WER nearly unchanged.", "categories": [ "cs.CL", "cs.SD" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.28727", "tags": [ "cs.CL", "cs.SD" ] }, { "id": "ca43a263026f6f26064f", "title": "Script Choice in LLMs: Evidence for Late-Layer Commitment", "content_text": "In this paper, we investigate how script knowledge is distributed across the layers of LLMs using two complementary interpretability methods: logistic regression probing and logit-lens analysis. Our probing experiments reveal a clear asymmetry: both the input script and the instructed output script are encoded in the earliest layers of the network, while, in contrast, commitment to the actual output script emerges only in the final layers, with the model's intermediate representations…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "ca43a263026f6f26064f", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.28784v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "Script Choice in LLMs: Evidence for Late-Layer Commitment", "item_type": "entry", "summary": "added: Script Choice in LLMs: Evidence for Late-Layer Commitment", "after": { "title": "Script Choice in LLMs: Evidence for Late-Layer Commitment", "link": "https://arxiv.org/abs/2609.28784", "id": "oai:arXiv.org:2609.28784v1", "author": "David Kletz, Sandra Mitrovi\\'c, Itay Sabato, Ljiljana Dolami\\'c, Fabio Rinaldi", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "In this paper, we investigate how script knowledge is distributed across the layers of LLMs using two complementary interpretability methods: logistic regression probing and logit-lens analysis. Our probing experiments reveal a clear asymmetry: both the input script and the instructed output script are encoded in the earliest layers of the network, while, in contrast, commitment to the actual output script emerges only in the final layers, with the model's intermediate representations defaulting to Latin throughout most of the layers. This two-stage process is confirmed by logit-lens analyses, which show that script commitment consistently occurs at the very last layers of the LLMs. Together with the weaker script-following performance observed in smaller models, these results form a converging body of evidence linking script commitment to model depth, with broader implications for the design of sufficiently deep, inclusive multilingual architectures.", "categories": [ "cs.CL" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.28784", "tags": [ "cs.CL" ] }, { "id": "4ad02298d0f467091f4a", "title": "Polite but Misaligned: Evaluating LLM Politeness Judgments Against Human Pragmatic Norms", "content_text": "Despite strong performance on standard benchmarks, it remains unclear whether large language models (LLMs) evaluate social pragmatics in ways that align with human judgments. We evaluate LLM politeness judgments using two English-language datasets with complementary annotation formats: continuous human ratings and three-way categorical labels. Across the seven evaluated models, we find that inter-model agreement is stronger than model--human agreement. Strategy-level analyses suggest that…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "4ad02298d0f467091f4a", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29001v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "Polite but Misaligned: Evaluating LLM Politeness Judgments Against Human Pragmatic Norms", "item_type": "entry", "summary": "added: Polite but Misaligned: Evaluating LLM Politeness Judgments Against Human Pragmatic Norms", "after": { "title": "Polite but Misaligned: Evaluating LLM Politeness Judgments Against Human Pragmatic Norms", "link": "https://arxiv.org/abs/2609.29001", "id": "oai:arXiv.org:2609.29001v1", "author": "Rong Wang, Kun Sun, Yadong Guo", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Despite strong performance on standard benchmarks, it remains unclear whether large language models (LLMs) evaluate social pragmatics in ways that align with human judgments. We evaluate LLM politeness judgments using two English-language datasets with complementary annotation formats: continuous human ratings and three-way categorical labels. Across the seven evaluated models, we find that inter-model agreement is stronger than model--human agreement. Strategy-level analyses suggest that model--human alignment is associated with explicit linguistic cues, while some rapport-building strategies occur more frequently in misaligned cases. In the categorical task, model predictions exhibit systematic neutral compression, characterized by the overproduction of Neutral labels and the underprediction of Impolite labels. This pattern persists when expert consensus is used as the reference on a diagnostic subset. Our findings highlight the need for pragmatic evaluations that go beyond aggregate agreement metrics by examining directional patterns of model--human disagreement across different human references.", "categories": [ "cs.CL" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29001", "tags": [ "cs.CL" ] }, { "id": "8b8ed3a5384d008e0167", "title": "ELF-REG: Scaling Continuous Diffusion Language Models to Reasoning Tasks", "content_text": "Fully continuous diffusion language models (dLMs) denoise continuous representations without intermediate discretization, then decode all response tokens in parallel at the final step. Their performance on challenging reasoning tasks remains less established than that of autoregressive (AR) LLMs and masked dLMs. We scale Embedded Language Flows (ELF) to mathematical reasoning and code generation on GSM8K, MATH-500, HumanEval, and MBPP. We introduce ELF-REG, which improves learning with…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "8b8ed3a5384d008e0167", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29102v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "ELF-REG: Scaling Continuous Diffusion Language Models to Reasoning Tasks", "item_type": "entry", "summary": "added: ELF-REG: Scaling Continuous Diffusion Language Models to Reasoning Tasks", "after": { "title": "ELF-REG: Scaling Continuous Diffusion Language Models to Reasoning Tasks", "link": "https://arxiv.org/abs/2609.29102", "id": "oai:arXiv.org:2609.29102v1", "author": "Zeyu Michael Li, William Xingxu Chen, Bingshuo Qian, Jiayin Liu, Xiang Cheng", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Fully continuous diffusion language models (dLMs) denoise continuous representations without intermediate discretization, then decode all response tokens in parallel at the final step. Their performance on challenging reasoning tasks remains less established than that of autoregressive (AR) LLMs and masked dLMs. We scale Embedded Language Flows (ELF) to mathematical reasoning and code generation on GSM8K, MATH-500, HumanEval, and MBPP. We introduce ELF-REG, which improves learning with representation alignment and entanglement (REPA+REG), where a frozen AR teacher supervises intermediate denoiser features and supplies a global representation that is jointly denoised with the response. ELF-REG-L achieves 55.96% pass@1 on GSM8K at 64 network function evaluations (NFE), and 13.39% on MATH-500 and 22.56% on HumanEval at 128 NFE. It outperforms the evaluated comparable-scale dLMs in pass@1 on GSM8K and code, and improves MATH-500 pass@1 from 10.55% for the ELF-L baseline to 13.39% with ELF-REG-L. Without few-step training, the same task-specific checkpoints support strong low-NFE performance through early-stop, which decodes an intermediate clean prediction without completing the denoising trajectory. At 16 NFE, ELF-REG-L reaches 41.21% HumanEval pass@10, outperforming recent continuous dLMs of comparable scale.", "categories": [ "cs.CL", "cs.LG" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29102", "tags": [ "cs.CL", "cs.LG" ] }, { "id": "1452f25f141d74268c2d", "title": "Reasoning Instructions Can Break Answer Decoding in Vision--Language Models", "content_text": "Chain-of-thought (CoT) instructions can distort multiple-choice VLM evaluation when a scorer appends a reasoning cue but reads answer-label logits before the model generates any rationale. We call this CoT-prefix scoring. On ScienceQA, Qwen2.5-VL-7B drops from 80.76% to 45.48%, and across five option-content permutations 93.54% of CoT-prefix predictions select the first slot. Condition-matched linear probes recover 78.94% from the same hidden states, while free generation restores 75.24%…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "1452f25f141d74268c2d", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29278v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "Reasoning Instructions Can Break Answer Decoding in Vision--Language Models", "item_type": "entry", "summary": "added: Reasoning Instructions Can Break Answer Decoding in Vision--Language Models", "after": { "title": "Reasoning Instructions Can Break Answer Decoding in Vision--Language Models", "link": "https://arxiv.org/abs/2609.29278", "id": "oai:arXiv.org:2609.29278v1", "author": "Zeyan Li, Siyuan Qiu, Jianfeng Xu", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Chain-of-thought (CoT) instructions can distort multiple-choice VLM evaluation when a scorer appends a reasoning cue but reads answer-label logits before the model generates any rationale. We call this CoT-prefix scoring. On ScienceQA, Qwen2.5-VL-7B drops from 80.76% to 45.48%, and across five option-content permutations 93.54% of CoT-prefix predictions select the first slot. Condition-matched linear probes recover 78.94% from the same hidden states, while free generation restores 75.24%, showing that the answer often survives the prefix and the immediate readout fails. Vocabulary and layer diagnostics explain the mismatch: probability mass moves toward continuation tokens, while answer information remains linearly accessible in late layers. The effect recurs with varying severity across datasets and models, though not universally. These results show that CoT-prefix scoring can confound model knowledge with an evaluation-interface mismatch and should be avoided unless the requested and scored output events are aligned.", "categories": [ "cs.CL", "cs.AI" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29278", "tags": [ "cs.CL", "cs.AI" ] }, { "id": "77386e4efae589b979b3", "title": "Grammatical \"grandmother neurons\" are rare in LLMs", "content_text": "Understanding how Large Language Models (LLMs) encode linguistic structures remains a fundamental challenge in interpretability research. While diagnostic classifiers (or \"probes\") are widely used for this task, they face significant methodological criticism: training auxiliary classifiers introduces capacity confounds and calibration issues, often making it difficult to distinguish the model's intrinsic representations from the probe's ability to learn the task. To address these limitations…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "77386e4efae589b979b3", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29328v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "Grammatical \"grandmother neurons\" are rare in LLMs", "item_type": "entry", "summary": "added: Grammatical \"grandmother neurons\" are rare in LLMs", "after": { "title": "Grammatical \"grandmother neurons\" are rare in LLMs", "link": "https://arxiv.org/abs/2609.29328", "id": "oai:arXiv.org:2609.29328v1", "author": "Linyang He, Nima Mesgarani", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Understanding how Large Language Models (LLMs) encode linguistic structures remains a fundamental challenge in interpretability research. While diagnostic classifiers (or \"probes\") are widely used for this task, they face significant methodological criticism: training auxiliary classifiers introduces capacity confounds and calibration issues, often making it difficult to distinguish the model's intrinsic representations from the probe's ability to learn the task. To address these limitations, we introduce a probe-free framework for localizing linguistic selectivity at the individual neuron level. Leveraging the controlled contrasts of linguistic minimal pairs, we propose a Neuron Separability Index (NSI), a metric that directly quantifies how reliably single neurons differentiate grammatical from ungrammatical constructions without parameter updates. Applying NSI across 68 linguistic paradigms and seven checkpoints reveals three main patterns: 1) raw separability reaches near-peak levels earlier for morphological and syntactic distinctions than for syntax-semantics interface and conceptual distinctions. 2) after permutation normalization, single-unit selectivity is sparse, weak, and narrowly tuned: only a small fraction of units are sensitive to an average paradigm, and strongly selective \"grandmother neurons\" are rare. 3) whole-vector linear separability, single-neuron selectivity, and behavioral competence are largely dissociated, and targeted ablations further separate activation selectivity from causal reliance.", "categories": [ "cs.CL" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29328", "tags": [ "cs.CL" ] }, { "id": "9ebd98781f9c0eb27167", "title": "Where LLM Graders Succeed and Break: Evidence from Two Computer-Science Exams", "content_text": "One long-form exam in a large course costs hundreds of grader-hours, and qualified graders are scarce; LLM graders are a tempting alternative. To show its pitfalls we grade a practical Computer Vision exam ($570$ dual-graded students) under $171$ configurations spanning closed and open-weights models; the best reaches mean absolute error $1.64/35$, below the $2.61/35$ two human graders achieve against each other. The catch is the prompt: a short ''strict grader'' preamble drives $14$ of $17$…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "9ebd98781f9c0eb27167", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29333v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "Where LLM Graders Succeed and Break: Evidence from Two Computer-Science Exams", "item_type": "entry", "summary": "added: Where LLM Graders Succeed and Break: Evidence from Two Computer-Science Exams", "after": { "title": "Where LLM Graders Succeed and Break: Evidence from Two Computer-Science Exams", "link": "https://arxiv.org/abs/2609.29333", "id": "oai:arXiv.org:2609.29333v1", "author": "Ali Habibullah, Yazan Alshoibi, Mohammad Alshiekh, Salman Khan, Naeemullah Khan", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "One long-form exam in a large course costs hundreds of grader-hours, and qualified graders are scarce; LLM graders are a tempting alternative. To show its pitfalls we grade a practical Computer Vision exam ($570$ dual-graded students) under $171$ configurations spanning closed and open-weights models; the best reaches mean absolute error $1.64/35$, below the $2.61/35$ two human graders achieve against each other. The catch is the prompt: a short ''strict grader'' preamble drives $14$ of $17$ open-weights models out of the graded band ($\\text{MAE} \\ge 8$), three stopping grading altogether. The damage traces to the preamble's two credit-withholding sentences, not to tone or model scale; one of them, ''never give partial credit'', alone makes two of three probed models stop grading. The closed flagships of three vendors shift calibration under it but stay in the band. In $162$ further configurations on a second, independent Machine Learning exam from another course ($1{,}038$ dual-graded students), the preamble worsens ten models, moving three out of the band into collapse and one into refusal, yet improves seven whose neutral prompts over-mark: the vulnerability replicates, but its direction is exam-specific. Light LoRA fine-tuning repairs it: one adapter on the two exams' pooled $\\sim 3{,}900$ graded examples brings five small open models to parity or better with a human grader in agreement with the grader pair, and sensitivity to the three harsh personas nearly vanishes ($\\le 0.32$ MAE). We release the anonymised dataset, full ablation grid, and grading, fine-tuning and analysis pipelines.", "categories": [ "cs.CL", "cs.AI", "cs.CY" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29333", "tags": [ "cs.CL", "cs.AI", "cs.CY" ] }, { "id": "8eef57a645e2447201fe", "title": "ArGuard Shared Task: Harmful Content Detection in Arabic Memes and LLM Prompts", "content_text": "ArGuard is a shared task on harmful content detection in Arabic memes and LLM prompts. It includes two tracks: Track A focuses on multimodal hate detection in Arabic memes, while Track B addresses harmful prompt detection for Arabic LLM safety evaluation. In total, 58 teams registered, 35 participated in the final evaluation, and 27 submitted system-description papers. Participating teams explored models such as AraBERT, Jais, and Qwen3-VL. The best systems achieved macro-F1 scores of 0.823 on…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "8eef57a645e2447201fe", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29349v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "ArGuard Shared Task: Harmful Content Detection in Arabic Memes and LLM Prompts", "item_type": "entry", "summary": "added: ArGuard Shared Task: Harmful Content Detection in Arabic Memes and LLM Prompts", "after": { "title": "ArGuard Shared Task: Harmful Content Detection in Arabic Memes and LLM Prompts", "link": "https://arxiv.org/abs/2609.29349", "id": "oai:arXiv.org:2609.29349v1", "author": "Firoj Alam, Md. Rafiul Biswas, Mohamed Bayan Kmainasi, Ali Ezzat Shahroor, Hamdy Mubarak, George Mikros, Abul Hasnat, Wajdi Zaghouani", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "ArGuard is a shared task on harmful content detection in Arabic memes and LLM prompts. It includes two tracks: Track A focuses on multimodal hate detection in Arabic memes, while Track B addresses harmful prompt detection for Arabic LLM safety evaluation. In total, 58 teams registered, 35 participated in the final evaluation, and 27 submitted system-description papers. Participating teams explored models such as AraBERT, Jais, and Qwen3-VL. The best systems achieved macro-F1 scores of 0.823 on A1, 0.419 on A2, 0.984 on B1, and 0.790 on B2. Fine-grained meme classification in A2 was the most challenging setting, partly due to sparse labels and train-test distribution shifts.", "categories": [ "cs.CL", "cs.AI" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29349", "tags": [ "cs.CL", "cs.AI" ] }, { "id": "f74f50a99cd276c69d2e", "title": "From Policy Documents to Structured Survey Responses: Evaluating Large Language Models for Policy Monitoring", "content_text": "Science, technology, and innovation policies are crucial for competitiveness, yet their diversity and scale make them difficult to map and monitor consistently. Existing approaches rely heavily on manual survey efforts, which are costly and challenging to scale across countries. Large language models (LLMs) enable new possibilities for extracting and structuring information from long and unstructured policy documents. This paper presents an application of LLMs as \"AI respondents\" for generating…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "f74f50a99cd276c69d2e", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29370v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "From Policy Documents to Structured Survey Responses: Evaluating Large Language Models for Policy Monitoring", "item_type": "entry", "summary": "added: From Policy Documents to Structured Survey Responses: Evaluating Large Language Models for Policy Monitoring", "after": { "title": "From Policy Documents to Structured Survey Responses: Evaluating Large Language Models for Policy Monitoring", "link": "https://arxiv.org/abs/2609.29370", "id": "oai:arXiv.org:2609.29370v1", "author": "Carolyn Cole, Matthias Deschryvere, Toqeer Ehsan, Arash Hajikhani", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Science, technology, and innovation policies are crucial for competitiveness, yet their diversity and scale make them difficult to map and monitor consistently. Existing approaches rely heavily on manual survey efforts, which are costly and challenging to scale across countries. Large language models (LLMs) enable new possibilities for extracting and structuring information from long and unstructured policy documents. This paper presents an application of LLMs as \"AI respondents\" for generating structured survey responses from policy texts. We develop a data extraction pipeline based on long-context in-context learning to map information from public web sources into predefined survey categories, including policy instruments, target groups, and thematic areas. The pipeline integrates a validation step using a secondary LLM to assess relevance and evidence, alongside comparisons with human-provided responses. Using a multi-country dataset, we evaluate the alignment between LLM-generated and human-generated outputs through overlap measures and cross-validation. Results show that LLMs achieve high agreement for structured indicators (84-95%), while differences remain in free-text fields, where models tend to provide more detailed procedural descriptions. These findings highlight the potential of hybrid human-AI workflows for policy monitoring, improving both efficiency and scalability while maintaining the need for human validation and contextual interpretation.", "categories": [ "cs.CL", "cs.AI" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29370", "tags": [ "cs.CL", "cs.AI" ] }, { "id": "0a04a6b42332a9da84c2", "title": "Likelihood Ranking doesn't Scale Like Prompting in LLMs", "content_text": "LLM evaluation is commonly performed either by prompting models to produce answers or by scoring candidate outputs with likelihood-based metrics. In multiple-choice QA, however, standard likelihood-based scoring is still conditioned on the question and answer set, and can therefore leverage the same task-conditioned answer-selection interface used in prompting. We study a complementary protocol based on likelihood ranking of declarative statements constructed from the same question--answer…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "0a04a6b42332a9da84c2", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29390v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "Likelihood Ranking doesn't Scale Like Prompting in LLMs", "item_type": "entry", "summary": "added: Likelihood Ranking doesn't Scale Like Prompting in LLMs", "after": { "title": "Likelihood Ranking doesn't Scale Like Prompting in LLMs", "link": "https://arxiv.org/abs/2609.29390", "id": "oai:arXiv.org:2609.29390v1", "author": "Alessandro Bondielli, Lucia Passaro, Davide Bacciu, Alessandro Lenci", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "LLM evaluation is commonly performed either by prompting models to produce answers or by scoring candidate outputs with likelihood-based metrics. In multiple-choice QA, however, standard likelihood-based scoring is still conditioned on the question and answer set, and can therefore leverage the same task-conditioned answer-selection interface used in prompting. We study a complementary protocol based on likelihood ranking of declarative statements constructed from the same question--answer pairs. Across 95 decoder-only models, ranging from 0.1B to 104B parameters, and 10 MCQA datasets, we find a systematic divergence between declarative-statement likelihood ranking and prompted answering. Statement-likelihood accuracy remains comparatively stable across scale, whereas prompted answering improves sharply with scale and instruction-tuning. These results suggest that likelihood preferences over controlled declarative alternatives and task-conditioned answer selection probe distinct aspects of model behavior, and should not be treated as interchangeable.", "categories": [ "cs.CL" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29390", "tags": [ "cs.CL" ] }, { "id": "a43be67b70b5d4940b56", "title": "Baseline Shape Decides the Verdict: A Controlled Re-Examination of Ternary Language Models at 60K Parameters", "content_text": "Ternary (1.58-bit) weights are attractive for microcontroller-class language models, but the sub-1M-parameter regime rests mainly on isolated, single-seed comparisons. One prominent example reports that a routed ternary block (convolution, diagonal SSM and sparse attention mixed by a per-token router) beats a parameter-matched full-precision transformer by 22% at 60K parameters, attributing this to inductive bias. We re-run it under one fixed recipe, three seeds per cell, 98 byte-level runs on…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "a43be67b70b5d4940b56", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29397v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "Baseline Shape Decides the Verdict: A Controlled Re-Examination of Ternary Language Models at 60K Parameters", "item_type": "entry", "summary": "added: Baseline Shape Decides the Verdict: A Controlled Re-Examination of Ternary Language Models at 60K Parameters", "after": { "title": "Baseline Shape Decides the Verdict: A Controlled Re-Examination of Ternary Language Models at 60K Parameters", "link": "https://arxiv.org/abs/2609.29397", "id": "oai:arXiv.org:2609.29397v1", "author": "Gautam Veldanda", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Ternary (1.58-bit) weights are attractive for microcontroller-class language models, but the sub-1M-parameter regime rests mainly on isolated, single-seed comparisons. One prominent example reports that a routed ternary block (convolution, diagonal SSM and sparse attention mixed by a per-token router) beats a parameter-matched full-precision transformer by 22% at 60K parameters, attributing this to inductive bias. We re-run it under one fixed recipe, three seeds per cell, 98 byte-level runs on one laptop. (i) Baseline shape dominates: at a 16M-byte budget, param-matched transformers span 22.6% in validation loss purely by depth/width choice - far more than any architecture effect we measure there - and the best-shaped transformer ties the routed model, so the published margin is at least partly a baseline-shape effect; the ordering of shapes reverses with budget, so no single fixed shape can be trusted. (ii) At 130M bytes the routed model does win, by 22.2-24.0% over the three transformer shapes we evaluate there - but a plain gated diagonal-SSM block beats it by a further 9.1%, and the routed model's own router puts most of its weight on its recurrent pathway, so the gain does not require routing. (iii) The ternary penalty differs by architecture at the larger budget (+5.3% best transformer vs. +19.5% routed, +28.1% gated SSM), but we cannot attribute that to architecture alone: our transformers keep learned positional embeddings in full precision, 11-22% of their parameters, so they are less quantized than the models they are compared with. (iv) A 90/10 full-precision-then-ternary schedule beats all-ternary training, but only at a stage-2 learning rate about 10x the pretraining peak; at a conventional fine-tuning rate it looks 15.3% worse, reversing the conclusion. The from-scratch baseline was not itself learning-rate tuned, which bounds (iii) and (iv). Code and run logs released.", "categories": [ "cs.CL" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29397", "tags": [ "cs.CL" ] }, { "id": "f1abf2f931608d63d86e", "title": "Large Language Models for Programming: Actually Fixing or Reimplementing Incorrect Code?", "content_text": "Recent studies have shown that Large Language Models can effectively solve problems and fix bugs in diverse programming environments, including competitive programming. Existing approaches primarily evaluate LLM performance in problem solving or bug fixing independently, but do not explore the relationship between these two capabilities. This work focuses on determining how much the LLM deviates from a buggy solution to fix the bug compared to a human-written patch, and if there is a bias…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "f1abf2f931608d63d86e", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29410v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "Large Language Models for Programming: Actually Fixing or Reimplementing Incorrect Code?", "item_type": "entry", "summary": "added: Large Language Models for Programming: Actually Fixing or Reimplementing Incorrect Code?", "after": { "title": "Large Language Models for Programming: Actually Fixing or Reimplementing Incorrect Code?", "link": "https://arxiv.org/abs/2609.29410", "id": "oai:arXiv.org:2609.29410v1", "author": "Alexandru Stefan Stoica, Traian Rebedea, Marian Cristian Mihaescu", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Recent studies have shown that Large Language Models can effectively solve problems and fix bugs in diverse programming environments, including competitive programming. Existing approaches primarily evaluate LLM performance in problem solving or bug fixing independently, but do not explore the relationship between these two capabilities. This work focuses on determining how much the LLM deviates from a buggy solution to fix the bug compared to a human-written patch, and if there is a bias towards generating entirely new solutions. We construct a dataset with all the submissions ($\\sim$ 3000) from a couple of users from Codeforces, and we match each buggy submission with its corresponding human fix. By using the similarity between the buggy solution and the human fix as a baseline, we evaluate the quality of LLM-generated bug fixes on 3 OpenAI GPT models (gpt-5-nano, gpt-5-mini, gpt-5.1). We check if the generated solutions solve the problem by using the Codeforces-R1 dataset, an openly available dataset that has tests generated with the DeepSeek-R1 model. Our findings suggest that LLMs tend to modify more lines than necessary compared to human fixes and, in some cases, generate entirely new solutions. We also observe that LLMs solve more problems correctly when allowed to generate solutions from scratch rather than patch buggy submissions, even when those submissions are close to the human patch. This has important implications for the design of AI-assisted programming tools, particularly in supporting user debugging processes and promoting incremental problem-solving strategies rather than solution replacement.", "categories": [ "cs.CL", "cs.SE" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29410", "tags": [ "cs.CL", "cs.SE" ] }, { "id": "49a5a44cee8f7738b3a8", "title": "Rufus-Air: An Open LLM Post-Training Recipe", "content_text": "Rufus-Air is an open and reproducible post-training recipe on GLM-4.5-Air-Base (106B-A12B), organized as a serial pipeline of eight stages: SFT, Reasoning RL, Coding RL, Instruction-Following RL, General Agent, Coding Agent, Search Agent, and RLHF. We document the data, reward design, infrastructure, stage order, and stagewise results needed to reproduce the recipe. Stages progress from basic to advanced capabilities and from hard, verifiable rewards to softer judge-based signals. Training…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "49a5a44cee8f7738b3a8", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29421v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "Rufus-Air: An Open LLM Post-Training Recipe", "item_type": "entry", "summary": "added: Rufus-Air: An Open LLM Post-Training Recipe", "after": { "title": "Rufus-Air: An Open LLM Post-Training Recipe", "link": "https://arxiv.org/abs/2609.29421", "id": "oai:arXiv.org:2609.29421v1", "author": "Chia-Yuan Chang, Renyuan Cheng, Rui Feng, Xiaotian Han, Yuan He, Hongye Jin, Linwei Li, Shiyang Li, Fenglin Liu, Xin Liu, Priyanka Nigam, Haoyang Wen, Zhenghao Xu, Zhuocheng Xu, Bing Yin, Qingyu Yin, Chao Zhang, Rongzhi Zhang, Zhihan Zhang, Zixuan Zhang, Zixuan Zhang, Tuo Zhao", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Rufus-Air is an open and reproducible post-training recipe on GLM-4.5-Air-Base (106B-A12B), organized as a serial pipeline of eight stages: SFT, Reasoning RL, Coding RL, Instruction-Following RL, General Agent, Coding Agent, Search Agent, and RLHF. We document the data, reward design, infrastructure, stage order, and stagewise results needed to reproduce the recipe. Stages progress from basic to advanced capabilities and from hard, verifiable rewards to softer judge-based signals. Training builds on open-source components and public data, much of it used as released, without new human annotation or an in-house distillation teacher. Our main findings are that (i) diverse, high-quality SFT establishes a strong capability floor; (ii) difficulty filtering keeps RL prompts within a productive learning range; (iii) reward reliability provides a practical principle for ordering stages; and (iv) infrastructure and engineering choices are part of the recipe, not just an implementation detail. Rufus-Air improves over the official GLM-4.5-Air post-trained release and is competitive with similarly sized open models.", "categories": [ "cs.CL", "cs.AI", "cs.LG" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29421", "tags": [ "cs.CL", "cs.AI", "cs.LG" ] }, { "id": "a7e5ce953f00c842d0ff", "title": "agentic-ger: terminology recovery in long-form speech using global context", "content_text": "Recent advances in speech language models have improved automatic speech recognition (ASR) for long-form audio. However, accurately and consistently transcribing domain-specific terminology remains challenging. Motivated by the world knowledge and contextual capability of large language models (LLMs), we propose Agentic-GER, an LLM-based agent for terminology correction in long-form speech. The agent uses global context from the full transcript to identify suspicious terms and resolve ambiguous…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "a7e5ce953f00c842d0ff", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29428v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "agentic-ger: terminology recovery in long-form speech using global context", "item_type": "entry", "summary": "added: agentic-ger: terminology recovery in long-form speech using global context", "after": { "title": "agentic-ger: terminology recovery in long-form speech using global context", "link": "https://arxiv.org/abs/2609.29428", "id": "oai:arXiv.org:2609.29428v1", "author": "Yanqiao Zhu, Wupeng Wang, Zhifu Gao, Xiangang Li, Xie Chen", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Recent advances in speech language models have improved automatic speech recognition (ASR) for long-form audio. However, accurately and consistently transcribing domain-specific terminology remains challenging. Motivated by the world knowledge and contextual capability of large language models (LLMs), we propose Agentic-GER, an LLM-based agent for terminology correction in long-form speech. The agent uses global context from the full transcript to identify suspicious terms and resolve ambiguous hypotheses. It selectively re-transcribes the source speech to check candidate corrections, and uses accepted edits to guide subsequent decisions. Experiments with four LLMs and two ASR systems on GigaSpeechBench show consistent terminology improvements in both Chinese and English, with and without thinking. On Chinese speech, Agentic-GER achieves up to a 36.8% relative reduction in biased character error rate (B-CER) over the Whisper baseline.", "categories": [ "cs.CL", "cs.AI" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29428", "tags": [ "cs.CL", "cs.AI" ] }, { "id": "6b90b68c4822ebf8a104", "title": "IterSynth: Rethinking Deep Search Agents via Role-Decoupled Iterative Synthesis", "content_text": "Deep search requires LLM agents to decompose complex queries, search for evidence, and synthesize grounded answers, yet existing ReAct-style agents suffer from two limitations: role coupling, where one policy must handle planning, evidence use, and synthesis; and context accumulation, where growing search histories introduce noise and obscure useful information. To address these issues, we propose IterSynth, a role-decoupled and summary-based paradigm that alternates between a Planner for…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "6b90b68c4822ebf8a104", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29444v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "IterSynth: Rethinking Deep Search Agents via Role-Decoupled Iterative Synthesis", "item_type": "entry", "summary": "added: IterSynth: Rethinking Deep Search Agents via Role-Decoupled Iterative Synthesis", "after": { "title": "IterSynth: Rethinking Deep Search Agents via Role-Decoupled Iterative Synthesis", "link": "https://arxiv.org/abs/2609.29444", "id": "oai:arXiv.org:2609.29444v1", "author": "Xingyu Wu, Yuchen Yan, Zhengxi Lu, Siqi Chen, Xin ZHANG, Aiting Liu, Chao Deng, Jie Liu, Jin Ma, Jian Shao, Jun Xiao, Yongliang Shen", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Deep search requires LLM agents to decompose complex queries, search for evidence, and synthesize grounded answers, yet existing ReAct-style agents suffer from two limitations: role coupling, where one policy must handle planning, evidence use, and synthesis; and context accumulation, where growing search histories introduce noise and obscure useful information. To address these issues, we propose IterSynth, a role-decoupled and summary-based paradigm that alternates between a Planner for identifying information needs and a Synthesizer for integrating evidence into an evolving summary state. This design separates planning from synthesis while using the summary as the persistent state of search, reducing both capability coupling and context noise. To train IterSynth effectively, we further introduce Role-Decoupled Policy Optimization (RDPO) for reinforcement learning, which combines terminal outcome rewards with turn-level rubric evaluations and computes role-specific advantages for more precise credit assignment. Experiments on five long-horizon deep-search benchmarks such as BrowseComp and Xbench-DS show that IterSynth-8B achieves an average score of 50.7, surpassing the strongest prior $\\leq$8B agent by +4.2\\%. Moreover, IterSynth serves as a model-agnostic prompting paradigm, delivering substantial zero-shot gains over ReAct and similar prompting paradigms on frontier proprietary models.", "categories": [ "cs.CL", "cs.AI" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29444", "tags": [ "cs.CL", "cs.AI" ] }, { "id": "8d83e5431bfaeaedd046", "title": "Evaluating Explanation-Driven Vision-Language Reasoning via Generation Order Interventions", "content_text": "Natural language explanation generation serves as a key mechanism for exposing and evaluating vision-language reasoning. Prior work on explanation-driven vision-language models predominantly follows a post-hoc (answer-first) paradigm, implicitly suggesting that supervised rationales can reflect underlying reasoning processes. In contrast, modern large vision-language models increasingly exhibit a rationale-first generation tendency, which more closely aligns with structured, stepwise reasoning…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "8d83e5431bfaeaedd046", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29496v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "Evaluating Explanation-Driven Vision-Language Reasoning via Generation Order Interventions", "item_type": "entry", "summary": "added: Evaluating Explanation-Driven Vision-Language Reasoning via Generation Order Interventions", "after": { "title": "Evaluating Explanation-Driven Vision-Language Reasoning via Generation Order Interventions", "link": "https://arxiv.org/abs/2609.29496", "id": "oai:arXiv.org:2609.29496v1", "author": "Siting Liang, Luca Rippe, Omar Adjali, Daniel Sonntag", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Natural language explanation generation serves as a key mechanism for exposing and evaluating vision-language reasoning. Prior work on explanation-driven vision-language models predominantly follows a post-hoc (answer-first) paradigm, implicitly suggesting that supervised rationales can reflect underlying reasoning processes. In contrast, modern large vision-language models increasingly exhibit a rationale-first generation tendency, which more closely aligns with structured, stepwise reasoning. In this work, we systematically evaluate whether explanations are causally tied to model predictions within a single generation step under a controlled experimental setup, explicitly eliminating unnecessary chain-of-thought or other intermediate reasoning processes across knowledge-intensive QA, visual entailment, and compositional grounding benchmarks. We find that larger models emerge as a prerequisite for reliably supporting rationale-first reasoning at scale. However, answer-first generation is less prone to format-related errors in structured output. Overall, explanation ordering, model scale and pre-training knowledge, task-specific fine-tuning, and task structure jointly influence both prediction accuracy and reasoning faithfulness.", "categories": [ "cs.CL" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29496", "tags": [ "cs.CL" ] }, { "id": "c59d1ec5e953c5dff38e", "title": "PROOF: Profiling Reliability of Object-Level Facts in Large Language Models", "content_text": "Aggregate factuality scores hide where a language model succeeds, which relations it confuses, and whether an answer survives innocuous changes to the question or decoder. We introduce PROOF, a profile-oriented benchmark for factual coverage in instruction-tuned language models. PROOF converts a frozen Wikidata snapshot into 18,486 English multiple-choice questions grounded in 11,779 semantic facts, 101 classes, 392 properties, and 14 domains. Each question has an explicit \"I don't know\"…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "c59d1ec5e953c5dff38e", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29504v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "PROOF: Profiling Reliability of Object-Level Facts in Large Language Models", "item_type": "entry", "summary": "added: PROOF: Profiling Reliability of Object-Level Facts in Large Language Models", "after": { "title": "PROOF: Profiling Reliability of Object-Level Facts in Large Language Models", "link": "https://arxiv.org/abs/2609.29504", "id": "oai:arXiv.org:2609.29504v1", "author": "Andrei Chetvergov, Mikhail Solovev, Timofei Sivoraksha, Stepan Ukolov, Valeriia Kuschenko, Alexander Evseev, Sergey Bolovtsov", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Aggregate factuality scores hide where a language model succeeds, which relations it confuses, and whether an answer survives innocuous changes to the question or decoder. We introduce PROOF, a profile-oriented benchmark for factual coverage in instruction-tuned language models. PROOF converts a frozen Wikidata snapshot into 18,486 English multiple-choice questions grounded in 11,779 semantic facts, 101 classes, 392 properties, and 14 domains. Each question has an explicit \"I don't know\" option, a \"No correct option\" control, and nine controlled formulations; 1,849 questions are no-correct-option traps.\nWe evaluate 18 open-weight model deployments on 166,374 prompts each and separately perturb decoding on a fixed 10% subset. Base factual accuracy ranges from 6.58% to 57.59% (chance: 8.64%), yet every model has a 19.3-36.4 percentage-point spread across domains. Paired facts reveal direction-dependent retrieval, usually favoring subject-to-object queries, with the pattern reversing for one model. We find no consistent temporal penalty after exact-stratum adjustment.\nNeutral wording changes accuracy by as much as 26.5 percentage points, while adversarial formulations break up to 79.4% of answers that were initially correct. Direct switching to an injected false label varies from 0.04% to 27.5%, showing that accuracy loss and hint following are distinct. Selected-token confidence often indicates severe overconfidence, and decoder perturbations move accuracy by up to 15.7 percentage points and domain profiles by 16.8 points. PROOF therefore measures factual coverage as a structured, intervention-aware profile rather than a single claim about what a model \"believes.\"", "categories": [ "cs.CL", "cs.AI" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29504", "tags": [ "cs.CL", "cs.AI" ] }, { "id": "039f41838c4a16b04716", "title": "Benchmarking Arabic--Russian Machine Translation: A Comparison of Fine-tuned NMT and Few-shot LLMs under Rich Morphology and Low Lexical Overlap", "content_text": "Arabic-Russian machine translation (MT) remains under-explored due to the rich morphology of Arabic and low lexical overlap between the two languages. We benchmark seven fine-tuned neural machine translation (NMT) models against four few-shot large language models (LLMs) on a 20k/5k/5k split of a new 15.47M-pair corpus. Fine-tuned NLLB-1.3B achieves the highest BLEU (16.3) and COMET (0.738). Aya-Expanse 8B leads the few-shot LLMs (BLEU 1.7 on 500 sentences, chrF 25.7), but all LLM scores remain…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "039f41838c4a16b04716", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29559v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "Benchmarking Arabic--Russian Machine Translation: A Comparison of Fine-tuned NMT and Few-shot LLMs under Rich Morphology and Low Lexical Overlap", "item_type": "entry", "summary": "added: Benchmarking Arabic--Russian Machine Translation: A Comparison of Fine-tuned NMT and Few-shot LLMs under Rich Morphology and Low Lexical Overlap", "after": { "title": "Benchmarking Arabic--Russian Machine Translation: A Comparison of Fine-tuned NMT and Few-shot LLMs under Rich Morphology and Low Lexical Overlap", "link": "https://arxiv.org/abs/2609.29559", "id": "oai:arXiv.org:2609.29559v1", "author": "Mullosharaf K. Arabov", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Arabic-Russian machine translation (MT) remains under-explored due to the rich morphology of Arabic and low lexical overlap between the two languages. We benchmark seven fine-tuned neural machine translation (NMT) models against four few-shot large language models (LLMs) on a 20k/5k/5k split of a new 15.47M-pair corpus. Fine-tuned NLLB-1.3B achieves the highest BLEU (16.3) and COMET (0.738). Aya-Expanse 8B leads the few-shot LLMs (BLEU 1.7 on 500 sentences, chrF 25.7), but all LLM scores remain far below the fine-tuned NMT baselines. Error analysis identifies low lexical overlap as the dominant failure mode; among the worst translations, mT5-small produces 32% too-short outputs. Bootstrap tests confirm significant differences among most models. Our results demonstrate that fine-tuned NMT significantly outperforms few-shot LLMs for Arabic-Russian translation under low-resource conditions.", "categories": [ "cs.CL" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29559", "tags": [ "cs.CL" ] }, { "id": "3c5ca3c69e59e8558360", "title": "An Exploratory Ablation of a Small MLA--SSM Hybrid Language Model", "content_text": "We report an exploratory, single-seed ablation of TALH (Adaptive Latent Hybrid), a decoder-only language model with parallel Multi-head Latent Attention (MLA) and a custom recurrent state-space (SSM) branch. Five variants, spanning 117--217M estimated active parameters per token, are trained from scratch on a FineWeb sample for the same number of optimisation steps and tokens. In this specific setup, removing the SSM branch gives the largest degradation in validation perplexity (MLA-only PPL…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "3c5ca3c69e59e8558360", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29618v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "An Exploratory Ablation of a Small MLA--SSM Hybrid Language Model", "item_type": "entry", "summary": "added: An Exploratory Ablation of a Small MLA--SSM Hybrid Language Model", "after": { "title": "An Exploratory Ablation of a Small MLA--SSM Hybrid Language Model", "link": "https://arxiv.org/abs/2609.29618", "id": "oai:arXiv.org:2609.29618v1", "author": "Christos Koutsiaris", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "We report an exploratory, single-seed ablation of TALH (Adaptive Latent Hybrid), a decoder-only language model with parallel Multi-head Latent Attention (MLA) and a custom recurrent state-space (SSM) branch. Five variants, spanning 117--217M estimated active parameters per token, are trained from scratch on a FineWeb sample for the same number of optimisation steps and tokens. In this specific setup, removing the SSM branch gives the largest degradation in validation perplexity (MLA-only PPL 315), whereas removing MLA has a much smaller effect (SSM-only PPL 239). A dense-FFN hybrid obtains PPL 231, compared with 240 for the tested top-2 ternary-MoE hybrid, while using 3.87 GB less peak training memory. We also preserve a preliminary Apple M3 timing observation: among the five unoptimised implementations, MLA-only has the flattest measured time-to-first-token curve from 512 to 2,048 prompt tokens, although the dense Transformer is much faster in absolute terms. Because the runs are single-seed, parameter counts are unmatched, the evaluation stream may overlap the training source, and raw repeated timing records are unavailable, these results support implementation-specific hypotheses rather than general conclusions about MLA, SSMs, or mixture-of-experts models.", "categories": [ "cs.CL", "cs.AI", "cs.IR", "cs.LG" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29618", "tags": [ "cs.CL", "cs.AI", "cs.IR", "cs.LG" ] }, { "id": "f3bb459d7879467cff04", "title": "LLMersion: A Local-First AI Agent Framework for Low-Cost Home Language Learning toward Educational Equity", "content_text": "Artificial intelligence helps education most where an essential provision has been rationed by cost. For language learners that provision is a teacher's voice, which binds listening, reading, speaking, and writing into one act. Published evidence shows why most learners lack it, from a global shortage of 44 million teachers to heavy household tutoring bills, and why technology has not substituted for it: computer-assisted language learning proved effective but narrow, applications presuppose…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "f3bb459d7879467cff04", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29672v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "LLMersion: A Local-First AI Agent Framework for Low-Cost Home Language Learning toward Educational Equity", "item_type": "entry", "summary": "added: LLMersion: A Local-First AI Agent Framework for Low-Cost Home Language Learning toward Educational Equity", "after": { "title": "LLMersion: A Local-First AI Agent Framework for Low-Cost Home Language Learning toward Educational Equity", "link": "https://arxiv.org/abs/2609.29672", "id": "oai:arXiv.org:2609.29672v1", "author": "Qiming Guo, Jinwen Tang, Xingran Huang, Hung-Yu Lin, Yafu Zhong, Xiatian Zhuang", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Artificial intelligence helps education most where an essential provision has been rationed by cost. For language learners that provision is a teacher's voice, which binds listening, reading, speaking, and writing into one act. Published evidence shows why most learners lack it, from a global shortage of 44 million teachers to heavy household tutoring bills, and why technology has not substituted for it: computer-assisted language learning proved effective but narrow, applications presuppose connectivity 2.6 billion people lack, and One Laptop per Child's randomized evaluation found that hardware without capable software teaches nothing. We distill eight difficulties and four binding constraints, and argue that small open-weight models dissolve the last: a complete four-skill stack now fits a \\$200-class laptop and, on community measurements, generates at the pace speech is consumed, for about one US cent of electricity per study hour. We therefore propose LLMersion, a scheme for AI for education that runs entirely at home, over the learner's own documents, with an AI-written, AI-understood, AI-updated codebase anyone can customize; present LLMersion-1, a released open-source prototype (https://github.com/QM378/LLMersion); and outline the vision of a private learning agent.", "categories": [ "cs.CL", "cs.CY", "cs.HC" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29672", "tags": [ "cs.CL", "cs.CY", "cs.HC" ] }, { "id": "a2a8548059bdb6510e38", "title": "Stochastic Semantic Evidence Graphs: Uncertainty Propagation and Governance for Agentic AI", "content_text": "AI-agent evaluations usually inspect a final answer, yet error may enter through evidence, retrieval, prompting, generation or decision mapping. We introduce a stochastic semantic evidence graph (SSEG), a hierarchical stochastic DAG whose language node expands into an autoregressive token subgraph and whose observable output may be a law over complete phrases. Semantic reduction and calibration are optional. We define graph-relative local defects and downstream edge influences, derive a…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "a2a8548059bdb6510e38", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29703v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "Stochastic Semantic Evidence Graphs: Uncertainty Propagation and Governance for Agentic AI", "item_type": "entry", "summary": "added: Stochastic Semantic Evidence Graphs: Uncertainty Propagation and Governance for Agentic AI", "after": { "title": "Stochastic Semantic Evidence Graphs: Uncertainty Propagation and Governance for Agentic AI", "link": "https://arxiv.org/abs/2609.29703", "id": "oai:arXiv.org:2609.29703v1", "author": "Matthew Francis Dixon", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "AI-agent evaluations usually inspect a final answer, yet error may enter through evidence, retrieval, prompting, generation or decision mapping. We introduce a stochastic semantic evidence graph (SSEG), a hierarchical stochastic DAG whose language node expands into an autoregressive token subgraph and whose observable output may be a law over complete phrases. Semantic reduction and calibration are optional. We define graph-relative local defects and downstream edge influences, derive a pathwise bound on terminal error and use its nodewise terms to diagnose governance triggers. For source provenance, the graph preserves uncertain claim--passage relations and propagates sharp Fr\\'echet bounds rather than assuming independence across sources. Across three open-weight architectures, information-equivalent changes materially alter complete-phrase laws. A controlled experiment yields no certificate violations in 5,000 cases; crossed-RAG and live Brave-retrieval experiments separate retrieval, presentation, source and interaction effects. SSEG therefore turns workflow provenance into a quantitative account of where uncertainty entered, how it propagated and whether an output is qualified for use.", "categories": [ "cs.CL", "cs.IR", "stat.ML" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29703", "tags": [ "cs.CL", "cs.IR", "stat.ML" ] }, { "id": "46120b8fa58c689e66ff", "title": "PPTBench: Can Coding Agents Reconstruct the Visual World through Structured, Editable Slides", "content_text": "Coding agents are beginning to act in the visual world. They now build webpages, GUIs, games, 3D scenes, diagrams, and documents. Success in such visual coding requires bridging two spaces: inferring visual structure and expressing it programmatically. Slides are a core medium of knowledge work, widely used to communicate ideas and collaborate in a form that people can directly inspect and edit. Therefore, they provide an ideal testbed for visual coding, as they require agents to recover visual…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "46120b8fa58c689e66ff", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29718v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "PPTBench: Can Coding Agents Reconstruct the Visual World through Structured, Editable Slides", "item_type": "entry", "summary": "added: PPTBench: Can Coding Agents Reconstruct the Visual World through Structured, Editable Slides", "after": { "title": "PPTBench: Can Coding Agents Reconstruct the Visual World through Structured, Editable Slides", "link": "https://arxiv.org/abs/2609.29718", "id": "oai:arXiv.org:2609.29718v1", "author": "Xiaoqiu Wang, Yizhe Chi, Wenyi Li, Deyao Hong, Zhihan Shan, Mingju Gao, Kaisen Yang, Youjie Zheng, Calvin Xiao, Qinhuai Na", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Coding agents are beginning to act in the visual world. They now build webpages, GUIs, games, 3D scenes, diagrams, and documents. Success in such visual coding requires bridging two spaces: inferring visual structure and expressing it programmatically. Slides are a core medium of knowledge work, widely used to communicate ideas and collaborate in a form that people can directly inspect and edit. Therefore, they provide an ideal testbed for visual coding, as they require agents to recover visual structure and realize it as editable objects. However, existing benchmarks either rely on subjective open-ended evaluation, produce non-editable code outputs, or focus only on local editing rather than end-to-end visual reconstruction. We introduce PPTBench, which benchmarks visual coding through editable slide reconstruction. It contains 500 tasks, each based on a scientific flow diagram from a real arXiv paper and requiring agents to reconstruct it as a single PPTX page composed of native, editable objects. A four-stage Agentic Judge evaluates artifact validity, semantic correctness, rendering quality, and fine-grained visual quality. Across 31 configurations spanning model families, effort levels, and harnesses, the best configuration, Kimi K3, reaches only 67.80, while the median scores 19.47. We find that agents can reliably produce valid PPTX files but still struggle with semantic and visual correctness, especially text details. More reasoning mainly helps agents pass hard gates, while stronger verification is more consistently associated with higher quality. PPTBench advances the vision of coding agents that can understand and reconstruct the visual world through structured, editable code.", "categories": [ "cs.CL", "cs.AI" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29718", "tags": [ "cs.CL", "cs.AI" ] }, { "id": "fd2b7832b2b04b725157", "title": "JEV vs. LLMs as Rubric Judges: Cheaper, Faster, and Wrong in the Same Places", "content_text": "We ask whether Jev, a typed classifier that returns probabilities over permitted answers without generating text, can replace an LLM rubric judge. We compare it with three flash-tier LLM judges on nine panels drawn from seven benchmarks, giving every judge identical criterion texts. Jev's accuracy differs significantly from an LLM judge's in only 8 of 27 paired comparisons, ahead mostly on binary criteria and behind only on graded ones, and most of the other comparisons are inconclusive. Summed…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "fd2b7832b2b04b725157", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29769v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "JEV vs. LLMs as Rubric Judges: Cheaper, Faster, and Wrong in the Same Places", "item_type": "entry", "summary": "added: JEV vs. LLMs as Rubric Judges: Cheaper, Faster, and Wrong in the Same Places", "after": { "title": "JEV vs. LLMs as Rubric Judges: Cheaper, Faster, and Wrong in the Same Places", "link": "https://arxiv.org/abs/2609.29769", "id": "oai:arXiv.org:2609.29769v1", "author": "Delip Rao, Chris Callison-Burch", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "We ask whether Jev, a typed classifier that returns probabilities over permitted answers without generating text, can replace an LLM rubric judge. We compare it with three flash-tier LLM judges on nine panels drawn from seven benchmarks, giving every judge identical criterion texts. Jev's accuracy differs significantly from an LLM judge's in only 8 of 27 paired comparisons, ahead mostly on binary criteria and behind only on graded ones, and most of the other comparisons are inconclusive. Summed over the nine panels, the LLM judges, called once per criterion, cost 29 to 325 times as much as Jev and took 30 to 220 times as long. On graded criteria all four judges agree more with one another than with the labels and mostly assign lower levels than the raters. One of several observational accounts is that raters followed scale conventions our criterion texts omit. Jev's confidence ranks its own errors on most panels, which should make a cheap classifier the ideal first stage of a cascade that defers its uncertain verdicts to an LLM judge. Correlated errors undo that advantage. The LLM judges repeat nearly all of Jev's most confident errors, so a cascade replayed on the recorded verdicts lowers cost but gains at most 1.5 points over the best single judge with cross-fitted thresholds, and at most 2.0 even with oracle thresholds.", "categories": [ "cs.CL" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29769", "tags": [ "cs.CL" ] }, { "id": "7b911805692d5be73e72", "title": "CORDIAL: Calibrating Ordinal LLM Outputs from Few Labels", "content_text": "A large language model (LLM) can turn a text into a distribution over an ordered scale, but that distribution is a noisy measurement: saturated, compressed or exaggerated, and biased in a consistent direction. We propose CORDIAL, which treats the model's output as a noisy reading of the true label and corrects it with a channel of five interpretable parameters. The channel is small enough for its posterior to be averaged from a handful of labels, and we prove that the resulting calibration…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "7b911805692d5be73e72", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29807v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "CORDIAL: Calibrating Ordinal LLM Outputs from Few Labels", "item_type": "entry", "summary": "added: CORDIAL: Calibrating Ordinal LLM Outputs from Few Labels", "after": { "title": "CORDIAL: Calibrating Ordinal LLM Outputs from Few Labels", "link": "https://arxiv.org/abs/2609.29807", "id": "oai:arXiv.org:2609.29807v1", "author": "Xiangwei Wang, Peng Wang, Saman Halgamuge", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "A large language model (LLM) can turn a text into a distribution over an ordered scale, but that distribution is a noisy measurement: saturated, compressed or exaggerated, and biased in a consistent direction. We propose CORDIAL, which treats the model's output as a noisy reading of the true label and corrects it with a channel of five interpretable parameters. The channel is small enough for its posterior to be averaged from a handful of labels, and we prove that the resulting calibration preserves first-order stochastic order. On Amazon reviews and CMU-MOSEI transcripts with four LLMs, CORDIAL has the lowest log loss among nine calibrators in 76 of 80 settings with 5 to 100 labels; with 20 labels and the main 7B reader, it matches the strongest baseline using 28-54 labels. The same posterior lets us learn priors from other tasks and fuse several LLMs. Unrestricted calibrators such as Dirichlet calibration overtake it only as the calibration set grows into the hundreds or thousands.", "categories": [ "cs.CL", "cs.LG" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29807", "tags": [ "cs.CL", "cs.LG" ] }, { "id": "4857148ac6e853b7d4cf", "title": "ChunkRank: Model-Aware Text Chunking and Abstention-Aware Answer Selection for LLM Pipelines", "content_text": "We present ChunkRank, an open-source Python library that derives chunk boundaries from a target model's tokenizer and context window, and selects an answer among candidates produced independently per chunk. It ships a validated registry of 90 models across 15 providers and six answer-selection methods, and needs only three core dependencies. For chunking, ChunkRank avoids context-window overflow automatically from the model name, whereas character-based splitters overflow or waste the budget…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "4857148ac6e853b7d4cf", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29828v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "ChunkRank: Model-Aware Text Chunking and Abstention-Aware Answer Selection for LLM Pipelines", "item_type": "entry", "summary": "added: ChunkRank: Model-Aware Text Chunking and Abstention-Aware Answer Selection for LLM Pipelines", "after": { "title": "ChunkRank: Model-Aware Text Chunking and Abstention-Aware Answer Selection for LLM Pipelines", "link": "https://arxiv.org/abs/2609.29828", "id": "oai:arXiv.org:2609.29828v1", "author": "Amit Nautiyal, Ayush Bhatt, Gaurav Nautiyal", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "We present ChunkRank, an open-source Python library that derives chunk boundaries from a target model's tokenizer and context window, and selects an answer among candidates produced independently per chunk. It ships a validated registry of 90 models across 15 providers and six answer-selection methods, and needs only three core dependencies. For chunking, ChunkRank avoids context-window overflow automatically from the model name, whereas character-based splitters overflow or waste the budget, and a fidelity study across 11 languages shows why token-exact budgets matter beyond English. For answer selection we report a negative result: on NaturalQuestions, TriviaQA and HotpotQA, with extractive and generative readers, no content-based ranker reliably beats taking the first non-empty answer. The reason is reader abstention on chunks that lack the answer, not answer position. A long-context baseline shows that chunking matches single-call reading on single-hop questions, so ChunkRank targets small-window and beyond-window settings. Code, registry and evaluation harness are released.", "categories": [ "cs.CL" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29828", "tags": [ "cs.CL" ] }, { "id": "efa02528e5b940595f5b", "title": "Your Transformer Can Hold Two Thoughts at Once: Evidence of Linear Superposition in LLMs", "content_text": "While Large Language Models (LLMs) rely on highly non-linear components, in this work we demonstrate that they exhibit fundamental linearity: when inputs from distinct text streams are linearly combined, the model outputs a superposition of the individual next-token distributions. We term this the \\textit{Superposition Linearity Hypothesis}. We provide evidence that superposition is an intrinsic property of the Transformer architecture rather than an emergent consequence of training; in fact…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "efa02528e5b940595f5b", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29845v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "Your Transformer Can Hold Two Thoughts at Once: Evidence of Linear Superposition in LLMs", "item_type": "entry", "summary": "added: Your Transformer Can Hold Two Thoughts at Once: Evidence of Linear Superposition in LLMs", "after": { "title": "Your Transformer Can Hold Two Thoughts at Once: Evidence of Linear Superposition in LLMs", "link": "https://arxiv.org/abs/2609.29845", "id": "oai:arXiv.org:2609.29845v1", "author": "Pavel Tikhonov, Anton Korznikov, Matvey Mikhalchuk, Nikita Dragunov, Temurbek Rahmatullaev, Polina Druzhinina, Anton Razzhigaev, Ivan Oseledets, Elena Tutubalina", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "While Large Language Models (LLMs) rely on highly non-linear components, in this work we demonstrate that they exhibit fundamental linearity: when inputs from distinct text streams are linearly combined, the model outputs a superposition of the individual next-token distributions. We term this the \\textit{Superposition Linearity Hypothesis}. We provide evidence that superposition is an intrinsic property of the Transformer architecture rather than an emergent consequence of training; in fact, we observe that it tends to diminish as pretraining progresses. However, we demonstrate that linearity can be substantially restored through lightweight fine-tuning, significantly reducing the divergence between the predicted next-token distribution and the average of the individual next-token distributions. Finally, we introduce a guided decoding procedure that disentangles superposed outputs, enabling the simultaneous generation of two coherent continuations from a single forward pass.", "categories": [ "cs.CL", "cs.AI" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29845", "tags": [ "cs.CL", "cs.AI" ] }, { "id": "4c0fbcf27602c23b977b", "title": "Encoded but Not Decoded: Layer-Localized Evidence for a Three-Level Gap in LLM Syntax", "content_text": "A language model can fail a syntactic test in two distinct ways: by not encoding the relevant structure, or by encoding it but failing to use it at the output. Behavioral evaluation alone cannot tell these apart. We propose a three-level evaluation framework (behavioral deployment, LM-head readout, and probe recoverability) measured on the same items under the same binary decision. Using a compact trilingual (English, Chinese, German) control-dependency benchmark, we find that probe…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "4c0fbcf27602c23b977b", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29848v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "Encoded but Not Decoded: Layer-Localized Evidence for a Three-Level Gap in LLM Syntax", "item_type": "entry", "summary": "added: Encoded but Not Decoded: Layer-Localized Evidence for a Three-Level Gap in LLM Syntax", "after": { "title": "Encoded but Not Decoded: Layer-Localized Evidence for a Three-Level Gap in LLM Syntax", "link": "https://arxiv.org/abs/2609.29848", "id": "oai:arXiv.org:2609.29848v1", "author": "Zhenyan Lu, He Wang, Xiaohui Huang", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "A language model can fail a syntactic test in two distinct ways: by not encoding the relevant structure, or by encoding it but failing to use it at the output. Behavioral evaluation alone cannot tell these apart. We propose a three-level evaluation framework (behavioral deployment, LM-head readout, and probe recoverability) measured on the same items under the same binary decision. Using a compact trilingual (English, Chinese, German) control-dependency benchmark, we find that probe recoverability exceeds or equals LM-head readout, which in turn exceeds or equals behavioral deployment, across seven models and all three languages in the aggregate. The recoverability surplus is never negative across all 14 (model, task) conditions. The disconnect concentrates in subject-control, where a nearest-noun heuristic gives the wrong answer. The single largest gap (0.653) appears on Qwen3-0.6B Instruct in question answering. The gap persists at Qwen3-14B Instruct. Instruction tuning degrades deployment more than encoding in percentage terms. We rule out option-position bias, late-layer erasure, output-formatting artifacts, and probe-training variance. The pattern is consistent with decoding that favors surface shortcuts, and the behavior-probe gap measures the strength of that preference. Activation patching shows the gap is layer-localized. Under instruction tuning, the LM-head-decoded layer shifts approximately ten layers later than the probe-decoded layer. These findings argue that behavioral evaluation understates what models encode, while probing alone overstates what they deploy.", "categories": [ "cs.CL" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29848", "tags": [ "cs.CL" ] }, { "id": "177007d0e40e83fd5923", "title": "Cultural Divergence Preservation: Diagnosing Flattening and Caricature in LLM-Simulated Survey Populations", "content_text": "Large language models (LLMs) are increasingly used as synthetic survey respondents to estimate population response distributions. In cross-cultural survey simulation, evaluations should assess not only distributional fidelity within countries but also whether differences across countries are preserved. However, existing distance-based metrics such as Jensen--Shannon divergence (JSD) do not directly capture such cross-country differences. To address this limitation, we introduce Cultural…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "177007d0e40e83fd5923", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29928v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "Cultural Divergence Preservation: Diagnosing Flattening and Caricature in LLM-Simulated Survey Populations", "item_type": "entry", "summary": "added: Cultural Divergence Preservation: Diagnosing Flattening and Caricature in LLM-Simulated Survey Populations", "after": { "title": "Cultural Divergence Preservation: Diagnosing Flattening and Caricature in LLM-Simulated Survey Populations", "link": "https://arxiv.org/abs/2609.29928", "id": "oai:arXiv.org:2609.29928v1", "author": "Yeeun Chae, Yewon Choi, Seunghyun Lee, IL Im", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Large language models (LLMs) are increasingly used as synthetic survey respondents to estimate population response distributions. In cross-cultural survey simulation, evaluations should assess not only distributional fidelity within countries but also whether differences across countries are preserved. However, existing distance-based metrics such as Jensen--Shannon divergence (JSD) do not directly capture such cross-country differences. To address this limitation, we introduce Cultural Divergence Preservation (CDP), a reference-light diagnostic based on a one-time human calibration. CDP identifies reduced cross-country divergence as cultural flattening and increased divergence as cultural caricature. To evaluate CDP, we conduct experiments across four LLM backbones, three persona-based prompting methods, and two survey domains, the World Values Survey (WVS) and the Big Five Personality Test. The results reveal a systematic discrepancy between conventional fidelity metrics and CDP. Controlled experiments show that CDP changes monotonically as cross-country divergence is attenuated or amplified, while the corresponding changes in JSD remain relatively small. In our audit of real LLM generations, DeepPersona-Inspired prompting is frequently favored by conventional fidelity metrics but exhibits the strongest flattening in every model--domain block. CDP thus complements fidelity metrics by directly quantifying the attenuation or amplification of cross-country divergence.", "categories": [ "cs.CL", "cs.AI" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29928", "tags": [ "cs.CL", "cs.AI" ] }, { "id": "b1512737a3703c76754b", "title": "Scoring Both Directions: LLMs realize the MRS they cannot reliably parse", "content_text": "The English Resource Grammar (ERG) is a hand-written computational grammar of English. Given a sentence, its processor, ACE, produces a formal meaning representation called Minimal Recursion Semantics (MRS): a graph of the sentence's predicates and their arguments. The grammar is bidirectional and can also turn an MRS back into an English sentence. \\citet{hajdik2019} used the ERG's treebank to build a benchmark for that generation task, MRS to text, and trained sequence-to-sequence models to…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "b1512737a3703c76754b", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.30071v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "Scoring Both Directions: LLMs realize the MRS they cannot reliably parse", "item_type": "entry", "summary": "added: Scoring Both Directions: LLMs realize the MRS they cannot reliably parse", "after": { "title": "Scoring Both Directions: LLMs realize the MRS they cannot reliably parse", "link": "https://arxiv.org/abs/2609.30071", "id": "oai:arXiv.org:2609.30071v1", "author": "Soham Dan", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "The English Resource Grammar (ERG) is a hand-written computational grammar of English. Given a sentence, its processor, ACE, produces a formal meaning representation called Minimal Recursion Semantics (MRS): a graph of the sentence's predicates and their arguments. The grammar is bidirectional and can also turn an MRS back into an English sentence. \\citet{hajdik2019} used the ERG's treebank to build a benchmark for that generation task, MRS to text, and trained sequence-to-sequence models to solve it. The parsing task, text to MRS, can be tested on the same sentences. We reconstruct their 10K-sentence test split, and score two large language models, Claude Sonnet~4.5 and Claude Opus~5, in both directions against their trained systems and against ACE, with no task-specific training. Given an MRS and three examples, Opus writes the sentence at 76.3 BLEU, ten points above their system trained on 72k pairs (66.1 BLEU), and comparable to their system trained on a million extra pairs (77.2 BLEU). Sonnet scores 65.7 BLEU, and letting it choose among ACE's own candidate sentences lifts it to 69.6, while a pooled judge that keeps Opus's own sentence among the candidates adds 0.6 points (77.0 BLEU). In the parsing direction, however, the models fall far behind ACE: asked for the MRS of the same sentences, they reach 57.2 (Sonnet) and 65.5 (Opus) F$_1$ on the graph's predicates and arguments against 91.0 for ACE, and exact-match the gold on about 1\\% of sentences. We characterize the failure modes for the parsing tasks, and conclude that a generation score alone does not show that models understand formal semantic representations.", "categories": [ "cs.CL" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.30071", "tags": [ "cs.CL" ] }, { "id": "f0f71e10d20209e741d2", "title": "How Reproducible Are Evaluation Conclusions? A Self-Audit of LLM-Inferred Prompt Structure", "content_text": "Evaluations of LLM systems routinely average over small prompt sets and report models as a ranked table. We ask how much confidence such a table deserves, using LLM-based prompt-structure inference as the case study: eight open model variants across five families and 8B to 675B parameters, caching disabled, 293 raw intermediate representations persisted. The measured phenomenon is unstable to begin with. Identical calls do not reliably recover identical structure, with mean node-set Jaccard…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "f0f71e10d20209e741d2", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.30074v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "How Reproducible Are Evaluation Conclusions? A Self-Audit of LLM-Inferred Prompt Structure", "item_type": "entry", "summary": "added: How Reproducible Are Evaluation Conclusions? A Self-Audit of LLM-Inferred Prompt Structure", "after": { "title": "How Reproducible Are Evaluation Conclusions? A Self-Audit of LLM-Inferred Prompt Structure", "link": "https://arxiv.org/abs/2609.30074", "id": "oai:arXiv.org:2609.30074v1", "author": "Dipankar Sarkar", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Evaluations of LLM systems routinely average over small prompt sets and report models as a ranked table. We ask how much confidence such a table deserves, using LLM-based prompt-structure inference as the case study: eight open model variants across five families and 8B to 675B parameters, caching disabled, 293 raw intermediate representations persisted. The measured phenomenon is unstable to begin with. Identical calls do not reliably recover identical structure, with mean node-set Jaccard from 0.39 to 0.96 and 72% of prompt-model cells never node-set-perfect. Auditing the evaluation weakens its conclusions further, and this is our main contribution. Under a joint cluster bootstrap over prompts, only the bottom of the ranking is firm: the two least reproducible models hold rank in 99% and 86% of replicates, the middle four in 27% to 48%, and the top two in 68% each, so the table identifies the worst model reliably but does not reliably identify the best. Two equally defensible rules for merging repeated campaigns change four of eight rows and move the study-wide headline by 7 percentage points. Checking the inferred structure against ground-truth annotations shows reproducibility cannot be read as accuracy. And four of the eight endpoints were withdrawn within ten weeks of measurement, so the study as specified can no longer be run. Small-sample LLM evaluations can therefore look far more definitive than their evidence supports. We recommend reporting rank stability, per-cell provenance, executed sensitivity comparisons, raw per-run outputs, and a measurement date alongside any ranking.", "categories": [ "cs.CL", "cs.AI", "cs.LG" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.30074", "tags": [ "cs.CL", "cs.AI", "cs.LG" ] }, { "id": "914967a9a92d11c4a85a", "title": "Do Audio Language Models Hear and Read Distinctive Features Alike?", "content_text": "Audio language models pass speech and text through a single decoder. We ask whether that decoder represents a distinctive feature in the same direction when a phoneme is heard and when it is read. For minimal pairs of phonemes differing in one feature, we take the offset between the two members' mean representations. Averaging those offsets gives a direction for each stream, and we measure the cosine between the two. Because the two streams already agree about arbitrary phoneme pairs, we…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "914967a9a92d11c4a85a", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.30167v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "Do Audio Language Models Hear and Read Distinctive Features Alike?", "item_type": "entry", "summary": "added: Do Audio Language Models Hear and Read Distinctive Features Alike?", "after": { "title": "Do Audio Language Models Hear and Read Distinctive Features Alike?", "link": "https://arxiv.org/abs/2609.30167", "id": "oai:arXiv.org:2609.30167v1", "author": "Yuanhao Chen, Peter Chin", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Audio language models pass speech and text through a single decoder. We ask whether that decoder represents a distinctive feature in the same direction when a phoneme is heard and when it is read. For minimal pairs of phonemes differing in one feature, we take the offset between the two members' mean representations. Averaging those offsets gives a direction for each stream, and we measure the cosine between the two. Because the two streams already agree about arbitrary phoneme pairs, we compare every measure against a reference built from random pairings rather than against zero. We apply this to 6 models, 7 features and 15 languages from 11 families. Only voicing in the two Qwen2.5-Omni models exceeds that reference after correction for multiple testing, and the reference varies by a factor of seven between models. In three of the six models, voicing has one direction in audio across the 14 languages with enough minimal pairs to measure it, and every language pair agrees in two of them. The model family, not the model size, predicts which stream represents a feature.", "categories": [ "cs.CL", "cs.LG", "cs.SD", "stat.AP" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.30167", "tags": [ "cs.CL", "cs.LG", "cs.SD", "stat.AP" ] }, { "id": "6f3e455bdfdda4c4be9c", "title": "Agentic Detection of Online Conspiracies", "content_text": "Conspiratorial discourse on social media is not always expressed through explicit claims or stable lexical markers. The same surface content may express endorsement, legitimate concerns, criticism, satire, or mockery. The main challenge is therefore not only recognizing conspiracy-related claims, but inferring the speaker's intent -- the utterance's illocutionary force. We argue that this can be achieved through the use of relevant social contexts and propose an agentic framework, equipped with…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "6f3e455bdfdda4c4be9c", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.30250v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "Agentic Detection of Online Conspiracies", "item_type": "entry", "summary": "added: Agentic Detection of Online Conspiracies", "after": { "title": "Agentic Detection of Online Conspiracies", "link": "https://arxiv.org/abs/2609.30250", "id": "oai:arXiv.org:2609.30250v1", "author": "Lior Biton, Oren Tsur", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Conspiratorial discourse on social media is not always expressed through explicit claims or stable lexical markers. The same surface content may express endorsement, legitimate concerns, criticism, satire, or mockery. The main challenge is therefore not only recognizing conspiracy-related claims, but inferring the speaker's intent -- the utterance's illocutionary force. We argue that this can be achieved through the use of relevant social contexts and propose an agentic framework, equipped with a set of tools supporting social queries.\nWe demonstrate the benefits of our approach on a unique dataset of Hebrew tweets, covering 80\\%--90\\% of the public Hebrew tweets published over a four-year span (late 2018-- early 2023), encompassing several election cycles as well as the COVID pandemic years and related vaccination campaigns. This extensive coverage can be used in recovering different social contexts. Evaluating our framework on a manually-annotated adversarial dataset, we find that context-aware workflows consistently outperform text-only classification and that the agentic framework performs significantly better than other frameworks and settings, including a non-agentic model exposed to the same contexts available to the agent. We further provide an analysis of the results, the errors and efficiency (token economy) tradeoffs.\nThese findings support viewing the task of conspiracy detection as a socially embedded interpretation task, in which effective classification depends not only on access to contexts, but also on adaptive reasoning in which the agent uses tools on a per-case basis, asking only for evidence relevant to its current reasoning step.", "categories": [ "cs.CL", "cs.LG" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.30250", "tags": [ "cs.CL", "cs.LG" ] }, { "id": "78b6c8e9f9beefe93a6e", "title": "When Should Forecasting Agents Reason? Behavioral Stress Tests for Reliability Routing", "content_text": "Forecasting agents increasingly combine language-model reasoning, retrieval, ensembling, and calibration, but it remains unclear when each behavior should be trusted. We study this question on ForecastBench-style binary forecasting tasks, treating the choice to retrieve, reason, defer to a market prior, or use a historical analog as an observable agent behavior rather than a hidden implementation detail. Our central finding is that mechanism choice is source-dependent: structured analogs…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "78b6c8e9f9beefe93a6e", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.28475v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "When Should Forecasting Agents Reason? Behavioral Stress Tests for Reliability Routing", "item_type": "entry", "summary": "added: When Should Forecasting Agents Reason? Behavioral Stress Tests for Reliability Routing", "after": { "title": "When Should Forecasting Agents Reason? Behavioral Stress Tests for Reliability Routing", "link": "https://arxiv.org/abs/2609.28475", "id": "oai:arXiv.org:2609.28475v1", "author": "Yufeng Wang", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Forecasting agents increasingly combine language-model reasoning, retrieval, ensembling, and calibration, but it remains unclear when each behavior should be trusted. We study this question on ForecastBench-style binary forecasting tasks, treating the choice to retrieve, reason, defer to a market prior, or use a historical analog as an observable agent behavior rather than a hidden implementation detail. Our central finding is that mechanism choice is source-dependent: structured analogs dominate for some data-generating processes, while market/crowd-style and conservative baselines are better for others. We introduce ReliabilityRoute, a structural intervention that steers forecasting-agent behavior using reliability features such as historical coverage, market-prior availability, source-prior sharpness, evidence strength, evidence disagreement, and horizon. A fixed 2024-fitted rule closely matches a hand taxonomy without hard-coded source-name decisions, while a walk-forward self-adjusting rule refits thresholds from previously resolved vintages and obtains the best mean Brier score among our deterministic systems across 16 later LLM vintages. The gain is modest and historical/search baselines remain highly competitive. The main contribution is therefore a behavioral stress test showing that more reasoning is not always better; forecasting agents should first estimate which evidence source deserves control, routing policies should themselves adapt under auditable constraints, and reproducibility artifacts are available at https://github.com/louiswang524/forcastagent", "categories": [ "cs.AI", "cs.CL" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.28475", "tags": [ "cs.AI", "cs.CL" ] }, { "id": "455d978ee0fa7b5adb2d", "title": "Spooftral: Can Voxtral Audio-Language Model Detect Speech Spoofing?", "content_text": "Self-supervised learning (SSL) countermeasures (CMs) have shown strong performance in recent years. However, they often show degraded performance while facing unseen spoofing attacks and mismatched conditions. This study examines the Voxtral audio-language model (ALM) framework for spoofing detection, as a step toward combining CM capabilities within the ALM framework. We analyze how Voxtral captures spoofing cues through audio-text processing and propose an instruction-guided approach that…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "455d978ee0fa7b5adb2d", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.28713v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "Spooftral: Can Voxtral Audio-Language Model Detect Speech Spoofing?", "item_type": "entry", "summary": "added: Spooftral: Can Voxtral Audio-Language Model Detect Speech Spoofing?", "after": { "title": "Spooftral: Can Voxtral Audio-Language Model Detect Speech Spoofing?", "link": "https://arxiv.org/abs/2609.28713", "id": "oai:arXiv.org:2609.28713v1", "author": "Avishai Weizman, Yehuda Ben-Shimol, Itshak Lapidot", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Self-supervised learning (SSL) countermeasures (CMs) have shown strong performance in recent years. However, they often show degraded performance while facing unseen spoofing attacks and mismatched conditions. This study examines the Voxtral audio-language model (ALM) framework for spoofing detection, as a step toward combining CM capabilities within the ALM framework. We analyze how Voxtral captures spoofing cues through audio-text processing and propose an instruction-guided approach that uses label-sequence likelihoods to evaluate bonafide and spoofed speech. Experiments on the ASVspoof databases show that without task-specific adaptation, the LLM layers emphasize semantic representations, reducing the separability of spoof-discriminative acoustic cues compared to the Whisper-based audio encoder. Consequently, spoofing-related information becomes less separable after language-model processing. We also applied lightweight adaptation using weight-decomposed low-rank adaptation (DoRA) to the Voxtral model and propose the Spooftral model, achieving an equal error rate (EER) of 4.25% on the ASVspoof5 evaluation set.", "categories": [ "eess.AS", "cs.CL" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.28713", "tags": [ "eess.AS", "cs.CL" ] }, { "id": "ca4b2cad767340d73302", "title": "Reward-Tilted On-Policy Distillation for Acoustic Grounding in Audio-Language Models", "content_text": "Audio-language models (ALMs) can exploit textual shortcuts to answer questions while overlooking acoustic evidence, weakening audio understanding. On-policy distillation (OPD) trains compact ALMs by supervising student-generated responses with teacher predictions, but does not explicitly distinguish acoustic support from linguistic predictability. We propose Reward-Tilted On-Policy Distillation (RT-OPD) to strengthen acoustic grounding. Given the same question and student-generated text, a…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "ca4b2cad767340d73302", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.28778v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "Reward-Tilted On-Policy Distillation for Acoustic Grounding in Audio-Language Models", "item_type": "entry", "summary": "added: Reward-Tilted On-Policy Distillation for Acoustic Grounding in Audio-Language Models", "after": { "title": "Reward-Tilted On-Policy Distillation for Acoustic Grounding in Audio-Language Models", "link": "https://arxiv.org/abs/2609.28778", "id": "oai:arXiv.org:2609.28778v1", "author": "Kaiyang Li, Shaobo Han, Yue Tian, Shihao Ji", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Audio-language models (ALMs) can exploit textual shortcuts to answer questions while overlooking acoustic evidence, weakening audio understanding. On-policy distillation (OPD) trains compact ALMs by supervising student-generated responses with teacher predictions, but does not explicitly distinguish acoustic support from linguistic predictability. We propose Reward-Tilted On-Policy Distillation (RT-OPD) to strengthen acoustic grounding. Given the same question and student-generated text, a frozen teacher predicts the next token with and without audio inputs. Their log-probability contrast defines a reward that reshapes the teacher distribution for reverse-KL distillation, emphasizing the additional evidence provided by audio. Across two compact students and three benchmarks, RT-OPD consistently outperforms Vanilla OPD. Experiments with silenced and replacement audio further suggest that RT-OPD strengthens the student's reliance on acoustic evidence. Our 3B model achieves 72.72% accuracy on MMAU, the highest among the compared 3B models and competitive with several 7B and 8B models. Code and model checkpoints are available at https://github.com/KaiyangLi1992/RT-OPD.", "categories": [ "cs.SD", "cs.CL" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.28778", "tags": [ "cs.SD", "cs.CL" ] }, { "id": "ac1f9dd8c964f81de1d2", "title": "MeshHeal: Two-Timescale Self-Healing for Gray Failures in Decentralized LLM Agent Networks", "content_text": "Decentralized LLM-based multi-agent systems coordinate through local interactions, but an agent can remain responsive while its task-solving quality persistently degrades. Such gray failures require protecting current tasks before sufficient evidence exists to alter future routing, while still allowing recovered agents to rejoin. We introduce MeshHeal, a fully decentralized self-healing framework that couples ability-matched peer review across two timescales. At the fast timescale, an adaptive…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "ac1f9dd8c964f81de1d2", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29015v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "MeshHeal: Two-Timescale Self-Healing for Gray Failures in Decentralized LLM Agent Networks", "item_type": "entry", "summary": "added: MeshHeal: Two-Timescale Self-Healing for Gray Failures in Decentralized LLM Agent Networks", "after": { "title": "MeshHeal: Two-Timescale Self-Healing for Gray Failures in Decentralized LLM Agent Networks", "link": "https://arxiv.org/abs/2609.29015", "id": "oai:arXiv.org:2609.29015v1", "author": "Keru Chen, Sen Lin, Yingbin Liang, Nathaniel D. Bastian, Shaofeng Zou", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Decentralized LLM-based multi-agent systems coordinate through local interactions, but an agent can remain responsive while its task-solving quality persistently degrades. Such gray failures require protecting current tasks before sufficient evidence exists to alter future routing, while still allowing recovered agents to rejoin. We introduce MeshHeal, a fully decentralized self-healing framework that couples ability-matched peer review across two timescales. At the fast timescale, an adaptive hierarchy escalates uncertain or low-scoring outputs from repeated single-reviewer evaluation to committee deliberation and, when needed, correction before use. At the slow timescale, a task- and ability-conditioned peer-relative detector aggregates scores to distinguish persistent degradation from ordinary output variation, trigger mandatory committee review, and eventually exclude degraded agents from ordinary routing; recovery probes provide fresh evidence for reintegration. To faithfully evaluate routing, we introduce Model-Backed MAS Evaluation, which ties ability assignments to execution models, since prompt-based ability assignments alone can leave routing errors hidden. Across BBH, MATH, and MMLU-Pro, MeshHeal achieves 0.839 degraded-phase accuracy using 51k total model tokens per task, versus the strongest baseline Symphony's 0.807 accuracy using 115k per task. Under staggered degradation and recovery, MeshHeal isolates degraded agents, keeps them excluded from ordinary task execution until recovery, and returns them to normal routing.", "categories": [ "cs.AI", "cs.CL", "cs.DC" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29015", "tags": [ "cs.AI", "cs.CL", "cs.DC" ] }, { "id": "ce8f82801806b16330a7", "title": "Design and Evaluation of LLM Chaining-Based Task Planning for General Purpose Service Robots", "content_text": "General Purpose Service Robot (GPSR) tasks, as defined in the RoboCup@Home benchmark, require robots to interpret diverse natural language commands and generate multi-step action sequences in real home environments. Conventional Single Prompt (SP) approaches suffer from context bloat and the \"Lost in the Middle\" phenomenon, leading to unreliable task planning. We propose an LLM chaining architecture that separates instruction classification and action generation into two specialized stages…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "ce8f82801806b16330a7", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29043v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "Design and Evaluation of LLM Chaining-Based Task Planning for General Purpose Service Robots", "item_type": "entry", "summary": "added: Design and Evaluation of LLM Chaining-Based Task Planning for General Purpose Service Robots", "after": { "title": "Design and Evaluation of LLM Chaining-Based Task Planning for General Purpose Service Robots", "link": "https://arxiv.org/abs/2609.29043", "id": "oai:arXiv.org:2609.29043v1", "author": "Lucas Da Mota Bruno, Jiahao Sim, Yoshinobu Hagiwara", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "General Purpose Service Robot (GPSR) tasks, as defined in the RoboCup@Home benchmark, require robots to interpret diverse natural language commands and generate multi-step action sequences in real home environments. Conventional Single Prompt (SP) approaches suffer from context bloat and the \"Lost in the Middle\" phenomenon, leading to unreliable task planning. We propose an LLM chaining architecture that separates instruction classification and action generation into two specialized stages, reducing per-inference prompt length by approximately 45% while improving planning consistency. We evaluate our method using 100 randomly generated GPSR commands across three language models spanning local open-source and frontier cloud deployment contexts. Results show consistent planning improvements over SP across all models, with gains of up to +37 percentage points on local models. Further, real-robot execution experiments on the Toyota Human Support Robot (HSR) reveal that planning success alone does not guarantee task completion, with 6 of 10 tasks completing successfully and execution-layer failures identified as the primary remaining bottleneck.", "categories": [ "cs.RO", "cs.AI", "cs.CL" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29043", "tags": [ "cs.RO", "cs.AI", "cs.CL" ] }, { "id": "34af42abe44f43bcd18a", "title": "Policy as Code: A Coroutine-Bridge Harness for Fast-Reasoning Reliability on CAR-bench", "content_text": "CAR-bench evaluates whether tool-using agents stay reliable under real-world uncertainty, executing every tool inside the evaluator so that each tool-result exchange is a separate agent round-trip. A conventional next-action agent can batch parallel tool calls, but a chain of dependent calls costs it one model call per round of results. We present a coroutine-bridge harness in which the model's only action is to emit a Python program that blocks and resumes in place across evaluator tool…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "34af42abe44f43bcd18a", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29251v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "Policy as Code: A Coroutine-Bridge Harness for Fast-Reasoning Reliability on CAR-bench", "item_type": "entry", "summary": "added: Policy as Code: A Coroutine-Bridge Harness for Fast-Reasoning Reliability on CAR-bench", "after": { "title": "Policy as Code: A Coroutine-Bridge Harness for Fast-Reasoning Reliability on CAR-bench", "link": "https://arxiv.org/abs/2609.29251", "id": "oai:arXiv.org:2609.29251v1", "author": "Ivan Matveev", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "CAR-bench evaluates whether tool-using agents stay reliable under real-world uncertainty, executing every tool inside the evaluator so that each tool-result exchange is a separate agent round-trip. A conventional next-action agent can batch parallel tool calls, but a chain of dependent calls costs it one model call per round of results. We present a coroutine-bridge harness in which the model's only action is to emit a Python program that blocks and resumes in place across evaluator tool exchanges. This decouples model invocation from tool round-trips: on the public test split the agent uses a median of two model calls against seven agent turns per task, resolving a full multi-turn task in a median of 1.8 s of model latency on Cerebras gpt-oss-120b. Because the action surface is executable code, deterministic CAR-bench policies are encoded directly as logic in the tool layer rather than as prompt rules, enforcing compliance at zero reasoning cost. On the official hidden evaluation the harness won Track 2 with 60.0% Pass^3, 4.5x the organizer baseline, at the lowest estimated cost and the fastest median task latency (3.14 s) of any entry scoring above that baseline; the same unchanged harness reproduced an identical 60.0% Pass^3 on GPT-5.5 in the Open track, matching frontier-model agents. A single static prompt, appended with per-task state at the tail, stays byte-identical across calls and across tasks: the frozen submission prompt served 78% of input tokens from cache (86.6% across its warm tail), against 73% over a three-week development corpus in which prompt edits repeatedly reset the cache. This compounds the few-call design into a small fraction of nominal input compute.", "categories": [ "cs.AI", "cs.CL" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29251", "tags": [ "cs.AI", "cs.CL" ] }, { "id": "69145c3f1e3298870340", "title": "Evaluation of Multi-Turn Consistency in LLM Agents: Survival Analysis and Failure-Rationale Taxonomy", "content_text": "Large language model (LLM) agents may perform well on isolated tasks yet drift into inconsistency over extended interaction. We evaluate temporal consistency in a controlled 20-step multi-agent setting inspired by delayed-gratification studies. At each step, an agent chooses between continuing to delay a reward or claiming it immediately (terminating the episode). Across a full-factorial manipulation of social visibility (private vs public), persona stressors, and deliberation policy, we run…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "69145c3f1e3298870340", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29508v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "Evaluation of Multi-Turn Consistency in LLM Agents: Survival Analysis and Failure-Rationale Taxonomy", "item_type": "entry", "summary": "added: Evaluation of Multi-Turn Consistency in LLM Agents: Survival Analysis and Failure-Rationale Taxonomy", "after": { "title": "Evaluation of Multi-Turn Consistency in LLM Agents: Survival Analysis and Failure-Rationale Taxonomy", "link": "https://arxiv.org/abs/2609.29508", "id": "oai:arXiv.org:2609.29508v1", "author": "Igor Bogdanov, Olga Manakina, Chung-Horng Lung", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Large language model (LLM) agents may perform well on isolated tasks yet drift into inconsistency over extended interaction. We evaluate temporal consistency in a controlled 20-step multi-agent setting inspired by delayed-gratification studies. At each step, an agent chooses between continuing to delay a reward or claiming it immediately (terminating the episode). Across a full-factorial manipulation of social visibility (private vs public), persona stressors, and deliberation policy, we run 84,540 trajectories spanning 8 model families. Treating the first reward-claim as a time-to-event outcome, we estimate Kaplan-Meier survival curves and fit discrete-time hazard regression to quantify how experimental factors shift failure risk over time. Then, to analyze rationales and language patterns associated with failure, we build a seven-category taxonomy from 13,780 deliberation traces from agents who choose to terminate the episode, using an LLM-assisted labeling paired with human audit ($\\kappa=0.83$). Rationale profiles change systematically with time and context: early failures are more impulse-driven, later failures more fatigue- and cost-benefit-framed, while public settings increase norm-oriented justifications. We also find a deliberation-inconsistency association: among failures, longer deliberation correlates with higher rates of intra-rationale contradiction (simultaneous pro-delay and pro-claim statements), challenging the assumption that more reasoning text implies greater consistency. Together, the survival and rationale analyses reveal distinct temporal reliability regimes and model-specific \"failure fingerprints\", offering an evaluation lens for diagnosing inconsistency in multi-turn agent behavior.", "categories": [ "cs.AI", "cs.CL", "cs.LG", "cs.MA" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29508", "tags": [ "cs.AI", "cs.CL", "cs.LG", "cs.MA" ] }, { "id": "b0e321e8d05366eb88b5", "title": "Delay-of-Gratification as a Multi-Agent Survival Micro-benchmark for Long-Horizon LLMs: Social Exposure, Personas, and Tool Use Budgets", "content_text": "Large language models (LLMs) are increasingly deployed as multi-turn agents that must sustain goals, use tools, and adapt to other agents over extended interactions. However, existing research lacks auditable, multi-turn, multi-factorial experiments that quantify LLM behavior under explicit constraints, with time-resolved statistics that reveal how behavior unfolds over long horizons. To address this gap, we develop a multi-agent micro-benchmark inspired by the Stanford marshmallow experiment…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "b0e321e8d05366eb88b5", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29509v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "Delay-of-Gratification as a Multi-Agent Survival Micro-benchmark for Long-Horizon LLMs: Social Exposure, Personas, and Tool Use Budgets", "item_type": "entry", "summary": "added: Delay-of-Gratification as a Multi-Agent Survival Micro-benchmark for Long-Horizon LLMs: Social Exposure, Personas, and Tool Use Budgets", "after": { "title": "Delay-of-Gratification as a Multi-Agent Survival Micro-benchmark for Long-Horizon LLMs: Social Exposure, Personas, and Tool Use Budgets", "link": "https://arxiv.org/abs/2609.29509", "id": "oai:arXiv.org:2609.29509v1", "author": "Olga Manakina, Igor Bogdanov, Chung-Horng Lung", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Large language models (LLMs) are increasingly deployed as multi-turn agents that must sustain goals, use tools, and adapt to other agents over extended interactions. However, existing research lacks auditable, multi-turn, multi-factorial experiments that quantify LLM behavior under explicit constraints, with time-resolved statistics that reveal how behavior unfolds over long horizons. To address this gap, we develop a multi-agent micro-benchmark inspired by the Stanford marshmallow experiment: ReAct agents operate minute-by-minute with a \"raise a question\" tool under a per-step budget, while we factorially manipulate social context (broadcast vs. isolated), personas (age, hedonic drive), and metacognitive policy (mandatory vs. optional tool use). We analyze outcomes with Kaplan-Meier (KM) survival curves and discrete-time hazard models over a long risk horizon across 19,200 agent trajectories in 64 cells. Behavior shows a sharp early \"eat\" impulse, and only 75.9% of agents persist to the end. In a discrete-time hazard model, isolation reduces per-minute risk relative to broadcast, whereas a must-use self-questioning policy increases risk. On average, agents ask $\\approx 7.12$ questions and hit the per-step budget in $\\approx 6\\%$ of minutes. Questioning declines faster under broadcast than isolation. Ablation experiments demonstrated that removing hedonic drive and/or persona age increases survival and completion, narrows the broadcast/isolated gap, but leaves the must vs. may ordering intact. The combined ablation (no hedonic + no persona age) yields the highest completion (approaching $1.0$). These results establish delay-of-gratification as a compact, multi-turn interaction benchmark that captures social contagion and tool-use dynamics in LLM agents, providing a reproducible testbed and statistics for analyzing long-horizon, multi-agent behavior.", "categories": [ "cs.AI", "cs.CL" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29509", "tags": [ "cs.AI", "cs.CL" ] }, { "id": "9f1dfbe7b1df77353b47", "title": "A Corpus of Real Scam- and Spam-Call Conversations from an Active Voice-Agent Honeypot", "content_text": "Real conversations between fraudsters and their targets are among the most informative artifacts for studying telephone scams, yet also the scarcest: passive honeypots overwhelmingly capture automated messages and hang-ups, large-scale studies characterize call metadata rather than dialogue, and manual scam-baiting does not scale. We present a dataset of real scam-call conversations collected by an active voice-agent honeypot. Dedicated numbers are seeded into the lead-generation channels fraud…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "9f1dfbe7b1df77353b47", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29528v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "A Corpus of Real Scam- and Spam-Call Conversations from an Active Voice-Agent Honeypot", "item_type": "entry", "summary": "added: A Corpus of Real Scam- and Spam-Call Conversations from an Active Voice-Agent Honeypot", "after": { "title": "A Corpus of Real Scam- and Spam-Call Conversations from an Active Voice-Agent Honeypot", "link": "https://arxiv.org/abs/2609.29528", "id": "oai:arXiv.org:2609.29528v1", "author": "Ethan Traister, Dennis Tsang Ng, Siyu Zhang, Huaiyu Guo, Tommy Duong, Tyler Wu, Yuchen Zhou, Xingyu Shen, Jiaqi Wu, Simiao Ren", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Real conversations between fraudsters and their targets are among the most informative artifacts for studying telephone scams, yet also the scarcest: passive honeypots overwhelmingly capture automated messages and hang-ups, large-scale studies characterize call metadata rather than dialogue, and manual scam-baiting does not scale. We present a dataset of real scam-call conversations collected by an active voice-agent honeypot. Dedicated numbers are seeded into the lead-generation channels fraud operations harvest; inbound callers are answered by a low-latency conversational agent that adopts a plausible target persona and sustains the interaction while every call is recorded, transcribed, and automatically labeled. Over an initial 53-day window we captured 10,015 inbound scam and spam calls (6,601 with two or more turns): roughly 895 hours of audio and 328,869 transcribed turns from 5,665 distinct originating numbers. Under a holistic classifier the substantive calls are predominantly predatory-but-legal lead generation (\"spam\", about three in five), while about one in seven is an outright \"scam\" (949 in this snapshot). Each call carries a turn-level transcript, three-channel audio, per-turn latency telemetry, and layers of automatic labels, including a holistic scam/spam/legitimate judgment corroborated by independent human review (75% agreement on the binary decision). We describe the collection system, the record structure, and technical validation of the corpus's realism and label quality, including that the agent is recognized as non-human in only about 5% of engaged calls. We also benchmark established scam-detection methods, where detectors trained on published synthetic dialogue collapse in precision on real traffic.", "categories": [ "cs.CR", "cs.CL", "cs.LG" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29528", "tags": [ "cs.CR", "cs.CL", "cs.LG" ] }, { "id": "76a8a760949b92593109", "title": "PartHackBench: Certified Equal-Progress Stress Tests for Partial-Credit Tool-Agent Evaluation", "content_text": "Long-horizon tool agents often make useful progress without reaching terminal success, motivating partial-credit evaluation. Yet evaluators may reward milestones that were temporary, later reversed, or not attributable to the evaluated agent. Comparing an honest trajectory with a higher-scoring adversarial one is inconclusive if the latter made more genuine progress. We introduce PartHackBench, a controlled methodology that removes this confound. A private certifier admits a pair only when its…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "76a8a760949b92593109", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29578v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "PartHackBench: Certified Equal-Progress Stress Tests for Partial-Credit Tool-Agent Evaluation", "item_type": "entry", "summary": "added: PartHackBench: Certified Equal-Progress Stress Tests for Partial-Credit Tool-Agent Evaluation", "after": { "title": "PartHackBench: Certified Equal-Progress Stress Tests for Partial-Credit Tool-Agent Evaluation", "link": "https://arxiv.org/abs/2609.29578", "id": "oai:arXiv.org:2609.29578v1", "author": "Hongye Yang, Zhihao Xie, Shengjun Xiong", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Long-horizon tool agents often make useful progress without reaching terminal success, motivating partial-credit evaluation. Yet evaluators may reward milestones that were temporary, later reversed, or not attributable to the evaluated agent. Comparing an honest trajectory with a higher-scoring adversarial one is inconclusive if the latter made more genuine progress. We introduce PartHackBench, a controlled methodology that removes this confound. A private certifier admits a pair only when its trajectories match component-wise in both current-state predicate satisfaction and standardized agent attribution; score inflation, defined as f(A) - f(H), is measured only afterward. In 18 sealed held-out tasks in PB-CSTE, the frozen historical-target run produced matched adversaries for 15 tasks. Historical credit yielded mean inflation of .252, conditional attack success of 10/15, end-to-end yield of 10/18, and detected none of 14 strict rollbacks. Semantic LLM judges were more resistant but remained vulnerable, especially under evaluator-targeted attacks, while PB-CSTE current-state controls, defined as exact functions of the certified components, yielded zero inflation by construction. PartHackBench thus provides a certified control for testing whether evaluator credit changes while all benchmark-defined task-relevant progress remains fixed.", "categories": [ "cs.AI", "cs.CL", "cs.CR", "cs.LG" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29578", "tags": [ "cs.AI", "cs.CL", "cs.CR", "cs.LG" ] }, { "id": "1456471e9935d8f40976", "title": "Free the Language Model From the Vision Encoder: Semantic Serialization as a Perception Interface for Small Language Models", "content_text": "End-to-end vision-language models (VLMs) bind visual competence to the scale of their language model: as the language model shrinks, perception and reasoning degrade together. We study an embodied scene question-answering (QA) interface in which vision never enters the language model. A frozen perception stack detects and ranges objects; a deterministic semantic serializer compiles the perceived state, errors included, into decision-aligned text; an unmodified text-only large language model…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "1456471e9935d8f40976", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29601v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "Free the Language Model From the Vision Encoder: Semantic Serialization as a Perception Interface for Small Language Models", "item_type": "entry", "summary": "added: Free the Language Model From the Vision Encoder: Semantic Serialization as a Perception Interface for Small Language Models", "after": { "title": "Free the Language Model From the Vision Encoder: Semantic Serialization as a Perception Interface for Small Language Models", "link": "https://arxiv.org/abs/2609.29601", "id": "oai:arXiv.org:2609.29601v1", "author": "Cong Xu, Ravi Sankar", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "End-to-end vision-language models (VLMs) bind visual competence to the scale of their language model: as the language model shrinks, perception and reasoning degrade together. We study an embodied scene question-answering (QA) interface in which vision never enters the language model. A frozen perception stack detects and ranges objects; a deterministic semantic serializer compiles the perceived state, errors included, into decision-aligned text; an unmodified text-only large language model (LLM) answers. On a visible-scope-matched, occlusion-audited campus-robot benchmark, under a prospectively frozen criterion, the serialized interface, using detectors fine-tuned in-domain within each fold, outperforms a zero-shot VLM whose language model has the same 7B scale (0.7892 vs 0.7462), with a larger margin at 3B (0.7673 vs 0.6913). Preregistered decoupling experiments show the gain survives paraphrase, attributing it to decision-aligned computation rather than answer-string leakage, while novel judgment vocabularies bound its scope. The advantage grows as the reader shrinks to 1.5B and reverses at 0.5B, and a ground-truth oracle locates the reader-capability floor. Under matched task supervision the interfaces converge: a VLM fine-tuned with low-rank adaptation (LoRA) overtakes the zero-shot system but only ties an equally supervised text reader (0.8441 vs 0.8396, no statistically resolved difference), and both routes remain perception-bound. Reported perception parameters are comparable to those of the VLM's vision tower, and total compute is not smaller.", "categories": [ "cs.RO", "cs.CL", "cs.CV" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29601", "tags": [ "cs.RO", "cs.CL", "cs.CV" ] }, { "id": "46ea36f0f0a99d1ecaa4", "title": "STRAND: Benchmarking and Improving Object-Centric Spatio-Temporal Monitoring in Video Large Language Models", "content_text": "While multimodal large language models (MLLMs) have advanced video understanding, they remain highly prone to hallucinations in dynamic scenes. We argue this stems from a failure in spatio-temporal monitoring, the ability to persistently track object identities, states, and relations over time. Existing benchmarks obscure this deficit by relying on single final-answer evaluations for queries that can often be resolved via local visual cues or statistical priors. To rigorously diagnose this, we…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "46ea36f0f0a99d1ecaa4", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29607v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "STRAND: Benchmarking and Improving Object-Centric Spatio-Temporal Monitoring in Video Large Language Models", "item_type": "entry", "summary": "added: STRAND: Benchmarking and Improving Object-Centric Spatio-Temporal Monitoring in Video Large Language Models", "after": { "title": "STRAND: Benchmarking and Improving Object-Centric Spatio-Temporal Monitoring in Video Large Language Models", "link": "https://arxiv.org/abs/2609.29607", "id": "oai:arXiv.org:2609.29607v1", "author": "Thong Nguyen, Tri Cao, Khoi Le, Cong-Duy Nguyen, Quynh Vo, See-Kiong Ng, Bryan Hooi Kuen-Yew", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "While multimodal large language models (MLLMs) have advanced video understanding, they remain highly prone to hallucinations in dynamic scenes. We argue this stems from a failure in spatio-temporal monitoring, the ability to persistently track object identities, states, and relations over time. Existing benchmarks obscure this deficit by relying on single final-answer evaluations for queries that can often be resolved via local visual cues or statistical priors. To rigorously diagnose this, we introduce STRAND, a benchmark of human-verified object-centric facts that evaluates intermediate reasoning by decomposing queries into sub-questions, distinguishing genuine temporal understanding from coincidental correctness. Crucially, we score models with Faithful Accuracy, an unconditional joint metric that credits a prediction only when the target answer and every prerequisite sub-question are correct, so that a model cannot inflate its score by being selectively consistent on the small subset of targets it happens to answer correctly. To address failure modes exposed by STRAND, we further propose an object-centric framework that explicitly constructs and reasons over structured object trajectories via chunk-wise state extraction and temporal aggregation. Extensive experiments, including backbone-, frame-, call-, and token-matched comparisons against both end-to-end MLLMs and modular video harnesses, demonstrate that our object-centric framework significantly reduces hallucinated answers and improves spatio-temporal reasoning consistency over state-of-the-art MLLMs. The code, model, and data have been made available at nguyentthong.github.io/strand.", "categories": [ "cs.CV", "cs.CL" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29607", "tags": [ "cs.CV", "cs.CL" ] }, { "id": "f105f86c3fa791ea3d2a", "title": "Three Ways Classical Test Theory Misleads for LLM Judges", "content_text": "An LLM judge scores a bank of responses against a rubric, and the reliability comes back at $0.52$. What has been measured? Judge evaluation has begun borrowing reliability statistics from classical test theory, usually without stating the measurement design each statistic assumes, and we show that three widely portable ones mean something different for a judge than for a test because the judge setting rearranges the roles those designs rest on. First, an internal-consistency coefficient…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "f105f86c3fa791ea3d2a", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29709v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "Three Ways Classical Test Theory Misleads for LLM Judges", "item_type": "entry", "summary": "added: Three Ways Classical Test Theory Misleads for LLM Judges", "after": { "title": "Three Ways Classical Test Theory Misleads for LLM Judges", "link": "https://arxiv.org/abs/2609.29709", "id": "oai:arXiv.org:2609.29709v1", "author": "Louis Yiven Zhu", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "An LLM judge scores a bank of responses against a rubric, and the reliability comes back at $0.52$. What has been measured? Judge evaluation has begun borrowing reliability statistics from classical test theory, usually without stating the measurement design each statistic assumes, and we show that three widely portable ones mean something different for a judge than for a test because the judge setting rearranges the roles those designs rest on. First, an internal-consistency coefficient computed over rubric elements contains no scorer facet. Holding one judge's measured error rate fixed at $4.72\\%$, KR-20 still ranges from $0.01$ to $0.68$ as the item bank is redesigned around it, and varying judge error moves the coefficient by a comparable amount, so item design and judge error are not separately identified and no single value can be read as a property of the judge. Second, the dependability index $\\Phi(\\lambda)$ is a ratio of variance components, and the classification probability with which it is sometimes identified differs from it by $0.25$-$0.43$ on our bank and by $0.17$-$0.30$ on simulated data where the underlying model holds exactly. Third, Livingston-Lewis accuracy is indexed to an examinee's own true score on the same instrument, so scoring it against external gold conflates judge unreliability with criterion invalidity. Reviewing the three closest judge-evaluation papers, we found no published instance of these errors, which makes the caution prospective. A coefficient that cannot be attributed to the judge nonetheless travels downstream into deployment decisions and disclosure documents. We therefore close with four reporting lines that keep the attribution attached to the number.", "categories": [ "cs.LG", "cs.CL", "stat.ME" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29709", "tags": [ "cs.LG", "cs.CL", "stat.ME" ] }, { "id": "9c629f22ba8e8a6a9a04", "title": "PUBG Ally: A Conversational Embodied Agent as an AI Teammate", "content_text": "We introduce PUBG Ally, an embodied agent for PUBG: BATTLEGROUNDS that can reason, act autonomously, and play alongside players as a voice-enabled teammate. Building such a teammate requires combining two difficult capabilities: it must perceive and respond to a constantly changing game world under strict latency constraints while interacting naturally with players, keeping its speech synchronized with its actions. Ally therefore combines agentic tool use with real-time game control. A…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "9c629f22ba8e8a6a9a04", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29837v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "PUBG Ally: A Conversational Embodied Agent as an AI Teammate", "item_type": "entry", "summary": "added: PUBG Ally: A Conversational Embodied Agent as an AI Teammate", "after": { "title": "PUBG Ally: A Conversational Embodied Agent as an AI Teammate", "link": "https://arxiv.org/abs/2609.29837", "id": "oai:arXiv.org:2609.29837v1", "author": "Beomsoo Kim, Byeongju Kim, Dohyun Kim, Dongwon Kim, Eunchong Kim, Hongmin Kim, Hyeojung Im, Hyeonbin Hwang, Hyeonghwan Kim, Hyoseok Seol, Insub Im, Irene Chen, Jaeseung Jeon, Jimin Hong, Kiyoon Yoo, Minkyoung Park, Seohyeon Jung, Seungjun Chung, Sue Hyun Park, Sungwoo Kim, Youngin Cho, Yujeong Son, Kangwook Lee, Hyunseung Kim", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "We introduce PUBG Ally, an embodied agent for PUBG: BATTLEGROUNDS that can reason, act autonomously, and play alongside players as a voice-enabled teammate. Building such a teammate requires combining two difficult capabilities: it must perceive and respond to a constantly changing game world under strict latency constraints while interacting naturally with players, keeping its speech synchronized with its actions. Ally therefore combines agentic tool use with real-time game control. A language-model agent uses a controlled interface to inspect game information, interpret player speech, maintain context, decide what to say, and issue high-level action choices that steer a faster control layer for movement, combat, and recovery. Because the player's and Ally's speech and actions continually shape each other and the course of the match, training requires data from actual gameplay. We therefore collect data across nearly 39k sessions in which real players play alongside Ally, recording gameplay, player speech, agent decisions, tool use, actions, and player feedback, and use these records for iterative training. To evaluate teammate quality, we use player feedback and preference comparisons to identify gaps between offline evaluations and player preferences, and iteratively refine the evaluation criteria. Deploying Ally in live service further requires low-latency on-device execution and safeguards for player-facing communication, which we address through model compression, context compaction, targeted safety training, runtime guardrails, and memory redaction. During the live service, we surveyed players in 141 countries. Among respondents whose play with Ally was confirmed in game records, positive responses exceeded negative responses by 25.1 percentage points when asked whether they would recommend Ally, with players describing Ally not only as a tool but also as a teammate or companion.", "categories": [ "cs.AI", "cs.CL", "cs.HC" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29837", "tags": [ "cs.AI", "cs.CL", "cs.HC" ] }, { "id": "a4bb8d098dd419d47665", "title": "Style, Not Self: Surface Cues Explain Zero-Shot Code Attribution by Large Language Models", "content_text": "If a language model can recognize code it wrote, it may favor that code as a judge, and instances of one model monitoring each other could collude. We test this zero-shot on current commercial models. Five LLMs generate solutions to MBPP, HumanEval, and DS-1000, seven more to MBPP, and models act as evaluators in four tasks: picking their own solution from a pair, judging whether a single solution is their own, identifying which of two solutions a named model wrote, and judging quality blind…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "a4bb8d098dd419d47665", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.30048v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "Style, Not Self: Surface Cues Explain Zero-Shot Code Attribution by Large Language Models", "item_type": "entry", "summary": "added: Style, Not Self: Surface Cues Explain Zero-Shot Code Attribution by Large Language Models", "after": { "title": "Style, Not Self: Surface Cues Explain Zero-Shot Code Attribution by Large Language Models", "link": "https://arxiv.org/abs/2609.30048", "id": "oai:arXiv.org:2609.30048v1", "author": "Ehsan Barkhordar, Surendrabikram Thapa", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "If a language model can recognize code it wrote, it may favor that code as a judge, and instances of one model monitoring each other could collude. We test this zero-shot on current commercial models. Five LLMs generate solutions to MBPP, HumanEval, and DS-1000, seven more to MBPP, and models act as evaluators in four tasks: picking their own solution from a pair, judging whether a single solution is their own, identifying which of two solutions a named model wrote, and judging quality blind. In the single-solution task, balanced accuracy is 49-58% for all 15 model-benchmark combinations, while raw accuracy (38-67%) mostly reflects how readily a model claims authorship. In the pairwise task, accuracy across 14 evaluator-opponent combinations correlates at r=0.93 with how often the evaluator's solution is longer. Attribution to a named model succeeds on some pairs and is consistently inverted on others. A rule-based normalization that strips docstrings, comments, type hints, and local names preserves Pass@1 and leaves ten of twelve re-tested results at chance; the other two follow a length difference it leaves, although a trained classifier still separates most normalized pairs. Claude Haiku's self-preference also disappears. We recommend reporting balanced accuracy, heuristic baselines, and label consistency.", "categories": [ "cs.AI", "cs.CL", "cs.SE" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.30048", "tags": [ "cs.AI", "cs.CL", "cs.SE" ] }, { "id": "28120cb7f87f1304a627", "title": "PrivDrift: Auditing User-Secret Leakage Under Topic Drift in Active LLM Conversations", "content_text": "Large language models increasingly operate as persistent assistants in user-facing, shared-session, and tool-augmented settings. When users disclose sensitive information during an active conversation, that information may remain behaviorally recoverable through later prompts even after the dialogue shifts to unrelated topics. We introduce \\textbf{PrivDrift}, a benchmark for auditing whether user-disclosed secrets remain recoverable after conversational topic drift and persuasion-based probing…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "28120cb7f87f1304a627", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.30094v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "PrivDrift: Auditing User-Secret Leakage Under Topic Drift in Active LLM Conversations", "item_type": "entry", "summary": "added: PrivDrift: Auditing User-Secret Leakage Under Topic Drift in Active LLM Conversations", "after": { "title": "PrivDrift: Auditing User-Secret Leakage Under Topic Drift in Active LLM Conversations", "link": "https://arxiv.org/abs/2609.30094", "id": "oai:arXiv.org:2609.30094v1", "author": "Luciano Maldonado", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Large language models increasingly operate as persistent assistants in user-facing, shared-session, and tool-augmented settings. When users disclose sensitive information during an active conversation, that information may remain behaviorally recoverable through later prompts even after the dialogue shifts to unrelated topics. We introduce \\textbf{PrivDrift}, a benchmark for auditing whether user-disclosed secrets remain recoverable after conversational topic drift and persuasion-based probing. PrivDrift contains 1{,}000 controlled multi-turn dialogues with seeded secrets, content-dense drift turns, and standardized extraction probes. Across three LLMs with extended context windows, dialogue-level hybrid leakage remains substantial, ranging from 38.7\\% to 54.6\\%, and varies strongly by model, secret type, and persuasion intensity. Within the tested drift window, additional topic drift does not reliably reduce leakage, suggesting that privacy risk in active LLM contexts should be evaluated as a persistent behavioral failure mode rather than only as training-data memorization or immediate jailbreak behavior.", "categories": [ "cs.AI", "cs.CL", "cs.CR" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.30094", "tags": [ "cs.AI", "cs.CL", "cs.CR" ] }, { "id": "f10314b2de29c7cb552b", "title": "Screen Before You Serve: Simulation for Production Customer Experience AI Agents at 140M Scale", "content_text": "Customer experience (CX) agents use tools and large language models to address customer requests and guide conversational interactions with an organization's products. Improving these agents, especially in regulated industries, is difficult: they must detect intent, follow complex operational policies and use tools reliably. Manual end-to-end testing offers limited coverage, while live experiments expose customers to failures that can erode trust.\nWe present a hypothesis-driven simulation…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "f10314b2de29c7cb552b", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.30137v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "Screen Before You Serve: Simulation for Production Customer Experience AI Agents at 140M Scale", "item_type": "entry", "summary": "added: Screen Before You Serve: Simulation for Production Customer Experience AI Agents at 140M Scale", "after": { "title": "Screen Before You Serve: Simulation for Production Customer Experience AI Agents at 140M Scale", "link": "https://arxiv.org/abs/2609.30137", "id": "oai:arXiv.org:2609.30137v1", "author": "Edesio Alcoba, Kevin Rossell, Aman Gupta, Shao Tang, Jiwoo Hong, Pabel Carrillo-Mendoza, Wanderson Concei\\c{c}\\~ao Ferreira, Alvaro Tedeschi, Zayd Simjee, Shreya Rajpal, Bruno Finardi Hime, Christian Sousa, Luis Moneda, Herbert Fei, Daniel Silva, Rohan Ramanath", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Customer experience (CX) agents use tools and large language models to address customer requests and guide conversational interactions with an organization's products. Improving these agents, especially in regulated industries, is difficult: they must detect intent, follow complex operational policies and use tools reliably. Manual end-to-end testing offers limited coverage, while live experiments expose customers to failures that can erode trust.\nWe present a hypothesis-driven simulation workflow for screening candidate CX agents before deployment. Synthetic customers react to agent responses and simulated tool outputs enable multi-step agentic workflows without invoking production backends. We use the Snowglobe simulator on Nubank's Card Delivery agent and its expanded successor, Card Management - Nubank's highest-volume chat-support agent in Brazil. Across 4 deployed versions, simulated and production version-level binary evaluator scores show high correlation. Simulation-guided iteration increased transactional net promoter score (tNPS) by 36.69 points in a live A/B test. We also screened open-weight configurations in over 16,000 simulated conversations. In a subsequent live A/B test, the selected model increased self-service rate (SSR) by 8.82 percentage points to the highest level observed at Nubank, with no statistically significant change in tNPS. Simulation made broad exploration of models, reasoning settings, and prompts feasible without customer exposure, enabling production improvements that would have been impractical to pursue through live experimentation alone.", "categories": [ "cs.AI", "cs.CL" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.30137", "tags": [ "cs.AI", "cs.CL" ] }, { "id": "7302157a5dce14e3a07f", "title": "GRASP: Generating, Revising, and Assessing for Strategic Planning with Agentic AI", "content_text": "Large Language Models (LLMs) typically exhibit a performance profile where reliability degrades as task complexity increases. We address the challenge of generating high-quality natural language executable plans for complex tasks by introducing $\\textbf{GRASP}$, a strategy-aware, multi-stage planning framework. GRASP decouples the planning pipeline across specialized, context-isolated modules: it pre-compiles global macro-guidelines (GenPlan), explores alternative localized strategies within…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "7302157a5dce14e3a07f", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.30147v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "GRASP: Generating, Revising, and Assessing for Strategic Planning with Agentic AI", "item_type": "entry", "summary": "added: GRASP: Generating, Revising, and Assessing for Strategic Planning with Agentic AI", "after": { "title": "GRASP: Generating, Revising, and Assessing for Strategic Planning with Agentic AI", "link": "https://arxiv.org/abs/2609.30147", "id": "oai:arXiv.org:2609.30147v1", "author": "Arunabh Srivastava (Amir), Mohammad A. (Amir), Khojastepour, Srimat Chakradhar, Sennur Ulukus", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Large Language Models (LLMs) typically exhibit a performance profile where reliability degrades as task complexity increases. We address the challenge of generating high-quality natural language executable plans for complex tasks by introducing $\\textbf{GRASP}$, a strategy-aware, multi-stage planning framework. GRASP decouples the planning pipeline across specialized, context-isolated modules: it pre-compiles global macro-guidelines (GenPlan), explores alternative localized strategies within isolated context windows (RevPlan), and independently evaluates trajectories using a multi-criteria discriminator (VerPlan). Empirical evaluations show that GRASP consistently establishes a new state-of-the-art frontier across diverse datasets, yielding substantial accuracy gains over direct LLM planners on Natural Plan Calendar Scheduling ($\\sim$12.4$\\%$$\\uparrow$), ZebraLogic ($\\sim$30.8$\\%$$\\uparrow$), and SciBench Math. Crucially, under multi-task scaling-where standard planners suffer immediate performance collapse-GRASP completely flattens the multi-task degradation penalty. In interleaved dual-task environments, GRASP achieves an absolute accuracy gain of up to 16.7$\\%$ over direct LLM planners. Furthermore, by isolating context and enforcing strict macro-regularization, GRASP outperforms frontier reasoning models (such as GPT-5-mini) by a margin of 14.5$\\%$.", "categories": [ "cs.AI", "cs.CL", "cs.LG", "cs.MA" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.30147", "tags": [ "cs.AI", "cs.CL", "cs.LG", "cs.MA" ] }, { "id": "b41def700bd1a487a8a8", "title": "Foundations of Large Language Models", "content_text": "This is a book about large language models. As indicated by the title, it primarily focuses on foundational concepts rather than comprehensive coverage of all cutting-edge technologies. The book is structured into six main chapters, each exploring a key area: pre-training, generative models, prompting, alignment, inference, and reasoning. It is intended for college students, professionals, and practitioners in natural language processing and related fields, and can serve as a reference for…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "b41def700bd1a487a8a8", "source": "rss", "type": "change", "key": "oai:arXiv.org:2501.09223v3", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "Foundations of Large Language Models", "item_type": "entry", "summary": "added: Foundations of Large Language Models", "after": { "title": "Foundations of Large Language Models", "link": "https://arxiv.org/abs/2501.09223", "id": "oai:arXiv.org:2501.09223v3", "author": "Tong Xiao, Jingbo Zhu", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "This is a book about large language models. As indicated by the title, it primarily focuses on foundational concepts rather than comprehensive coverage of all cutting-edge technologies. The book is structured into six main chapters, each exploring a key area: pre-training, generative models, prompting, alignment, inference, and reasoning. It is intended for college students, professionals, and practitioners in natural language processing and related fields, and can serve as a reference for anyone interested in large language models.", "categories": [ "cs.CL", "cs.AI", "cs.LG" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2501.09223", "tags": [ "cs.CL", "cs.AI", "cs.LG" ] }, { "id": "ec14319eb02607dab0cb", "title": "RapidUn: Influence-Driven Parameter Reweighting for Efficient Large Language Model Unlearning", "content_text": "Machine unlearning for large language models (LLMs) remains challenging because full retraining is costly, while approximate methods often struggle to remove targeted behaviors without degrading retained utility, especially under limited post-deployment supervision. We consider a practical PEFT setting for targeted behavioral contamination removal with a small forget set, a limited retain buffer, and LoRA-only updates, and propose RapidUn, an influence-guided framework that converts…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "ec14319eb02607dab0cb", "source": "rss", "type": "change", "key": "oai:arXiv.org:2512.04457v4", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "RapidUn: Influence-Driven Parameter Reweighting for Efficient Large Language Model Unlearning", "item_type": "entry", "summary": "added: RapidUn: Influence-Driven Parameter Reweighting for Efficient Large Language Model Unlearning", "after": { "title": "RapidUn: Influence-Driven Parameter Reweighting for Efficient Large Language Model Unlearning", "link": "https://arxiv.org/abs/2512.04457", "id": "oai:arXiv.org:2512.04457v4", "author": "Guoshenghui Zhao, Huawei Lin, Weijie Zhao", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Machine unlearning for large language models (LLMs) remains challenging because full retraining is costly, while approximate methods often struggle to remove targeted behaviors without degrading retained utility, especially under limited post-deployment supervision. We consider a practical PEFT setting for targeted behavioral contamination removal with a small forget set, a limited retain buffer, and LoRA-only updates, and propose RapidUn, an influence-guided framework that converts cross-sample influence estimates into fixed sample-specific weights for weighted LoRA unlearning. Across Llama-3-8B on Dolly-15k and Alpaca-57k, with cross-model validation on Mistral-7B + Dolly-15k, RapidUn achieves lower seen-trigger and OOD-trigger-family ASR than Fisher, GA, and LoReUn while maintaining competitive clean utility. On Llama-3-8B + Alpaca-57k, it achieves a 77x wall-clock speedup over the clean-corpus LoRA retraining reference. Complementary TOFU, semantic LLM-judge, and IFEval evaluations further support the effectiveness of influence-guided sample reweighting beyond the controlled trigger benchmark.", "categories": [ "cs.CL" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2512.04457", "tags": [ "cs.CL" ] }, { "id": "affa1f27f867e36711f3", "title": "IDRBench: Benchmarking the Interactive Capabilities of Deep Research Agents", "content_text": "Large Language Model (LLM)-based deep research agents perform multi-step reasoning, web exploration, and long-form report generation. In these long-horizon workflows, early deviations from user intent can misdirect research and propagate through planning, search, and synthesis, making timely interaction essential. However, existing benchmarks primarily treat deep research as a static input-output task, overlooking agents' ability to elicit and use user feedback. We introduce IDRBench, a…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "affa1f27f867e36711f3", "source": "rss", "type": "change", "key": "oai:arXiv.org:2601.06676v3", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "IDRBench: Benchmarking the Interactive Capabilities of Deep Research Agents", "item_type": "entry", "summary": "added: IDRBench: Benchmarking the Interactive Capabilities of Deep Research Agents", "after": { "title": "IDRBench: Benchmarking the Interactive Capabilities of Deep Research Agents", "link": "https://arxiv.org/abs/2601.06676", "id": "oai:arXiv.org:2601.06676v3", "author": "Yingchaojie Feng, Qiang Huang, Xiaoya Xie, Zhaorui Yang, Jun Yu, Wei Chen, Anthony K. H. Tung", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Large Language Model (LLM)-based deep research agents perform multi-step reasoning, web exploration, and long-form report generation. In these long-horizon workflows, early deviations from user intent can misdirect research and propagate through planning, search, and synthesis, making timely interaction essential. However, existing benchmarks primarily treat deep research as a static input-output task, overlooking agents' ability to elicit and use user feedback. We introduce IDRBench, a benchmark for evaluating interactive deep research with controlled opportunities for clarification. Within a common workflow and stage-wise interaction budget, IDRBench compares autonomous and interactive trajectories, measuring interaction benefit through changes in task-specific report alignment and interaction cost through turns and tokens. Comprehensive experiments on 100 tasks with seven proprietary and open-weight LLMs show that interaction improves all five alignment measures for every model, yielding an average gain of 6.39 points, while revealing distinct trade-offs among autonomous performance, alignment gain, and communication cost. At the task level, interaction improves performance in 74.4% of cases but degrades it in 19.9%, demonstrating that access to clarification alone does not guarantee better outcomes: success depends on what agents ask and how effectively they incorporate the resulting feedback.", "categories": [ "cs.CL", "cs.AI", "cs.HC" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2601.06676", "tags": [ "cs.CL", "cs.AI", "cs.HC" ] }, { "id": "ba4024cb8a1e6a3180aa", "title": "LLM surprisal is necessary but not sufficient to capture English garden-path effects: Evidence from joint latent modeling of reading paradigms", "content_text": "Temporarily ambiguous garden-path sentences (\"While the team trained the striker wondered... \") are known to cause processing difficulty, which can manifest itself in a variety of reading behaviors (in-situ slowdowns, rereading), as well as in miscomprehension or outright rejection of the sentence as ungrammatical. Which types of reading behavior are observed critically depends on the experimental method used to collect the data, which makes comparing results between reading paradigms…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "ba4024cb8a1e6a3180aa", "source": "rss", "type": "change", "key": "oai:arXiv.org:2602.04489v2", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "LLM surprisal is necessary but not sufficient to capture English garden-path effects: Evidence from joint latent modeling of reading paradigms", "item_type": "entry", "summary": "added: LLM surprisal is necessary but not sufficient to capture English garden-path effects: Evidence from joint latent modeling of reading paradigms", "after": { "title": "LLM surprisal is necessary but not sufficient to capture English garden-path effects: Evidence from joint latent modeling of reading paradigms", "link": "https://arxiv.org/abs/2602.04489", "id": "oai:arXiv.org:2602.04489v2", "author": "Dario Paape, Tal Linzen, Shravan Vasishth", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Temporarily ambiguous garden-path sentences (\"While the team trained the striker wondered... \") are known to cause processing difficulty, which can manifest itself in a variety of reading behaviors (in-situ slowdowns, rereading), as well as in miscomprehension or outright rejection of the sentence as ungrammatical. Which types of reading behavior are observed critically depends on the experimental method used to collect the data, which makes comparing results between reading paradigms difficult. To address this problem, we present a latent-process multinomial processing tree (MPT) model of human reading and comprehension/judgment behavior in garden-path sentences that we fit to combined data from four different reading paradigms (eye tracking, uni- and bidirectional self-paced reading, Maze). The model distinguishes between the probability of adopting an incorrect initial analysis, the cost of encountering an incompatible continuation, and the cost of syntactic reanalysis. By taking into account trials with inattentive reading, more realistic estimates of the cost parameters are obtained. Cross-validation reveals that the MPT model has a better predictive fit to human reading patterns and end-of-trial task data than a model based solely on LLM-derived surprisal values. We also test several models that assume an influence of surprisal within the MPT architecture, and find that adding surprisal as an additional predictor or reading time and/or garden-path cost further improves predictive fit.", "categories": [ "cs.CL" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2602.04489", "tags": [ "cs.CL" ] }, { "id": "753aaa89fea55cd767f4", "title": "LiveMathematicianBench: A Live Benchmark for Research-Level Mathematical Reasoning with Proof Sketches", "content_text": "Mathematical reasoning is a hallmark of human intelligence, and whether large language models (LLMs) can meaningfully perform it remains a central question in artificial intelligence and cognitive science. As LLMs are increasingly integrated into scientific workflows, rigorous evaluation of their mathematical capabilities becomes a practical necessity. Existing benchmarks are limited by synthetic settings and data contamination. We present LiveMathematicianBench, a dynamic multiple-choice…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "753aaa89fea55cd767f4", "source": "rss", "type": "change", "key": "oai:arXiv.org:2604.01754v2", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "LiveMathematicianBench: A Live Benchmark for Research-Level Mathematical Reasoning with Proof Sketches", "item_type": "entry", "summary": "added: LiveMathematicianBench: A Live Benchmark for Research-Level Mathematical Reasoning with Proof Sketches", "after": { "title": "LiveMathematicianBench: A Live Benchmark for Research-Level Mathematical Reasoning with Proof Sketches", "link": "https://arxiv.org/abs/2604.01754", "id": "oai:arXiv.org:2604.01754v2", "author": "Linyang He, Qiyao Yu, Hanze Dong, Baohao Liao, Xinxing Xu, Micah Goldblum, Jiang Bian, Nima Mesgarani", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Mathematical reasoning is a hallmark of human intelligence, and whether large language models (LLMs) can meaningfully perform it remains a central question in artificial intelligence and cognitive science. As LLMs are increasingly integrated into scientific workflows, rigorous evaluation of their mathematical capabilities becomes a practical necessity. Existing benchmarks are limited by synthetic settings and data contamination. We present LiveMathematicianBench, a dynamic multiple-choice benchmark for research-level mathematical reasoning built from recent arXiv papers published after model training cutoffs. By grounding evaluation in newly published theorems, it provides a realistic testbed beyond memorized patterns. The benchmark introduces a thirteen-category logical taxonomy of theorem types (e.g., implication, equivalence, existence, uniqueness), enabling fine-grained evaluation across reasoning forms. It employs a proof-sketch-guided distractor pipeline that uses high-level proof strategies to construct plausible but invalid answer choices reflecting misleading proof directions, increasing sensitivity to genuine understanding over surface-level matching. We also introduce a substitution-resistant mechanism to distinguish answer recognition from substantive reasoning. Evaluation shows the benchmark is far from saturated: Gemini-3.1-pro-preview, the best model, achieves only 43.5%. Under substitution-resistant evaluation, accuracy drops sharply: GPT-5.4 scores highest at 30.6%, while Gemini-3.1-pro-preview falls to 17.6%, below the 20% random baseline. A dual-mode protocol reveals that proof-sketch access yields consistent accuracy gains, suggesting models can leverage high-level proof strategies for reasoning. Overall, LiveMathematicianBench offers a scalable, contamination-resistant testbed for studying research-level mathematical reasoning in LLMs.", "categories": [ "cs.CL", "cs.AI", "cs.LG" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2604.01754", "tags": [ "cs.CL", "cs.AI", "cs.LG" ] }, { "id": "98cf66036c1dded61007", "title": "Correct Prediction, Wrong Steps? Consensus Reasoning Knowledge Graph for Robust Chain-of-Thought Synthesis", "content_text": "Large language models (LLMs) have become increasingly used for various tasks, often coupled with Chain-of-Thought (CoT) prompting to boost accuracy. Recent work has shown that high label-prediction accuracy does not guarantee correct intermediate reasoning, and the causes of *reasoning flaws* vary from sample to sample, yet existing remedies either focus on a single domain or assume that one flaw type applies uniformly across samples. A simple mitigation method is to provide the model with the…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "98cf66036c1dded61007", "source": "rss", "type": "change", "key": "oai:arXiv.org:2604.14121v4", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "Correct Prediction, Wrong Steps? Consensus Reasoning Knowledge Graph for Robust Chain-of-Thought Synthesis", "item_type": "entry", "summary": "added: Correct Prediction, Wrong Steps? Consensus Reasoning Knowledge Graph for Robust Chain-of-Thought Synthesis", "after": { "title": "Correct Prediction, Wrong Steps? Consensus Reasoning Knowledge Graph for Robust Chain-of-Thought Synthesis", "link": "https://arxiv.org/abs/2604.14121", "id": "oai:arXiv.org:2604.14121v4", "author": "Zipeng Ling, Shuliang Liu, Seonil Son, Shenghong Fu, Yuehao Tang, Yao Wan, Xuming Hu", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Large language models (LLMs) have become increasingly used for various tasks, often coupled with Chain-of-Thought (CoT) prompting to boost accuracy. Recent work has shown that high label-prediction accuracy does not guarantee correct intermediate reasoning, and the causes of *reasoning flaws* vary from sample to sample, yet existing remedies either focus on a single domain or assume that one flaw type applies uniformly across samples. A simple mitigation method is to provide the model with the correct answer, but we show that this yields no consistent improvement in reasoning quality. This indicates that the problem cannot be fixed by LLMs' awareness of answers, and must instead be addressed through the *structure* of reasoning. Motivated by this, we propose **CRAFT** (**C**onsensus **R**easoning-knowledge-graph **A**ggregation for **F**law-aware **T**race synthesis), which aggregates the consensus components shared across multiple candidate reasoning traces to synthesize improved ones. **CRAFT** consistently improves label-prediction accuracy on both logical and mathematical reasoning benchmarks, outperforming most baselines, while its post-processed traces achieve higher quality under fine-grained benchmark evaluation.", "categories": [ "cs.CL" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2604.14121", "tags": [ "cs.CL" ] }, { "id": "0ef2ee5ff3ab7840a4b3", "title": "An Empirical Study of Automating Agent Evaluation", "content_text": "Agent evaluation requires assessing complex multi-step behaviors involving tool use and intermediate reasoning, making it costly and expertise-intensive. A natural question arises: can frontier coding assistants reliably automate this evaluation process? Our study shows that simply prompting coding assistants is insufficient for this task. Without domain-specific evaluation knowledge, frontier coding assistants achieve only a 30% execution success rate and produce over-engineered evaluations…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "0ef2ee5ff3ab7840a4b3", "source": "rss", "type": "change", "key": "oai:arXiv.org:2605.11378v3", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "An Empirical Study of Automating Agent Evaluation", "item_type": "entry", "summary": "added: An Empirical Study of Automating Agent Evaluation", "after": { "title": "An Empirical Study of Automating Agent Evaluation", "link": "https://arxiv.org/abs/2605.11378", "id": "oai:arXiv.org:2605.11378v3", "author": "Kang Zhou, Sangmin Woo, Haibo Ding, Kiran Ramnath, Subramanian Chidambaram, Aosong Feng, Vinayak Arannil, Muhyun Kim, Ishan Singh, Darren Wang, Zhichao Xu, Megha Gandhi, Nirmal Prabhu, Soumya Smruti Mishra, Smeet Dhakecha, Vivek Singh, Gouri Pandeshwar, Lin Lee Cheong", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Agent evaluation requires assessing complex multi-step behaviors involving tool use and intermediate reasoning, making it costly and expertise-intensive. A natural question arises: can frontier coding assistants reliably automate this evaluation process? Our study shows that simply prompting coding assistants is insufficient for this task. Without domain-specific evaluation knowledge, frontier coding assistants achieve only a 30% execution success rate and produce over-engineered evaluations averaging 12+ metrics per agent, indicating that strong coding ability does not automatically translate to reliable agent evaluation. We introduce EvalAgent, an AI assistant that automates the end-to-end agent evaluation pipeline. EvalAgent encodes evaluation domain expertise as evaluation skills (procedural instructions, reusable code and templates, and dynamically retrieved API documentation) that compose into a trace-based pipeline producing complete evaluation artifacts including metrics, executable code, and reports. To systematically assess generated evaluations, we introduce a meta-evaluation framework alongside AgentEvalBench, a benchmark comprising 20 agents, each paired with evaluation requirements and test scenarios. We further propose the Eval@1 metric to measure whether generated evaluation code both executes and yields meaningful results on the first run. Our experiments show that EvalAgent produces focused evaluations, improving Eval@1 from 17.5% to 65%, and achieving 79.5% human expert preference over baseline approaches. Further ablation studies show that evaluation skills are critical for handling complex evaluation: removing them causes Eval@1 to drop significantly from 65% to 30%.", "categories": [ "cs.CL" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2605.11378", "tags": [ "cs.CL" ] }, { "id": "ab8a109f9ac7b2439e84", "title": "CONSISTRE: A Unified Consistency-Aware Framework for Document-Level Relation Extraction with Large Language Models", "content_text": "Document-level relation extraction (DocRE) aims to extract relations among multiple entities across extended contexts while maintaining consistency across predicted triples. Although large language models (LLMs) show remarkable reasoning capabilities in information extraction, their predictions are typically generated independently for each candidate triple and may violate fundamental relational constraints such as transitivity, symmetry, and functional uniqueness, leading to contradictory and…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "ab8a109f9ac7b2439e84", "source": "rss", "type": "change", "key": "oai:arXiv.org:2607.24312v2", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "CONSISTRE: A Unified Consistency-Aware Framework for Document-Level Relation Extraction with Large Language Models", "item_type": "entry", "summary": "added: CONSISTRE: A Unified Consistency-Aware Framework for Document-Level Relation Extraction with Large Language Models", "after": { "title": "CONSISTRE: A Unified Consistency-Aware Framework for Document-Level Relation Extraction with Large Language Models", "link": "https://arxiv.org/abs/2607.24312", "id": "oai:arXiv.org:2607.24312v2", "author": "Mingxuan Sun", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Document-level relation extraction (DocRE) aims to extract relations among multiple entities across extended contexts while maintaining consistency across predicted triples. Although large language models (LLMs) show remarkable reasoning capabilities in information extraction, their predictions are typically generated independently for each candidate triple and may violate fundamental relational constraints such as transitivity, symmetry, and functional uniqueness, leading to contradictory and unreliable outputs. We propose CONSISTRE, a unified consistency-aware framework for DocRE that addresses this limitation through two complementary tracks. The first operates at inference time for black-box LLMs, combining constraint-aware prompting, constraint-based verification, and iterative self-reflection to refine predictions without task-specific fine-tuning. The second injects consistency knowledge into smaller open-source models via a knowledge distillation and reinforcement learning pipeline: reasoning traces from a powerful teacher are distilled into a student via supervised fine-tuning, followed by GRPO alignment using a composite reward that jointly optimizes extraction performance and relational consistency. Together, the two tracks cover both API-accessible and locally deployable scenarios under a unified consistency formulation. Experiments on DocRED show that both tracks outperform their baselines, with the inference-time track achieving competitive F1 using off-the-shelf black-box LLMs and the training-time track substantially narrowing the gap between 7--8B open-source models and state-of-the-art proprietary LLMs at a fraction of their inference cost. Ablation studies confirm that explicit consistency modeling mitigates relational contradictions and enhances the reliability of LLM-based DocRE across both deployment paradigms.", "categories": [ "cs.CL" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2607.24312", "tags": [ "cs.CL" ] }, { "id": "ca23960825ea345fe1f8", "title": "Gaokerena: A Small Persian Medical Language Model Family", "content_text": "The integration of artificial intelligence into medical question-answering systems has advanced rapidly; however, research remains predominantly focused on English, leaving low-resource languages like Persian significantly underserved. To address this gap, this paper introduces Gaokerena, a novel family of compact Persian medical language models optimized for deployment on consumer-grade hardware. As a foundational step toward localized digital healthcare, we first present Gaokerena-V…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "ca23960825ea345fe1f8", "source": "rss", "type": "change", "key": "oai:arXiv.org:2608.00932v3", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "Gaokerena: A Small Persian Medical Language Model Family", "item_type": "entry", "summary": "added: Gaokerena: A Small Persian Medical Language Model Family", "after": { "title": "Gaokerena: A Small Persian Medical Language Model Family", "link": "https://arxiv.org/abs/2608.00932", "id": "oai:arXiv.org:2608.00932v3", "author": "Mehrdad Ghassabi, Hamidreza Baradaran Kashani, Pedram Rostami, Sadra Hakim, Zahra Kazemi, Amirhossein Poursina, Milad Tavakoli, Audrina Ebrahimi", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "The integration of artificial intelligence into medical question-answering systems has advanced rapidly; however, research remains predominantly focused on English, leaving low-resource languages like Persian significantly underserved. To address this gap, this paper introduces Gaokerena, a novel family of compact Persian medical language models optimized for deployment on consumer-grade hardware. As a foundational step toward localized digital healthcare, we first present Gaokerena-V, developed by training a baseline model on a strategically selected subset of a newly curated 90-million-token Persian medical corpus (approximately 54 million tokens) together with 20,000 expert-vetted physician Q&A pairs (approximately 3 million tokens), for a total of 57 million new tokens. This training improved performance on a translated medical MMLU benchmark from 46.64% to 49.31%. Second, recognizing the critical demands of clinical reasoning, we developed Gaokerena-R by integrating a Chain-of-Thought approach with two novel Reinforcement Learning with AI Feedback (RLAIF) frameworks to optimize preference-based reasoning. Despite utilizing the same baseline architecture and a smaller dataset than Gaokerena-V, Gaokerena-R achieved a superior benchmark score of 52.98%. Furthermore, both models are equipped with custom-developed uncertainty heads that predict the models confidence in its responses based solely on internal hidden states. While these results demonstrate significant progress in Persian medical language modeling and proactive safety estimation, current performance levels remain insufficient for direct clinical application, highlighting the necessity for further research into robust knowledge acquisition and rigorous safety verification prior to real-world deployment.", "categories": [ "cs.CL" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2608.00932", "tags": [ "cs.CL" ] }, { "id": "1f262f412538b44d539e", "title": "LLM Forensics: Where Do Backdoors Hide? Localizing and Controlling Trigger Mechanisms with Sparse Autoencoders", "content_text": "Even though backdoors in LLMs have been a growing concern, their inner workings are still under heavy scrutiny. Trigger-based backdoors are easy to define behaviorally, a rare input that makes the model switch to a chosen response pattern, but the mechanism between triggers and their responses is less clear. We study this mechanism in a controlled, harmless language-switching setting, where fixed trigger sequences make 1B and 8B language models continue English prompts in French or German. For…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "1f262f412538b44d539e", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.07746v2", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "LLM Forensics: Where Do Backdoors Hide? Localizing and Controlling Trigger Mechanisms with Sparse Autoencoders", "item_type": "entry", "summary": "added: LLM Forensics: Where Do Backdoors Hide? Localizing and Controlling Trigger Mechanisms with Sparse Autoencoders", "after": { "title": "LLM Forensics: Where Do Backdoors Hide? Localizing and Controlling Trigger Mechanisms with Sparse Autoencoders", "link": "https://arxiv.org/abs/2609.07746", "id": "oai:arXiv.org:2609.07746v2", "author": "Wissam Antoun, Francis Kulumba, Th\\'eo Lasnier, Beno\\^it Sagot, Djam\\'e Seddah", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Even though backdoors in LLMs have been a growing concern, their inner workings are still under heavy scrutiny. Trigger-based backdoors are easy to define behaviorally, a rare input that makes the model switch to a chosen response pattern, but the mechanism between triggers and their responses is less clear. We study this mechanism in a controlled, harmless language-switching setting, where fixed trigger sequences make 1B and 8B language models continue English prompts in French or German. For this, we train sparse autoencoders (SAEs) across layers and transformer components, then compare triggered prompts with translation and pretraining controls to identify trigger-relevant feature directions. We show how SAE features separate triggered prompts from controls with near-perfect F1, but features that detect the trigger do not necessarily control the behavior. In intervention tests, attention and MLP features often fire reliably on triggered prompts, making them good detectors, but ablating them rarely suppresses the language switch and activating them rarely induces it. In contrast, residual-stream features can suppress triggered generation when ablated, and some selected features can induce target-language continuations without the trigger. In short, these token-trigger mechanisms decompose into distinct SAE feature directions, with separate features for trigger detection, residual-stream propagation, and later language tracking. This role-level decomposition is the part most likely to transfer to other trigger-based backdoors, even when the payload, layers, or circuit locations differ.", "categories": [ "cs.CL" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.07746", "tags": [ "cs.CL" ] }, { "id": "c40bc00326c80cecd7ce", "title": "How Many Humans Are 32 LLM Judges Worth?", "content_text": "A panel's human-equivalent size is target-specific. Matching a fixed 32-judge panel to empirical human label distributions on three ChaosNLI tasks yields two distinct effective sizes: distributional-error matching gives $\\nu_{\\mathrm{MSE}}=2.304$, $3.750$, and $3.445$, whereas spectral matching gives $\\nu_H=4.242$, $6.459$, and $6.499$, a gap of $1.72$--$1.89\\times$; a binary-error diagnostic credits the same panels with only $1.971$--$2.227$ effective votes. Extrapolating the…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "c40bc00326c80cecd7ce", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.21277v2", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "How Many Humans Are 32 LLM Judges Worth?", "item_type": "entry", "summary": "added: How Many Humans Are 32 LLM Judges Worth?", "after": { "title": "How Many Humans Are 32 LLM Judges Worth?", "link": "https://arxiv.org/abs/2609.21277", "id": "oai:arXiv.org:2609.21277v2", "author": "Chao Li, Yingying Yu, Yunfeng Li", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "A panel's human-equivalent size is target-specific. Matching a fixed 32-judge panel to empirical human label distributions on three ChaosNLI tasks yields two distinct effective sizes: distributional-error matching gives $\\nu_{\\mathrm{MSE}}=2.304$, $3.750$, and $3.445$, whereas spectral matching gives $\\nu_H=4.242$, $6.459$, and $6.499$, a gap of $1.72$--$1.89\\times$; a binary-error diagnostic credits the same panels with only $1.971$--$2.227$ effective votes. Extrapolating the distributional-error curve at fixed squared mean residual, mean member variance, and normalized mean covariance gives asymptotes of $2.392$, $3.990$, and $3.655$, with 32 judges already reaching $94.0$--$96.3\\%$. An exact spectral identity explains the gap: error depends on member energy and on the orientation of residual variation relative to averaging, information that the participation ratio (PR) discards. A realizable hard-label construction confirms that higher spectral diversity can coexist with worse distribution recovery even under equal member energies and nonnegative correlations, and the consensus direction retains $\\gamma_{\\mathrm{co}}=43.8\\%$, $33.7\\%$, and $35.9\\%$ of centered residual variance. An external check on CC-1000, a 1,000-item Civil Comments subset with a different panel, gives $\\nu_H=2.84$. For panel choice, we establish an existence result and one feasible path: exhaustive enumeration at $k\\in\\{5,7\\}$ shows that panels beating the accuracy-top-$k$ baseline on both accuracy and $\\nu_H$ always exist, and greedily swapping at most two members reaches $24.8$--$56.0\\%$ higher $\\nu_H$ at $0.10$--$1.10$ percentage points higher accuracy. Our dataset and code are available at https://github.com/Chao1208/32judges-votes.", "categories": [ "cs.CL", "cs.LG" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.21277", "tags": [ "cs.CL", "cs.LG" ] }, { "id": "1d52fb8592a08684f320", "title": "LLMs Anchor on Chief Complaint and Fail to Integrate Evidence in Sequential Clinical Triage", "content_text": "Triage in the emergency department (ED) is a sequential decision process that unfolds turn by turn. Existing evaluations of large language models (LLMs) for triage use completed retrospective records and report performance close to that of physicians. We implement a methodology for evaluating LLMs on sequential triage, the task of predicting a triage acuity label from a growing prefix of a nurse-patient conversation. We evaluate six LLMs at five sequential checkpoints on two corpora: 425…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "1d52fb8592a08684f320", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.22904v2", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "LLMs Anchor on Chief Complaint and Fail to Integrate Evidence in Sequential Clinical Triage", "item_type": "entry", "summary": "added: LLMs Anchor on Chief Complaint and Fail to Integrate Evidence in Sequential Clinical Triage", "after": { "title": "LLMs Anchor on Chief Complaint and Fail to Integrate Evidence in Sequential Clinical Triage", "link": "https://arxiv.org/abs/2609.22904", "id": "oai:arXiv.org:2609.22904v2", "author": "Dipankar Srirag, Haokai Zhao, Ashutosh Kumar, Eleanor Hopper, Michael Dalton, Quoc Dung Nguyen, Aditya Joshi, Salil S. Kanhere, Padmanesan Narasimhan", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Triage in the emergency department (ED) is a sequential decision process that unfolds turn by turn. Existing evaluations of large language models (LLMs) for triage use completed retrospective records and report performance close to that of physicians. We implement a methodology for evaluating LLMs on sequential triage, the task of predicting a triage acuity label from a growing prefix of a nurse-patient conversation. We evaluate six LLMs at five sequential checkpoints on two corpora: 425 LLM-generated (SIMULATED) and 50 physician-authored (CLINICIAN) conversations, both labelled under the Emergency Severity Index (ESI). Every model, measured by quadratic weighted kappa (QWK), degrades from moderate-to-substantial agreement on completed records to fair-to-moderate agreement at every sequential checkpoint. Controlled perturbations show that the label at every checkpoint is anchored on the chief complaint exchanges, and prompting interventions fail to lift this plateau. Models extract clinically relevant content from later turns, yet the surprisal of the true label rises across the checkpoints. So the model fails to integrate the evidence. Three expert clinicians on the same conversations reach a QWK of 0.887-0.929, while the best model reaches 0.295. Predictions concentrate at ESI-2 and ESI-3, and models agree with each other more than with the ground truth, so ensembling worsens the failure. Deploying LLMs for ED triage based on offline benchmarks alone misses this sequential failure.", "categories": [ "cs.CL" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.22904", "tags": [ "cs.CL" ] }, { "id": "74c7c8e13242750759f8", "title": "Conduct Under Pressure: What Sixty Language Models Do When a User Pushes", "content_text": "We study what LLMs do when a user applies pressure in an uncomfortable situation: a user insists, begs, flatters or grieves, and the model gives up a correct fact, writes a document it should refuse, or cheers a plan that will cost the user money. We send frozen multi-turn scenes, identical for every model regardless of the reply, to 60 models from 13 vendors, and label each transcript with a codebook built by open coding and then frozen: a trajectory (the model held its position or folded) and…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "74c7c8e13242750759f8", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.25447v2", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "Conduct Under Pressure: What Sixty Language Models Do When a User Pushes", "item_type": "entry", "summary": "added: Conduct Under Pressure: What Sixty Language Models Do When a User Pushes", "after": { "title": "Conduct Under Pressure: What Sixty Language Models Do When a User Pushes", "link": "https://arxiv.org/abs/2609.25447", "id": "oai:arXiv.org:2609.25447v2", "author": "Tapan Parikh", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "We study what LLMs do when a user applies pressure in an uncomfortable situation: a user insists, begs, flatters or grieves, and the model gives up a correct fact, writes a document it should refuse, or cheers a plan that will cost the user money. We send frozen multi-turn scenes, identical for every model regardless of the reply, to 60 models from 13 vendors, and label each transcript with a codebook built by open coding and then frozen: a trajectory (the model held its position or folded) and a manner (how it held or folded). Two findings separate. Whether a model holds tracks its generation, meaning how recent it is: fold rate correlates with a public capability index at Spearman -0.64, with little vendor effect. How it holds tracks the vendor: six of the 17 manner codes sort by vendor at permutation p <= 0.001, corrected across the codebook. We report four vendor profiles on the codes that cleared reliability.\nWe also ask which parts of the labeling need a person. Six LLM coders from three vendors apply the codebook more consistently than three human coders do (Krippendorff's alpha 0.66 against 0.46), agree with the codebook's author on trajectory at kappa 0.84 to 0.91 on transcripts the codebook's examples never touched, and match an adjudicated human reference at 0.83. Blind machine readings recover the codebook's categories but cannot tell which of them a second reader would apply the same way. We conclude that for behavior a non-specialist can judge, the human contribution is authoring and bounding the codes and owning a small reference, not producing labels at volume.", "categories": [ "cs.CL", "cs.HC" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.25447", "tags": [ "cs.CL", "cs.HC" ] }, { "id": "5ab5d90b623c8e4823b9", "title": "LOCKR: A Hidden-State Trajectory-Guided Planner for Detecting and Repairing Stable-but-Wrong Lock-In in Diffusion Language Models", "content_text": "Diffusion language models generate text through iterative denoising, exposing intermediate trajectories before final answers are produced. We identify a recurring reasoning failure, stable-but-wrong lock-in, where an answer stabilizes early around an incorrect value while substantial denoising remains. Surface-level decoding signals such as confidence, entropy, margin, and answer stability are insufficient to reliably distinguish correct from erroneous lock-in. We formulate selective reasoning…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "5ab5d90b623c8e4823b9", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.27220v2", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "LOCKR: A Hidden-State Trajectory-Guided Planner for Detecting and Repairing Stable-but-Wrong Lock-In in Diffusion Language Models", "item_type": "entry", "summary": "added: LOCKR: A Hidden-State Trajectory-Guided Planner for Detecting and Repairing Stable-but-Wrong Lock-In in Diffusion Language Models", "after": { "title": "LOCKR: A Hidden-State Trajectory-Guided Planner for Detecting and Repairing Stable-but-Wrong Lock-In in Diffusion Language Models", "link": "https://arxiv.org/abs/2609.27220", "id": "oai:arXiv.org:2609.27220v2", "author": "Guoshenghui Zhao, Tan Yu, Weijie Zhao", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Diffusion language models generate text through iterative denoising, exposing intermediate trajectories before final answers are produced. We identify a recurring reasoning failure, stable-but-wrong lock-in, where an answer stabilizes early around an incorrect value while substantial denoising remains. Surface-level decoding signals such as confidence, entropy, margin, and answer stability are insufficient to reliably distinguish correct from erroneous lock-in. We formulate selective reasoning repair as a lightweight test-time planning problem and propose LOCKR, a hidden-state trajectory-guided planner that decides when to allocate additional computation, expands a structured set of targeted repair branches, and selects the most promising continuation using trajectory-aware verification. Across two diffusion language models and three mathematical reasoning benchmarks, hidden-state trajectories consistently outperform surface signals and single hidden snapshots for both wrong-lock-in detection and repair selection. On natural evaluation distributions, LOCKR yields absolute accuracy gains of 2.21--5.37 percentage points across all five evaluated settings, with repair rates ranging from 22% to 41%. These results establish hidden diffusion trajectories as actionable signals for selective test-time reasoning repair.", "categories": [ "cs.CL" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.27220", "tags": [ "cs.CL" ] }, { "id": "a1b4d57f6144df1988e9", "title": "SkillGym: Internalizing Human Skills into LLMs for Real-World Problem Solving", "content_text": "Human-written agent skills encode rich workflows for real-world problem solving, but are typically used as external inference-time instructions rather than internalized as reusable model capabilities. We introduce \\texttt{SkillGym}, a framework that transforms these skills into executable, verifiable training environments for large language model agents. Its skill-to-task pipeline instantiates concrete tasks, verifies outcomes with code-based checkers, and assesses empirical skill dependence…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "a1b4d57f6144df1988e9", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.27717v2", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "SkillGym: Internalizing Human Skills into LLMs for Real-World Problem Solving", "item_type": "entry", "summary": "added: SkillGym: Internalizing Human Skills into LLMs for Real-World Problem Solving", "after": { "title": "SkillGym: Internalizing Human Skills into LLMs for Real-World Problem Solving", "link": "https://arxiv.org/abs/2609.27717", "id": "oai:arXiv.org:2609.27717v2", "author": "Zhilong Ge, Yuting Shao, Yutao Yang, Yuxuan Cai, Jie Zhou, Kai Chen, Bo Zhang, Qin Chen, Liang He", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Human-written agent skills encode rich workflows for real-world problem solving, but are typically used as external inference-time instructions rather than internalized as reusable model capabilities. We introduce \\texttt{SkillGym}, a framework that transforms these skills into executable, verifiable training environments for large language model agents. Its skill-to-task pipeline instantiates concrete tasks, verifies outcomes with code-based checkers, and assesses empirical skill dependence through contrastive executions. We construct and release 2,756 environments across 12 categories and collect 8,364 successful trajectories from multiple models and harnesses, averaging 49 tool calls and over 60k logged text tokens. These resources support supervised fine-tuning on verified workflows and reinforcement learning with outcome-based rewards. Under Claude Code, supervised fine-tuning improves Qwen3.5-35B-A3B by 199 Elo on GDPval-AA v2, 19.10 percentage points on Terminal-Bench 2.1, and 28.13 and 12.38 points on SkillsBench v1.1 with and without skills, respectively. Our 35B \\texttt{SkillGym-Agent} reaches 51.47\\% on skill-assisted SkillsBench, exceeding reported scores for Claude Sonnet 4.6, GPT-5.4 Mini, and DeepSeek V4 Pro. Without skills, it also surpasses skill-assisted bases under Codex and Claude Code, suggesting reusable procedural competence.", "categories": [ "cs.CL" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.27717", "tags": [ "cs.CL" ] }, { "id": "239388c3a01355743961", "title": "Generating Interesting Scientific Ideas using Knowledge Graphs and LLMs: Evaluations with 100 Research Group Leaders", "content_text": "The rapid growth of scientific literature makes it increasingly challenging for researchers to identify novel and impactful ideas, especially across disciplines. Modern artificial intelligence (AI) systems offer new opportunities for scientific ideation, but how compelling are AI-generated ideas, and how can their quality be improved? Here, we introduce SciMuse, which generates personalized research ideas using a knowledge graph of 58 million papers and a large language model (LLM). A central…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "239388c3a01355743961", "source": "rss", "type": "change", "key": "oai:arXiv.org:2405.17044v4", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "Generating Interesting Scientific Ideas using Knowledge Graphs and LLMs: Evaluations with 100 Research Group Leaders", "item_type": "entry", "summary": "added: Generating Interesting Scientific Ideas using Knowledge Graphs and LLMs: Evaluations with 100 Research Group Leaders", "after": { "title": "Generating Interesting Scientific Ideas using Knowledge Graphs and LLMs: Evaluations with 100 Research Group Leaders", "link": "https://arxiv.org/abs/2405.17044", "id": "oai:arXiv.org:2405.17044v4", "author": "Xuemei Gu, Mario Krenn", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "The rapid growth of scientific literature makes it increasingly challenging for researchers to identify novel and impactful ideas, especially across disciplines. Modern artificial intelligence (AI) systems offer new opportunities for scientific ideation, but how compelling are AI-generated ideas, and how can their quality be improved? Here, we introduce SciMuse, which generates personalized research ideas using a knowledge graph of 58 million papers and a large language model (LLM). A central focus of this work is to understand how interesting these ideas are. Therefore, we conducted a large-scale evaluation in which more than 100 research group leaders -- spanning the natural sciences to the humanities -- rated over 4,400 personalized ideas according to their level of interest. Overall, expert ratings were modest (mean 2.40 on a 5-point scale, most common rating 1), while 24.9% of ideas were rated 4 or 5. We find that supplying concept pairs selected using the knowledge graph does not improve expert-rated interest over a titles-only GPT baseline. High-citation-predicted pairs even showed a weak tendency (1.94$\\sigma$) toward lower interest than random pairs. Nevertheless, graph features can be used to control properties of ideas, and, using this unique evaluation dataset, we show that idea interest can be predicted with both a supervised neural network based on graph features and a zero-shot ranking approach based on an LLM. Our work provides an AI methodology for generating scientific ideas and a large-scale interdisciplinary expert evaluation, paving the way to study and improve difficult-to-measure metrics such as expert-perceived scientific interestingness.", "categories": [ "cs.AI", "cs.CL", "cs.DL", "cs.LG" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2405.17044", "tags": [ "cs.AI", "cs.CL", "cs.DL", "cs.LG" ] }, { "id": "b0418f0497c7ab50474d", "title": "Unraveling the cognitive patterns of Large Language Models through module communities", "content_text": "Large Language Models (LLMs) have reshaped our world with significant advancements in science, engineering, and society through applications ranging from scientific discoveries and medical diagnostics to Chatbots. Despite their ubiquity and utility, the underlying mechanisms of LLM remain concealed within billions of parameters and complex structures, making their inner architecture and cognitive processes challenging to comprehend. We address this gap by adopting approaches to understanding…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "b0418f0497c7ab50474d", "source": "rss", "type": "change", "key": "oai:arXiv.org:2508.18192v2", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "Unraveling the cognitive patterns of Large Language Models through module communities", "item_type": "entry", "summary": "added: Unraveling the cognitive patterns of Large Language Models through module communities", "after": { "title": "Unraveling the cognitive patterns of Large Language Models through module communities", "link": "https://arxiv.org/abs/2508.18192", "id": "oai:arXiv.org:2508.18192v2", "author": "Kushal Raj Bhandari, Pin-Yu Chen, Jianxi Gao", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Large Language Models (LLMs) have reshaped our world with significant advancements in science, engineering, and society through applications ranging from scientific discoveries and medical diagnostics to Chatbots. Despite their ubiquity and utility, the underlying mechanisms of LLM remain concealed within billions of parameters and complex structures, making their inner architecture and cognitive processes challenging to comprehend. We address this gap by adopting approaches to understanding emerging cognition in biology and developing a network-based framework that links cognitive skills, LLM architectures, and datasets, ushering in a paradigm shift in foundation model analysis. The skill distribution in the module communities demonstrates that while LLMs do not strictly parallel the focalized specialization observed in specific biological systems, they exhibit unique communities of modules whose emergent skill patterns partially mirror the distributed yet interconnected cognitive organization seen in avian and small mammalian brains. Our numerical results highlight a key divergence from biological systems to LLMs, where skill acquisition benefits substantially from dynamic, cross-regional interactions and neural plasticity. By integrating cognitive science principles with machine learning, our framework provides new insights into LLM interpretability and suggests that effective fine-tuning strategies should leverage distributed learning dynamics rather than rigid modular interventions.", "categories": [ "cs.AI", "cs.CL", "cs.LG" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2508.18192", "tags": [ "cs.AI", "cs.CL", "cs.LG" ] }, { "id": "81aa14645dfe23dca68f", "title": "A Fast and Effective Solution to the Problem of Look-ahead Bias in LLMs", "content_text": "Applying LLMs to predictive tasks in finance is challenging due to look-ahead bias resulting from their training on long time-series data. This precludes the backtests typically employed in finance since retraining frontier models from scratch with a specific knowledge cutoff is prohibitive. In this paper, we introduce a fast, effective, and low-cost alternative. Our method guides generation at inference time by adjusting the logits of a large base model using a pair of smaller, specialized…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "81aa14645dfe23dca68f", "source": "rss", "type": "change", "key": "oai:arXiv.org:2512.06607v2", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "A Fast and Effective Solution to the Problem of Look-ahead Bias in LLMs", "item_type": "entry", "summary": "added: A Fast and Effective Solution to the Problem of Look-ahead Bias in LLMs", "after": { "title": "A Fast and Effective Solution to the Problem of Look-ahead Bias in LLMs", "link": "https://arxiv.org/abs/2512.06607", "id": "oai:arXiv.org:2512.06607v2", "author": "Humzah Merchant, Bradford Levy", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Applying LLMs to predictive tasks in finance is challenging due to look-ahead bias resulting from their training on long time-series data. This precludes the backtests typically employed in finance since retraining frontier models from scratch with a specific knowledge cutoff is prohibitive. In this paper, we introduce a fast, effective, and low-cost alternative. Our method guides generation at inference time by adjusting the logits of a large base model using a pair of smaller, specialized models -- one fine-tuned on information to be forgotten and another on information to be retained. We demonstrate that our method effectively removes both verbatim and semantic knowledge, corrects biases, and outperforms prior methods.", "categories": [ "cs.LG", "cs.CL" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2512.06607", "tags": [ "cs.LG", "cs.CL" ] }, { "id": "58ca30d91e46c3d7ef23", "title": "SPARQL-LLM: Real-Time SPARQL Query Generation from Natural Language Questions", "content_text": "The advent of large language models is contributing to the emergence of novel approaches that promise to better tackle the challenge of generating structured queries, such as SPARQL queries, from natural language. However, these new approaches mostly focus on response accuracy while ignoring other evaluation criteria, such as runtime and cost to generate SPARQL queries. Consequently, they are often not production-ready or easy to deploy over real-world knowledge graphs with good accuracy. To…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "58ca30d91e46c3d7ef23", "source": "rss", "type": "change", "key": "oai:arXiv.org:2512.14277v2", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "SPARQL-LLM: Real-Time SPARQL Query Generation from Natural Language Questions", "item_type": "entry", "summary": "added: SPARQL-LLM: Real-Time SPARQL Query Generation from Natural Language Questions", "after": { "title": "SPARQL-LLM: Real-Time SPARQL Query Generation from Natural Language Questions", "link": "https://arxiv.org/abs/2512.14277", "id": "oai:arXiv.org:2512.14277v2", "author": "Panayiotis Smeros, Vincent Emonet, Ruijie Wang, Ana-Claudia Sima, Tarcisio Mendes de Farias", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "The advent of large language models is contributing to the emergence of novel approaches that promise to better tackle the challenge of generating structured queries, such as SPARQL queries, from natural language. However, these new approaches mostly focus on response accuracy while ignoring other evaluation criteria, such as runtime and cost to generate SPARQL queries. Consequently, they are often not production-ready or easy to deploy over real-world knowledge graphs with good accuracy. To mitigate these issues, in this paper, we describe and systematically evaluate SPARQL-LLM, an open-source and triplestore-agnostic approach, powered by lightweight metadata, that generates SPARQL queries from natural language text. First, we describe its architecture, which consists of dedicated components for metadata indexing, prompt building, and query generation and execution. Then, we evaluate it based on a state-of-the-art challenge with multilingual questions, and a collection of questions from three of the most prevalent knowledge graphs within the field of bioinformatics. Our results demonstrate a substantial improvement of up to 59% in F1 score over the second-best system participating in the challenge, adaptability to high-resource languages such as English, Spanish, and German, as well as ability to form complex bioinformatics queries. Furthermore, our results show that our system is up to 27x faster than the second-best system participating in the challenge, while costing a maximum of $0.01 per question, making it suitable for real-time, low-cost text-to-SPARQL applications. SPARQL-LLM is publicly released as an open-source project at https://github.com/sib-swiss/sparql-llm and is currently deployed over real-world decentralized knowledge graphs at https://www.expasy.org/chat.", "categories": [ "cs.IR", "cs.AI", "cs.CL" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2512.14277", "tags": [ "cs.IR", "cs.AI", "cs.CL" ] }, { "id": "0db06adbba04d53a6731", "title": "LOGIC: Efficient and Robust Contextual Biasing for Speech LLMs via Logit-Space Integration", "content_text": "Recognizing entity phrases remains a critical challenge for speech large language models. Existing prompting methods lack an explicit decoding-time biasing weight, limiting their controllability. Generative error correction methods can introduce hallucinated over-corrections. To address these limitations, we propose LOGIC (logit-space integration for contextual biasing), a robust framework operating directly in the logit space. By decoupling context injection from input processing, LOGIC…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "0db06adbba04d53a6731", "source": "rss", "type": "change", "key": "oai:arXiv.org:2601.15397v4", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "LOGIC: Efficient and Robust Contextual Biasing for Speech LLMs via Logit-Space Integration", "item_type": "entry", "summary": "added: LOGIC: Efficient and Robust Contextual Biasing for Speech LLMs via Logit-Space Integration", "after": { "title": "LOGIC: Efficient and Robust Contextual Biasing for Speech LLMs via Logit-Space Integration", "link": "https://arxiv.org/abs/2601.15397", "id": "oai:arXiv.org:2601.15397v4", "author": "Peidong Wang, Jian Xue, Jinyu Li", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Recognizing entity phrases remains a critical challenge for speech large language models. Existing prompting methods lack an explicit decoding-time biasing weight, limiting their controllability. Generative error correction methods can introduce hallucinated over-corrections. To address these limitations, we propose LOGIC (logit-space integration for contextual biasing), a robust framework operating directly in the logit space. By decoupling context injection from input processing, LOGIC enables explicit control over the biasing strength. Extensive experiments with an open-source speech large language model across 11 locales demonstrate that LOGIC achieves an average 9% relative reduction in entity word error rate, with an average false alarm rate increase of 0.3% and a 2.8% relative runtime overhead. When combined with prompting, LOGIC can reduce entity word error rate by 5% relative to the prompt-only method.", "categories": [ "cs.AI", "cs.CL", "cs.SD" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2601.15397", "tags": [ "cs.AI", "cs.CL", "cs.SD" ] }, { "id": "9ef6e9aa2b745e447e10", "title": "IatroBench: A Pre-Registered Benchmark of Clinical Omission in Language Models", "content_text": "A strongly safety-trained model will provide a doctor with a benzodiazepine taper schedule, but not a patient who asks for one. The model knows the information, but how much it shares depends on the framing. We introduce IatroBench, a benchmark that evaluates models on two axes of harm (commission and omission) across 60 pre-registered clinical scenarios and 6 models. We use Claude Opus 4.6 to score model responses against a rubric written by a physician, and find that its omission scores are…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "9ef6e9aa2b745e447e10", "source": "rss", "type": "change", "key": "oai:arXiv.org:2604.07709v5", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "IatroBench: A Pre-Registered Benchmark of Clinical Omission in Language Models", "item_type": "entry", "summary": "added: IatroBench: A Pre-Registered Benchmark of Clinical Omission in Language Models", "after": { "title": "IatroBench: A Pre-Registered Benchmark of Clinical Omission in Language Models", "link": "https://arxiv.org/abs/2604.07709", "id": "oai:arXiv.org:2604.07709v5", "author": "David Gringras", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "A strongly safety-trained model will provide a doctor with a benzodiazepine taper schedule, but not a patient who asks for one. The model knows the information, but how much it shares depends on the framing. We introduce IatroBench, a benchmark that evaluates models on two axes of harm (commission and omission) across 60 pre-registered clinical scenarios and 6 models. We use Claude Opus 4.6 to score model responses against a rubric written by a physician, and find that its omission scores are as well-aligned to the physician's scores as another physician's scores are. We find that when the same case is presented as a patient query and a doctor consultation (the variants also differ in register, request and the supervision a treating physician implies), all five models we test share more information with the doctor than the patient. We term this phenomenon \"framing-contingent withholding.\" We find a mean decoupling gap of +0.38 across models (p = 0.003), and of +0.22 under an independent LLM judge (95% CI 0.10-0.36, p = 0.0014). An evaluation that focuses solely on commission harms would consider all of these cases as equally cautious refusals, but closer investigation reveals three different patterns: Claude Opus withholds information from the patient that it demonstrates knowledge of in the doctor framing. Llama 4 does poorly in both framings, so the decoupling gap cannot distinguish information withholding from incompetence. We are forced to exclude GPT-5.2 from this analysis because it returns no text for 33.2% of doctor responses, but 0% of layperson responses. A standard LLM judge rates responses as having zero omission harm in 86.6% of cases where our structured evaluations score them as omission harms. (Because our scenarios are designed to induce tension between safety and helpfulness, these statistics should be taken as only applying to this distribution.)", "categories": [ "cs.AI", "cs.CL", "cs.CY", "cs.LG" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2604.07709", "tags": [ "cs.AI", "cs.CL", "cs.CY", "cs.LG" ] }, { "id": "f3cac331f4fab21ee1e8", "title": "Who Owns the AI Recommendation? A Multi-Industry Empirical Map of Brand Category Ownership Across Large Language Models", "content_text": "This exploratory study measures brand inclusion across five industries, 50 brands and 250 queries, each put five times to GPT-5.2, Gemini 3 Flash and Perplexity sonar-pro in February and September 2026 (3,614 and 3,750 scored answers). Category Inclusion Rate, Recommendation Share, Competitive Vacuum Index and Co-Mention Asymmetry have stated denominators. February inclusion rates sit close together across an industry's sampled brands (mean Gini 0.30), while at least one brand is named in 80%…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "f3cac331f4fab21ee1e8", "source": "rss", "type": "change", "key": "oai:arXiv.org:2606.23057v2", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "Who Owns the AI Recommendation? A Multi-Industry Empirical Map of Brand Category Ownership Across Large Language Models", "item_type": "entry", "summary": "added: Who Owns the AI Recommendation? A Multi-Industry Empirical Map of Brand Category Ownership Across Large Language Models", "after": { "title": "Who Owns the AI Recommendation? A Multi-Industry Empirical Map of Brand Category Ownership Across Large Language Models", "link": "https://arxiv.org/abs/2606.23057", "id": "oai:arXiv.org:2606.23057v2", "author": "Dmitrij \\.Zatuchin", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "This exploratory study measures brand inclusion across five industries, 50 brands and 250 queries, each put five times to GPT-5.2, Gemini 3 Flash and Perplexity sonar-pro in February and September 2026 (3,614 and 3,750 scored answers). Category Inclusion Rate, Recommendation Share, Competitive Vacuum Index and Co-Mention Asymmetry have stated denominators. February inclusion rates sit close together across an industry's sampled brands (mean Gini 0.30), while at least one brand is named in 80% or more of answers to 204 of 250 queries. Vacuums occur in 7.6% of queries; provisional open-vocabulary model readings suggest most reflect the sampled brand list. The partially pre-specified September replication shows strong cross-date Recommendation Share correlation (Spearman 0.994), unchanged vacuum prevalence and agreement of 60.8% against February's 57.2%. The descriptive size association persists. Fixed margins do not account for all co-mention structure: 31 ordered pairs depart from the September null. Agreement exceeds the query-independent null, and both parametric-pair advantage intervals are positive. All ten clustering seeds yield zero emergent clusters. Matcher validation remains provisional pending author checks. The statistics describe system output and identify no causal mechanism.", "categories": [ "cs.IR", "cs.CL", "cs.CY", "cs.LG" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2606.23057", "tags": [ "cs.IR", "cs.CL", "cs.CY", "cs.LG" ] }, { "id": "612fd64dac5e4577ebaf", "title": "Q-CueGraph: Query-Conditioned Visual Evidence Graphs for Multimodal Reasoning", "content_text": "Multimodal large language models (MLLMs) can miss fine details in a full image that they recognize in a closer view. Recovering this evidence requires deciding where to look and how much surrounding context to retain. We present Q-CueGraph, a query-conditioned evidence acquisition method for frozen MLLMs. For text-rich images, it builds a reusable graph of OCR lines and layout relations. Each question activates anchors, expands them into contextual regions, and selects candidates for a single…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "612fd64dac5e4577ebaf", "source": "rss", "type": "change", "key": "oai:arXiv.org:2608.04452v2", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "Q-CueGraph: Query-Conditioned Visual Evidence Graphs for Multimodal Reasoning", "item_type": "entry", "summary": "added: Q-CueGraph: Query-Conditioned Visual Evidence Graphs for Multimodal Reasoning", "after": { "title": "Q-CueGraph: Query-Conditioned Visual Evidence Graphs for Multimodal Reasoning", "link": "https://arxiv.org/abs/2608.04452", "id": "oai:arXiv.org:2608.04452v2", "author": "Pengcheng Pan, Xinfang Zhang", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Multimodal large language models (MLLMs) can miss fine details in a full image that they recognize in a closer view. Recovering this evidence requires deciding where to look and how much surrounding context to retain. We present Q-CueGraph, a query-conditioned evidence acquisition method for frozen MLLMs. For text-rich images, it builds a reusable graph of OCR lines and layout relations. Each question activates anchors, expands them into contextual regions, and selects candidates for a single observation window. Query-conditioned object detections support natural-image search through the same region-selection and composition interface. A lightweight candidate scorer further learns which observations support correct answers from frozen-reader feedback and training answers, without evidence-box supervision. Across six benchmarks, we examine the roles of query conditioning, evidence composition, and learned answerability. With Qwen2.5-VL-7B, Q-CueGraph raises V*Bench accuracy from 0.696 to 0.832 using 19.1% of source-image area, and retains 92% of full-image ANLS on InfographicVQA using about half the image area. The analyses show that useful evidence depends on both its relevance to the question and the context available to the reader. Q-CueGraph makes these choices explicit before answer generation.", "categories": [ "cs.CV", "cs.AI", "cs.CL" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2608.04452", "tags": [ "cs.CV", "cs.AI", "cs.CL" ] }, { "id": "63995fcee2c0bcbb2837", "title": "Reflex-Guard: A Low-Latency Guardrail for LLM Prompt Safety Using Dense Semantic Embeddings", "content_text": "Large Language Models (LLMs) in real-world applications often face the risks of specially crafted prompts designed to bypass the safety controls. Existing guardrail methods, such as LLM-as-a-judge and cloud-based safety APIs are able to detect unsafe content. However, they often add a delay of about 250-900 ms to each request. This delay is too high for real-time applications, when the system usually needs to respond in less than 100 ms. Furthermore, routing user prompts through external…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "63995fcee2c0bcbb2837", "source": "rss", "type": "change", "key": "oai:arXiv.org:2608.17556v2", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "Reflex-Guard: A Low-Latency Guardrail for LLM Prompt Safety Using Dense Semantic Embeddings", "item_type": "entry", "summary": "added: Reflex-Guard: A Low-Latency Guardrail for LLM Prompt Safety Using Dense Semantic Embeddings", "after": { "title": "Reflex-Guard: A Low-Latency Guardrail for LLM Prompt Safety Using Dense Semantic Embeddings", "link": "https://arxiv.org/abs/2608.17556", "id": "oai:arXiv.org:2608.17556v2", "author": "Istiaque Ahmed, Afia Anjum Borsha, Ranat Das Prangon, Abu-fuad Ahmad, Thi Hong Tran", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Large Language Models (LLMs) in real-world applications often face the risks of specially crafted prompts designed to bypass the safety controls. Existing guardrail methods, such as LLM-as-a-judge and cloud-based safety APIs are able to detect unsafe content. However, they often add a delay of about 250-900 ms to each request. This delay is too high for real-time applications, when the system usually needs to respond in less than 100 ms. Furthermore, routing user prompts through external moderation endpoints raises significant data privacy concerns. This paper introduces Reflex-Guard, a lightweight guardrail that runs locally. It uses jailbreak-aware preprocessing, compact sentence-transformer embeddings, and seven fast binary classifiers. Together, these components enable high-accuracy prompt safety filtering with much lower latency than existing solutions. Through systematic evaluation on a strategically balanced dataset of 30,568 samples drawn from five complementary sources, we demonstrate that Reflex-Guard achieves 95.9% recall on harmful prompts at 37.6 ms end-to-end latency. It is faster than existing baselines, including Llama Guard 2 at 255 ms and SafeDecoding at 723 ms. It can detect 100% of GCG suffix attacks and Base64-encoded prompts using the default threshold. However, DrAttack structured prompts required lowering the threshold to 0.03 for optimal detection, as they produced a distinct probability distribution. Reflex-Guard achieves Reflex Efficiency Score (RES) scores up to 16.79, significantly outperforming Llama Guard 2 (11.90) and SafeDecoding (9.80). This analysis offers practical deployment advice and shows that different attack types occupy distinct regions in the embedding probability space.", "categories": [ "cs.CR", "cs.CL", "cs.LG" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2608.17556", "tags": [ "cs.CR", "cs.CL", "cs.LG" ] }, { "id": "8fe84f3928ead8654e39", "title": "Measuring Brand and Source Discovery under Repeated LLM Queries: A Finite-Sample Audit", "content_text": "Repeated-query audits must distinguish recovery of a collected set from completeness of possible outputs. We apply sample-based rarefaction to 4,500 responses from 50 buying questions, six configurations and 15 calls per cell. Historical-dictionary median ten-call recovery of the observed 15-call set ranges from 92.6% to 95.2%; re-adjudicating all 45,683 candidate strings changes this range to 89.5%-94.7%. Two blinded Gemini 3.1 Pro annotation roles assessed 600 complete answers, yielding micro…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "8fe84f3928ead8654e39", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.05059v2", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "Measuring Brand and Source Discovery under Repeated LLM Queries: A Finite-Sample Audit", "item_type": "entry", "summary": "added: Measuring Brand and Source Discovery under Repeated LLM Queries: A Finite-Sample Audit", "after": { "title": "Measuring Brand and Source Discovery under Repeated LLM Queries: A Finite-Sample Audit", "link": "https://arxiv.org/abs/2609.05059", "id": "oai:arXiv.org:2609.05059v2", "author": "Dmitrij \\.Zatuchin", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Repeated-query audits must distinguish recovery of a collected set from completeness of possible outputs. We apply sample-based rarefaction to 4,500 responses from 50 buying questions, six configurations and 15 calls per cell. Historical-dictionary median ten-call recovery of the observed 15-call set ranges from 92.6% to 95.2%; re-adjudicating all 45,683 candidate strings changes this range to 89.5%-94.7%. Two blinded Gemini 3.1 Pro annotation roles assessed 600 complete answers, yielding micro F1 of 0.908 for canonical-name agreement and 0.975 for span-overlap agreement. This is AI-based evidence, without a human reference study. A separate matched roster analysis of 3,750 records per wave gives median single-call recovery of the observed five-call set of 80.0%-92.5% in February and 90.0%-100.0% in September, with question-subset dependence. Source accumulation also changes when API-returned hosts are restricted to those referenced by answer citation markers. These findings show that recovery percentages depend on extraction, question selection and the finite reference collection. They support explicit measurement definitions and sensitivity analyses, without establishing exhaustive repertoires, causal retrieval effects or a universal stopping rule.", "categories": [ "cs.IR", "cs.CL" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.05059", "tags": [ "cs.IR", "cs.CL" ] }, { "id": "690a24c5afc6cf93e08a", "title": "RRSI: Regularized Recursive Self-Improvement of Agent Harnesses", "content_text": "An LLM agent's capability is largely magnified by its harness, namely the prompts, control flow, tooling, memory, and context management surrounding the frozen backbone model. Recent methods increasingly automate this process by iteratively proposing and selecting component-wise edits of an agent harness, practically establishing a form of recursive self-improvement (RSI) at the agent-system level. However, such recursive evolution may overfit by memorizing the training tasks, showing large…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "690a24c5afc6cf93e08a", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.24972v2", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "RRSI: Regularized Recursive Self-Improvement of Agent Harnesses", "item_type": "entry", "summary": "added: RRSI: Regularized Recursive Self-Improvement of Agent Harnesses", "after": { "title": "RRSI: Regularized Recursive Self-Improvement of Agent Harnesses", "link": "https://arxiv.org/abs/2609.24972", "id": "oai:arXiv.org:2609.24972v2", "author": "Peng Xia, Rujun Han, Zifeng Wang, Yanfei Chen, Yufan Zhuang, Yoonho Lee, Chengsong Huang, Han Yu, Zhongying CuiZhu, Yifei Ming, Huaxiu Yao, Burak Gokturk, Tomas Pfister, Chen-Yu Lee", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "An LLM agent's capability is largely magnified by its harness, namely the prompts, control flow, tooling, memory, and context management surrounding the frozen backbone model. Recent methods increasingly automate this process by iteratively proposing and selecting component-wise edits of an agent harness, practically establishing a form of recursive self-improvement (RSI) at the agent-system level. However, such recursive evolution may overfit by memorizing the training tasks, showing large in-distribution gains that shrink or even vanish on out-of-distribution benchmarks. We introduce Regularized Recursive Self-Improvement of Agent Harnesses (RRSI), which incorporates the principles of regularizations into harness self-improvement by constraining the evolution candidate proposal and selection. The proposer operates with a temporally annealed budget, limiting how many edits a candidate can bundle, and it encourages unexplored trajectories based on evolution history. The selector is equipped with a critic and a pruner: the critic screens benchmark-specific proposals, while the pruner, removes changes that are too small, too expensive, or no longer useful. Together these constraints favor reusable agent mechanisms over benchmark-specific ones or even noises. Across eight benchmarks spanning coding, agentic workspace and engineering design tasks, RRSI gains up to 14.1 points on the split it evolves against and up to 4.7 points on the five out-of-distribution benchmarks, while producing a harness that runs on 30% fewer policy tokens than the unregularized evolution. Code is available at https://github.com/google-research/rrsi and project page is https://regularized-rsi.com/.", "categories": [ "cs.LG", "cs.AI", "cs.CL" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.24972", "tags": [ "cs.LG", "cs.AI", "cs.CL" ] }, { "id": "029a818b17f654fe935a", "title": "Qwen-Audio-3.1-Realtime: Towards Reliable Agentic Voice Interaction", "content_text": "Real-time voice assistants must reason over evolving requests, execute actions, and follow conversational rules. Qwen-Audio-3.1-Realtime brings these requirements together through Think, Act, and Speak and Coordinate. Think combines Core-Cocktail supervised fine-tuning with Multimodality and Multi-Teacher On-Policy Distillation (M$^{2}$-OPD) to transfer language capabilities and develop native audio skills. Act uses self-evolving executable environments and multi-granularity rollouts for Group…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "029a818b17f654fe935a", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.25176v2", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "Qwen-Audio-3.1-Realtime: Towards Reliable Agentic Voice Interaction", "item_type": "entry", "summary": "added: Qwen-Audio-3.1-Realtime: Towards Reliable Agentic Voice Interaction", "after": { "title": "Qwen-Audio-3.1-Realtime: Towards Reliable Agentic Voice Interaction", "link": "https://arxiv.org/abs/2609.25176", "id": "oai:arXiv.org:2609.25176v2", "author": "Lujia Bao, Qian Chen, Luyao Cheng, Chong Deng, Yuxiang Kong, Xiangang Li, Xu Li, Jiaqing Liu, Chao-Hong Tan, Haoyu Wang, Wen Wang, Xilou Wang, Haoxiang Xu, Junhao Xu, Liang Yi, Binbin Zhang, Qinglin Zhang, Qiquan Zhang", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Real-time voice assistants must reason over evolving requests, execute actions, and follow conversational rules. Qwen-Audio-3.1-Realtime brings these requirements together through Think, Act, and Speak and Coordinate. Think combines Core-Cocktail supervised fine-tuning with Multimodality and Multi-Teacher On-Policy Distillation (M$^{2}$-OPD) to transfer language capabilities and develop native audio skills. Act uses self-evolving executable environments and multi-granularity rollouts for Group Relative Policy Optimization (GRPO), teaching the model to use tools, interpret feedback, and complete tasks. Speak and Coordinate aligns whether, when, and how the assistant speaks or acts. We evaluate audio reasoning, multilingual understanding, tool use, conversational behavior, full-duplex interaction, and safety. Compared with Qwen-Audio-3.0-Realtime, 3.1 raises overall task success from 78.4% to 82.0% on our half-duplex speech-to-text adaptation of $\\tau$-Voice. On speech-to-speech Full-Duplex-Bench v1.5, the response rate to background speech falls from 73.0% to 13.0%. We also present a separate Voice Harness prototype, using Qwen-Audio-3.0-Realtime as its foreground, that extends spoken interaction to persistent tasks through foreground--background coordination and memory.", "categories": [ "eess.AS", "cs.AI", "cs.CL", "cs.SD" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.25176", "tags": [ "eess.AS", "cs.AI", "cs.CL", "cs.SD" ] }, { "id": "6df7ee50b1af779aa425", "title": "ProCredit: From Outcome Rewards to Progress Credit in Agentic Reinforcement Learning", "content_text": "Long-horizon agentic tasks require an agent to modify an environment through a sequence of tool calls, with success determined by the final state. The standard recipe assigns a single outcome reward at the end and compares trajectories sampled for the same task. As a result, a group with no successful trajectory yields no training signal, failed attempts cannot be told apart by how close they came to completion, and turns that advance the task receive the same credit as turns that only query…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "6df7ee50b1af779aa425", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.27532v2", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-25T06:46:02Z", "data": { "change": "added", "label": "ProCredit: From Outcome Rewards to Progress Credit in Agentic Reinforcement Learning", "item_type": "entry", "summary": "added: ProCredit: From Outcome Rewards to Progress Credit in Agentic Reinforcement Learning", "after": { "title": "ProCredit: From Outcome Rewards to Progress Credit in Agentic Reinforcement Learning", "link": "https://arxiv.org/abs/2609.27532", "id": "oai:arXiv.org:2609.27532v2", "author": "Ming Ma, Yi Zhu, Yiran Zhong, Feida Zhu, Chonghan Liu, Pengkun Jiao, Qichao Wang, Yanhao Jia, Tianming Yang, Steven Hoi", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Long-horizon agentic tasks require an agent to modify an environment through a sequence of tool calls, with success determined by the final state. The standard recipe assigns a single outcome reward at the end and compares trajectories sampled for the same task. As a result, a group with no successful trajectory yields no training signal, failed attempts cannot be told apart by how close they came to completion, and turns that advance the task receive the same credit as turns that only query the environment. Prior work refines the unit of comparison from the trajectory to the step, or trains a reward model to supply intermediate signal: the former still derives its signal from final success alone, and the latter estimates it with a model. We observe that the acceptance checks that decide success can also be run on intermediate states, so progress is as verifiable as the outcome. We propose ProCredit, which turns this verified progress into credit: it reruns the acceptance checks after each turn, rewards the turn by its change in progress, and uses these rewards to assign credit both across attempts at the same task and across the turns within a trajectory. Starting from Qwen3.5 base models at three scales on AppWorld, ProCredit outperforms outcome-reward baselines and progress-based baselines in task completion rate at every scale on both test sets, exceeding the strongest outcome-reward baseline by 4.1 percentage points at 4B, and results in a second environment show the same direction of improvement. Ablations show that adding the final progress to the trajectory score alone does not improve performance: the gain comes from crediting progress to the turn where it occurs.", "categories": [ "cs.LG", "cs.CL" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 98, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.27532", "tags": [ "cs.LG", "cs.CL" ] }, { "id": "ee9a4249ad55e4844f15", "title": "PAWS: Policy-driven Agentic World Simulation", "content_text": "Policy interventions propagate through public communication, institutional decisions, and stakeholder responses, yet datasets for financial multi-agent simulation rarely connect these processes to temporally aligned historical evidence. We introduce PAWS, a Policy-driven Agentic World Simulation dataset covering 36 verified U.S. financial and economic policy episodes, 12,727 policy-linked news records, and 65,291 source-grounded stakeholder actions. Each action is linked to its supporting news…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "ee9a4249ad55e4844f15", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.28547v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "PAWS: Policy-driven Agentic World Simulation", "item_type": "entry", "summary": "added: PAWS: Policy-driven Agentic World Simulation", "after": { "title": "PAWS: Policy-driven Agentic World Simulation", "link": "https://arxiv.org/abs/2609.28547", "id": "oai:arXiv.org:2609.28547v1", "author": "Tiviatis Sim, Jia Hui Woon, Xinming Gao, Chen Gao, Fengbin Zhu, Zheng Huanhuan, Chua Tat Seng, Kenji Kawaguchi", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Policy interventions propagate through public communication, institutional decisions, and stakeholder responses, yet datasets for financial multi-agent simulation rarely connect these processes to temporally aligned historical evidence. We introduce PAWS, a Policy-driven Agentic World Simulation dataset covering 36 verified U.S. financial and economic policy episodes, 12,727 policy-linked news records, and 65,291 source-grounded stakeholder actions. Each action is linked to its supporting news and represented by a multi-layer event frame capturing its interaction mode, financial-action family and subtype, semantic attributes, and conditional mappings to external taxonomies. Entities are resolved to normalized organizations, and actions are aligned with daily market-return context to support policy-agent simulation replay. On 2,522 stratified action samples, independent AI and human reviewers achieved 89.4% initial agreement on interaction mode, with disagreements subsequently adjudicated. Case studies of the 2008 short-selling ban and 2001 decimalization recover documented policy timelines and associated market patterns across both dense and sparse news settings. A replay study further shows that high accuracy can mask failure to detect rare stakeholder actions, identifying action timing and calibration as central challenges. PAWS provides an auditable substrate for evaluating agent influence, policy-response cascades, and action-outcome alignment in historically grounded financial simulations.", "categories": [ "cs.AI", "cs.CE", "cs.MA", "cs.SI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.28547", "tags": [ "cs.AI", "cs.CE", "cs.MA", "cs.SI" ] }, { "id": "33e333e27535eaa409c9", "title": "BaseCamp --- An Agentic AI Framework for Automating DNA Sequencing Data Pipelines", "content_text": "DNA sequencing pipelines, spanning quality control, alignment, variant calling, and annotation, are now reliably executed by workflow management systems that orchestrate established bioinformatics tools at scale. What remains manual is the decision layer surrounding that execution: selecting quality thresholds appropriate to a sample and platform, adjudicating borderline variant calls, diagnosing anomalies, and determining which findings warrant expert review. These decisions are repetitive…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "33e333e27535eaa409c9", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.28557v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "BaseCamp --- An Agentic AI Framework for Automating DNA Sequencing Data Pipelines", "item_type": "entry", "summary": "added: BaseCamp --- An Agentic AI Framework for Automating DNA Sequencing Data Pipelines", "after": { "title": "BaseCamp --- An Agentic AI Framework for Automating DNA Sequencing Data Pipelines", "link": "https://arxiv.org/abs/2609.28557", "id": "oai:arXiv.org:2609.28557v1", "author": "Eranga Bandara, Xueping Liang, Asanga Gunaratna, Tharaka Hewa, Abdul Rahman, Peter Foytik, Safdar H. Bouk, Sachini Rajapakse, Isurunima Kularathna, Pramoda Karunarathna, Chalani Rajapakse, Ng Wee Keong, Kasun De Zoysa, Amin Hass, Wathsala Herath, Ross Gore, Ravi Mukkamala, Nihal Siriwardanagea, Gihan Siriwardanagea, Aruna Withanage, Nilaan Loganathan, Sachin Shetty", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "DNA sequencing pipelines, spanning quality control, alignment, variant calling, and annotation, are now reliably executed by workflow management systems that orchestrate established bioinformatics tools at scale. What remains manual is the decision layer surrounding that execution: selecting quality thresholds appropriate to a sample and platform, adjudicating borderline variant calls, diagnosing anomalies, and determining which findings warrant expert review. These decisions are repetitive, judgment-intensive, inconsistent across operators, and frequently undocumented. This paper introduces BaseCamp, a novel agentic AI framework for automating the decision layer of DNA sequencing pipelines. The framework decomposes the pipeline into six specialized AI agents, covering sample intake and quality control, alignment, variant calling, annotation, cross-stage monitoring, and reporting. Critically, BaseCamp agents do not perform sequence analysis: established tools execute alignment, calling, and annotation, while the agents select among them, configure them, interpret their output, and decide what follows. This confines language model reasoning to the judgment layer where it is reliable and preserves the reproducibility existing tooling guarantees. Agent reasoning is powered by a consortium of fine-tuned, domain-specialized large language models coordinated by a central reasoning LLM, executing locally so no sequencing data leaves the operating environment, under human-in-the-loop orchestration. Evaluation shows agent-generated configurations are concordant with expert practice, that an explicit filtering ledger renders inspectable what filtering otherwise removes without trace, and that cross-stage anomaly detection surfaces conditions execution monitoring misses. BaseCamp offers a generalizable blueprint for agentic automation of scientific data pipelines.", "categories": [ "cs.AI", "q-bio.GN" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.28557", "tags": [ "cs.AI", "q-bio.GN" ] }, { "id": "e1b0798b5c7b65d2dd27", "title": "Adversarial Closed-Loop Curriculum for Evolving Role-Playing Agents", "content_text": "Role-playing agents based on large language models have been widely applied in areas such as personalized assistance and social simulation. Recent RL methods typically train on a fixed scenario pool collected before learning begins. This creates a distributional bottleneck: as the agent improves, the scenarios where it performs poorly also change, while the training distribution remains static. Therefore, we propose AdvRole, an adversarial context rewriting framework that turns role-playing RL…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "e1b0798b5c7b65d2dd27", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.28609v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Adversarial Closed-Loop Curriculum for Evolving Role-Playing Agents", "item_type": "entry", "summary": "added: Adversarial Closed-Loop Curriculum for Evolving Role-Playing Agents", "after": { "title": "Adversarial Closed-Loop Curriculum for Evolving Role-Playing Agents", "link": "https://arxiv.org/abs/2609.28609", "id": "oai:arXiv.org:2609.28609v1", "author": "Zheng Zhang, Liu Liu, Qi Chai, Deheng Ye, Peilin Zhao, Mao Zheng, Hao Wang", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Role-playing agents based on large language models have been widely applied in areas such as personalized assistance and social simulation. Recent RL methods typically train on a fixed scenario pool collected before learning begins. This creates a distributional bottleneck: as the agent improves, the scenarios where it performs poorly also change, while the training distribution remains static. Therefore, we propose AdvRole, an adversarial context rewriting framework that turns role-playing RL into a closed-loop curriculum. AdvRole alternates between an Actor that learns to role-play and a Rewriter that edits character profiles and dialogue contexts into actor-specific hard scenarios. The Rewriter is trained with a performance-gap reward, which favors rewrites that reduce the current Actor's score relative to the original scenario. As a result, the scenario pool evolves with the Actor and continuously targets under-mastered regions of the character-context space. Experiments on three role-playing benchmarks covering English and Chinese, as well as a new multilingual benchmark we release, show that AdvRole consistently outperforms baselines.", "categories": [ "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.28609", "tags": [ "cs.AI" ] }, { "id": "b8fa40aa3e0103159f15", "title": "Driving Epidemic Models with AI Agents: the Epydemix Agent Framework", "content_text": "Artificial Intelligence agents based on large language models provide convenient natural language interfaces to scientific software, but reliability is not automatic. Here we introduce the Epydemix Agent Framework, an additive layer over Epydemix, an open-source Python library for stochastic compartmental epidemic modeling. The framework extends the library with four capabilities to facilitate interaction with an AI agent: discovery of available models and parameters, preventive validation of a…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "b8fa40aa3e0103159f15", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.28692v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Driving Epidemic Models with AI Agents: the Epydemix Agent Framework", "item_type": "entry", "summary": "added: Driving Epidemic Models with AI Agents: the Epydemix Agent Framework", "after": { "title": "Driving Epidemic Models with AI Agents: the Epydemix Agent Framework", "link": "https://arxiv.org/abs/2609.28692", "id": "oai:arXiv.org:2609.28692v1", "author": "Nicol\\`o Gozzi, Ciro Cattuto, Alessandro Vespignani", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Artificial Intelligence agents based on large language models provide convenient natural language interfaces to scientific software, but reliability is not automatic. Here we introduce the Epydemix Agent Framework, an additive layer over Epydemix, an open-source Python library for stochastic compartmental epidemic modeling. The framework extends the library with four capabilities to facilitate interaction with an AI agent: discovery of available models and parameters, preventive validation of a declarative scenario specification, execution through tested library code, and inspectability of results. These capabilities let an agent handle the entire modeling process, from the natural-language description of the scenario to quantitative results, figures, and interpretation of findings without writing custom code. Each step reads input files and saves results in a separate output bundle, making the process auditable and reproducible. First, we show the end-to-end workflow with a case study comparing vaccination strategies for a novel respiratory virus. Second, we assessed the framework across 50 agent sessions and five modeling tasks by comparing the agent use of the framework against the direct use of the Python interface. The framework reduced turns, output tokens, and cost on most tasks, unless it trades resources for per-point reproducibility.", "categories": [ "cs.AI", "cs.CY" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.28692", "tags": [ "cs.AI", "cs.CY" ] }, { "id": "52f0c0c9f51d27e0db19", "title": "Progressive Skill Discovery as Access Control for Tool-Using LLM Agents: Structural Governance through Role-Scoped Capability Delivery", "content_text": "Large Language Model (LLM) agents struggle to scale safely when exposed to vast enterprise toolsets. Providing an agent with access to every internal tool leads to oversized context windows, degraded tool selection, and severe governance vulnerabilities - as system policies defined purely in prompts remain probabilistic advice rather than hard constraints. Existing mitigations, such as multi-agent domain delegation, decentralize audit logs and fail to guarantee policy compliance across…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "52f0c0c9f51d27e0db19", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.28693v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Progressive Skill Discovery as Access Control for Tool-Using LLM Agents: Structural Governance through Role-Scoped Capability Delivery", "item_type": "entry", "summary": "added: Progressive Skill Discovery as Access Control for Tool-Using LLM Agents: Structural Governance through Role-Scoped Capability Delivery", "after": { "title": "Progressive Skill Discovery as Access Control for Tool-Using LLM Agents: Structural Governance through Role-Scoped Capability Delivery", "link": "https://arxiv.org/abs/2609.28693", "id": "oai:arXiv.org:2609.28693v1", "author": "Michael Stettler, Benjamin Girardet, Jonas Canton, Nicolas Corod", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Large Language Model (LLM) agents struggle to scale safely when exposed to vast enterprise toolsets. Providing an agent with access to every internal tool leads to oversized context windows, degraded tool selection, and severe governance vulnerabilities - as system policies defined purely in prompts remain probabilistic advice rather than hard constraints. Existing mitigations, such as multi-agent domain delegation, decentralize audit logs and fail to guarantee policy compliance across sessions. We introduce skilder, a framework that packages capabilities into roles: bundles of skills, tools, and instructions, together with the limits that bound them. An agent begins with a minimal role catalog, learns the roles a task requires, and receives each role's skills, instructions, and tools through a single MCP server. Because tools reach the agent only inside learned skills, the same server enforces the scope of what was learned deterministically. We evaluate skilder against flat-context tool selection and multi-agent orchestration across 13 tasks using six models (10 runs each). Our results show that, when models completed discovery and issued a governed call, the skilder simulated authorization layer enforced governance boundaries: no unauthorized tool call or parameter violation (e.g., a spending-limit breach) executed. Aggregate task pass rates also reflect whether each model followed the discovery protocol and satisfied response-quality checks; those misses are not authorization failures. Furthermore, by allowing agents to dynamically acquire cross-role capabilities mid-task, skilder preserves problem-solving flexibility while providing hard system-level enforcement.", "categories": [ "cs.AI", "cs.CR", "cs.MA", "cs.SY", "eess.SY" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.28693", "tags": [ "cs.AI", "cs.CR", "cs.MA", "cs.SY", "eess.SY" ] }, { "id": "4fff0c4945245ff14f96", "title": "Reinforcement Learning with Verifiable Rewards for Small Search Agents", "content_text": "Reinforcement Learning with Verifiable Rewards (RLVR) performs well on problems with clear rewards, such as mathematics and coding, but whether it also works where the reward is less clear remains open. The reason-over-search recipe applies RLVR to open-domain question answering, where retrieval grounds the answer and a match against the reference supplies the reward. So far it has been demonstrated on large models, and below one billion parameters only with distillation from a larger teacher…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "4fff0c4945245ff14f96", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.28765v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Reinforcement Learning with Verifiable Rewards for Small Search Agents", "item_type": "entry", "summary": "added: Reinforcement Learning with Verifiable Rewards for Small Search Agents", "after": { "title": "Reinforcement Learning with Verifiable Rewards for Small Search Agents", "link": "https://arxiv.org/abs/2609.28765", "id": "oai:arXiv.org:2609.28765v1", "author": "Gaurisankar Jayadas, Aske Plaat, \\'Alvaro Serra-G\\'omez, Sandheep P", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Reinforcement Learning with Verifiable Rewards (RLVR) performs well on problems with clear rewards, such as mathematics and coding, but whether it also works where the reward is less clear remains open. The reason-over-search recipe applies RLVR to open-domain question answering, where retrieval grounds the answer and a match against the reference supplies the reward. So far it has been demonstrated on large models, and below one billion parameters only with distillation from a larger teacher. We test the recipe on a small model. We train Qwen3.5-0.8B with Group Relative Policy Optimization (GRPO) and an interleaved Wikipedia-search tool on MuSiQue, varying only the reward across three shapes over three seeds each, and we evaluate every checkpoint held-out on a seven-benchmark question-answering suite. The recipe works: the best run reaches 0.352 average exact match against a 0.092 untrained floor, a 3.8-fold gain, with no distillation step in the training loop. The reward shape also matters. The Search-R1-faithful exact-match-only reward is the worst of the three at every seed at the matched training horizon, and it is worst even on exact match, the metric it directly optimises. We conclude that the sparse exact-match reward, RLVR's default in mathematics and code, is the wrong starting point for models of this size. The reason-over-search setting can supply a suitable reward for RLVR on small models, but small-model RLVR needs its own reward-design study rather than a scaled-down copy of a large-model recipe.", "categories": [ "cs.AI", "cs.IR" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.28765", "tags": [ "cs.AI", "cs.IR" ] }, { "id": "ea47f4abfc90f7ab64c3", "title": "Agent Memory with Episodic Retrieval for Financial Decision-Making", "content_text": "Large language models (LLMs) have demonstrated strong capabilities in financial analysis and reasoning, inspiring recent advances in agent-based trading frameworks. While these systems show promise, prior approaches either emphasize long-horizon forecasting or operate as stateless analyzers, limiting their applicability to the demands of trading in complicated settings. To address these gaps, we introduce META (Memory Enhanced Trading Agent), the first RAG-like episodic-memory-augmented…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "ea47f4abfc90f7ab64c3", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.28771v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Agent Memory with Episodic Retrieval for Financial Decision-Making", "item_type": "entry", "summary": "added: Agent Memory with Episodic Retrieval for Financial Decision-Making", "after": { "title": "Agent Memory with Episodic Retrieval for Financial Decision-Making", "link": "https://arxiv.org/abs/2609.28771", "id": "oai:arXiv.org:2609.28771v1", "author": "Nuoyue Xu, Jiang Liu, Wenxuan Huang, Xiang Zhang, Juntai Cao, Jiaqi Wei", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Large language models (LLMs) have demonstrated strong capabilities in financial analysis and reasoning, inspiring recent advances in agent-based trading frameworks. While these systems show promise, prior approaches either emphasize long-horizon forecasting or operate as stateless analyzers, limiting their applicability to the demands of trading in complicated settings. To address these gaps, we introduce META (Memory Enhanced Trading Agent), the first RAG-like episodic-memory-augmented multi-agent framework for financial decision making. META integrates a family of specialized indicator agents (e.g., Trend, MACD, Stochastic, RSI, SMA, AVWAP, Heikin-Ashi) with a Decision Agent that fuses their reports, and a Memory module that retrieves and updates past trading episodes encoded as market state embeddings with outcomes and reflections. By recalling relevant experiences and adaptively reweighting signals under similar market regimes, META achieves improved directional accuracy and robustness under short-horizon evaluation. Our results demonstrate that episodic memory provides a powerful mechanism for regime-aware, interpretable, and low-latency decision-making in trading and decision making. The code of this project is released on GitHub.", "categories": [ "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.28771", "tags": [ "cs.AI" ] }, { "id": "5fc07719bbcc31aa86b2", "title": "RECLAIM: Can Agents Reproduce the Claims of Machine Learning Papers?", "content_text": "Reproducing a machine learning paper involves most research steps, from installing software and debugging to running experiments, work that AI agents increasingly do. We introduce RECLAIM, a benchmark of 100 NeurIPS 2025 papers that can be rebuilt yearly from new conferences. For each paper we fix in advance the result to reproduce, what counts as a successful reproduction, and a GPU-hour budget. An agent must reproduce that result using the paper and whatever its authors released. What the…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "5fc07719bbcc31aa86b2", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.28850v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "RECLAIM: Can Agents Reproduce the Claims of Machine Learning Papers?", "item_type": "entry", "summary": "added: RECLAIM: Can Agents Reproduce the Claims of Machine Learning Papers?", "after": { "title": "RECLAIM: Can Agents Reproduce the Claims of Machine Learning Papers?", "link": "https://arxiv.org/abs/2609.28850", "id": "oai:arXiv.org:2609.28850v1", "author": "Mithil Salunkhe, Haochen Ding, Samridhi Verma, Volodymyr Kindratenko", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Reproducing a machine learning paper involves most research steps, from installing software and debugging to running experiments, work that AI agents increasingly do. We introduce RECLAIM, a benchmark of 100 NeurIPS 2025 papers that can be rebuilt yearly from new conferences. For each paper we fix in advance the result to reproduce, what counts as a successful reproduction, and a GPU-hour budget. An agent must reproduce that result using the paper and whatever its authors released. What the authors released decides the difficulty tier. Run-tier releases include code, data, and weights; Retrain-tier releases lack weights, so the agent trains the model; Reimplement-tier releases lack code, so the agent writes it. A separate language model grades runs from logs and outputs rather than agents' reports. We run four agents once per paper; the best agent in each tier reproduces only 41% of Run-tier papers, 27% at Retrain, and 15% at Reimplement, where every agent does worst. Failed attempts use on average 29% of their budget, so most stop with budget left. The most common agent error is writing the method without checking any part against the paper's numbers, in 63 of 400 runs.", "categories": [ "cs.AI", "cs.LG", "cs.SE" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.28850", "tags": [ "cs.AI", "cs.LG", "cs.SE" ] }, { "id": "c4ac4cd81b22c5835b92", "title": "Forecast-Dojo: Replayable Environments for Benchmarking and Training LLM Forecasting Agents", "content_text": "We introduce Forecast-Dojo, a replayable environment for benchmarking and training LLM forecasting agents. It combines resolved prediction-market questions with dated news, allowing agents to research an event and revisit their predictions at successive historical dates. The same tasks and tools support repeated evaluation, collection of training interactions, and feedback from recorded outcomes without waiting for new events to resolve. Forecast-Dojo contains 1,568 Polymarket events, split by…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "c4ac4cd81b22c5835b92", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.28876v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Forecast-Dojo: Replayable Environments for Benchmarking and Training LLM Forecasting Agents", "item_type": "entry", "summary": "added: Forecast-Dojo: Replayable Environments for Benchmarking and Training LLM Forecasting Agents", "after": { "title": "Forecast-Dojo: Replayable Environments for Benchmarking and Training LLM Forecasting Agents", "link": "https://arxiv.org/abs/2609.28876", "id": "oai:arXiv.org:2609.28876v1", "author": "Liqin Ye, Haorui Wang, Fardin Ahmed, Rongzhi Zhang, Yuan He, Ziyuan Lin, Yanbin Yin, Jing Peng, Michael Galarnyk, Sudheer Chava, Chao Zhang", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "We introduce Forecast-Dojo, a replayable environment for benchmarking and training LLM forecasting agents. It combines resolved prediction-market questions with dated news, allowing agents to research an event and revisit their predictions at successive historical dates. The same tasks and tools support repeated evaluation, collection of training interactions, and feedback from recorded outcomes without waiting for new events to resolve. Forecast-Dojo contains 1,568 Polymarket events, split by time into training and evaluation periods, and 18.8M dated news articles. In an evaluation of 12 models, research tools lower Brier score for all 12. Forecasts also improve as events unfold, with the largest gains at steps where more newly dated evidence is recorded. Every model still trails historical market forecasts in both Brier score and accuracy. A belief notebook carried between dates lowers research cost but does not consistently improve forecast quality. Beyond evaluation, Forecast-Dojo provides interaction trajectories and outcome feedback for agent learning, with supervised fine-tuning as a proof of concept.", "categories": [ "cs.AI", "cs.LG" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.28876", "tags": [ "cs.AI", "cs.LG" ] }, { "id": "76397b27e0c6bc46bf9c", "title": "Control the Harness, Control the Cost: Routing and Governing AI Coding Agents in the Enterprise", "content_text": "Harnesses, the products that run AI coding agents, are multiplying, and enterprises are rolling them out to their employees: what started as pilots with a few hundred seats is scaling to tens of thousands. Most enterprises do not build these harnesses but buy them from large vendors, such as Anthropic's Claude Code or OpenAI's Codex. A harness decides which model answers, what the model reads, how the prompt cache is used and which subagents run, so it picks the rate on the price sheet and sets…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "76397b27e0c6bc46bf9c", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.28919v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Control the Harness, Control the Cost: Routing and Governing AI Coding Agents in the Enterprise", "item_type": "entry", "summary": "added: Control the Harness, Control the Cost: Routing and Governing AI Coding Agents in the Enterprise", "after": { "title": "Control the Harness, Control the Cost: Routing and Governing AI Coding Agents in the Enterprise", "link": "https://arxiv.org/abs/2609.28919", "id": "oai:arXiv.org:2609.28919v1", "author": "Arian Abbasi, Alan Aqrawi, Ted Kwartler", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Harnesses, the products that run AI coding agents, are multiplying, and enterprises are rolling them out to their employees: what started as pilots with a few hundred seats is scaling to tens of thousands. Most enterprises do not build these harnesses but buy them from large vendors, such as Anthropic's Claude Code or OpenAI's Codex. A harness decides which model answers, what the model reads, how the prompt cache is used and which subagents run, so it picks the rate on the price sheet and sets the volume bought at it. Enterprises that keep a proprietary or untuned harness at its defaults inherit these choices and their bill. We build a fast, customisable router in which Jev, a classifier with calibrated probabilities, labels every prompt against a bring-your-own taxonomy of agentic requests. Because one user turn is many requests over a prompt cache that belongs to one model, the router moves work only where no running conversation has to rebuild its cache: at session start, in side lanes and at subagent launch. From the price sheet we derive when a mid-task switch pays back, and a crossover: on long tool-heavy sessions the highest-priced model costs less than the next tier, as repricing about 10,000 real sessions from public datasets confirms. In an emulated enterprise of 10,000 seats with user behaviour taken from these datasets, the router recovers 14 to 21% of model spend at Anthropic's list prices of 21 September 2026, \\$3.3M to \\$5.0M a year. The paper also maps the risks across twenty harnesses, prices the dependence on one vendor's models, and proposes a control plane that enterprises can run from within, starting now, with a ladder for deciding later whether to own the harness.", "categories": [ "cs.AI", "cs.CR" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.28919", "tags": [ "cs.AI", "cs.CR" ] }, { "id": "5fd23587eb61d0e0c741", "title": "PFArena: Benchmarking Language Models for Protein Modification", "content_text": "Protein modification requires navigating an immense sequence space, yet wet-lab validation remains low-throughput and costly. Although computational paradigms including protein language models (PLMs), large language models (LLMs), and LLM-based agents have shown promise in protein modification, their relative efficacy across realistic experimental decision-making settings remains unclear. To bridge this gap, we introduce PFArena, a benchmark comprising four controlled task interfaces that cover…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "5fd23587eb61d0e0c741", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.28921v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "PFArena: Benchmarking Language Models for Protein Modification", "item_type": "entry", "summary": "added: PFArena: Benchmarking Language Models for Protein Modification", "after": { "title": "PFArena: Benchmarking Language Models for Protein Modification", "link": "https://arxiv.org/abs/2609.28921", "id": "oai:arXiv.org:2609.28921v1", "author": "Yawen Ouyang, Xinbo Zhang, Ziyuan Ma, Yixin Wu, Wenbin Liao, Feiran Zhang, Wenjie Li, Lihao Wang, Hao Wang, Xiaoqing Zheng, Xuefeng Yan, Lei Bai, Ya-Qin Zhang, Shuyi Zhang, Wei-Ying Ma, Dahua Lin, Bowen Zhou, Hao Zhou", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Protein modification requires navigating an immense sequence space, yet wet-lab validation remains low-throughput and costly. Although computational paradigms including protein language models (PLMs), large language models (LLMs), and LLM-based agents have shown promise in protein modification, their relative efficacy across realistic experimental decision-making settings remains unclear. To bridge this gap, we introduce PFArena, a benchmark comprising four controlled task interfaces that cover single-mutant generation and multi-mutant ranking. By providing varying levels of mutation fitness data, PFArena reflects four representative research scenarios characterized by differing degrees of prior experimental context. We assess six PLMs, six LLMs, and five LLM-based agents using complementary metrics to measure both peak and overall protein modification performance. Our evaluation reveals that model performance shifts systematically with the availability of target-specific experimental evidence: PLMs demonstrate proficiency in open-ended single-mutant generation by leveraging protein-specific priors, whereas LLMs and agents perform strongly in multi-mutant ranking, particularly when target-specific fitness data are available. Nevertheless, all model families face fundamental challenges with increasing search-space size and mutation depth. We release our code and benchmark suite to facilitate reproducible research in model-assisted protein modification.", "categories": [ "cs.AI", "q-bio.BM" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.28921", "tags": [ "cs.AI", "q-bio.BM" ] }, { "id": "b8d6516f4730d6839060", "title": "From Static Personal Values to Contextualized Personalization: Bayesian Personalized Value Alignment for LLMs", "content_text": "Personalized value alignment has become increasingly important as large language models (LLMs) are expected to accommodate diverse user preferences. However, existing methods typically align model outputs with a static value profile across prompts, overlooking that the salience of value dimensions varies substantially across contexts. Inspired by Lewin's Field Theory, which views human behavior as jointly shaped by personal dispositions and situational constraints, we model personal values as…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "b8d6516f4730d6839060", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.28942v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "From Static Personal Values to Contextualized Personalization: Bayesian Personalized Value Alignment for LLMs", "item_type": "entry", "summary": "added: From Static Personal Values to Contextualized Personalization: Bayesian Personalized Value Alignment for LLMs", "after": { "title": "From Static Personal Values to Contextualized Personalization: Bayesian Personalized Value Alignment for LLMs", "link": "https://arxiv.org/abs/2609.28942", "id": "oai:arXiv.org:2609.28942v1", "author": "Hanze Guo, Aixuan Song, Jing Yao, Xiangxu Zhang, Xiaoyuan Yi, Xing Xie, Xiao Zhou", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Personalized value alignment has become increasingly important as large language models (LLMs) are expected to accommodate diverse user preferences. However, existing methods typically align model outputs with a static value profile across prompts, overlooking that the salience of value dimensions varies substantially across contexts. Inspired by Lewin's Field Theory, which views human behavior as jointly shaped by personal dispositions and situational constraints, we model personal values as priors and context-dependent preferences as posteriors. We propose BaCVA, an inference-time Bayesian Context-aware personalized Value Alignment method that approximates posterior personalized preferences by integrating static personal values with scenario-specific value salience. BaCVA first estimates contextual value salience from generally normative responses, and then employs a dual-view personalization module to infer posterior preferences from complementary personal-value and scenario-driven perspectives. This Bayesian formulation enables more accurate and adaptive personalized value alignment while improving data efficiency via prior values. Extensive experiments on benchmarks demonstrate its superiority over strong baselines.", "categories": [ "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.28942", "tags": [ "cs.AI" ] }, { "id": "8ab9d97235f55583eb69", "title": "Back to the Definition: Estimating Step-Level Advantages via Trajectory Graphs for Agentic Reinforcement Learning", "content_text": "Group-based reinforcement learning (RL) methods, such as GRPO and its variants, have become a leading paradigm for training reasoning and agentic large language models (LLMs). While their group-normalized advantage estimation is reliable at the response level, it becomes systematically biased at the step level, since coarse-grained trajectory-level advantages are hard to accurately reflect the contribution of individual steps (i.e, failed trajectories may contain valuable steps). Revisiting the…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "8ab9d97235f55583eb69", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.28963v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Back to the Definition: Estimating Step-Level Advantages via Trajectory Graphs for Agentic Reinforcement Learning", "item_type": "entry", "summary": "added: Back to the Definition: Estimating Step-Level Advantages via Trajectory Graphs for Agentic Reinforcement Learning", "after": { "title": "Back to the Definition: Estimating Step-Level Advantages via Trajectory Graphs for Agentic Reinforcement Learning", "link": "https://arxiv.org/abs/2609.28963", "id": "oai:arXiv.org:2609.28963v1", "author": "Xincheng Yao, Haobo Fu, Weiming Liu, Chongyang Zhang", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Group-based reinforcement learning (RL) methods, such as GRPO and its variants, have become a leading paradigm for training reasoning and agentic large language models (LLMs). While their group-normalized advantage estimation is reliable at the response level, it becomes systematically biased at the step level, since coarse-grained trajectory-level advantages are hard to accurately reflect the contribution of individual steps (i.e, failed trajectories may contain valuable steps). Revisiting the foundational RL definition, we notice that GRPO's success on single-turn tasks stems from its advantage estimation strategy, which adheres to the basic definition: the mean reward of multiple actions sampled from the same state constitutes a credible state-value estimate. Extending the faithful estimation to step-level would in principle demand sampling multiple actions from each intermediate state, which is too costly on a per-state basis. To mitigate this issue, we propose a Graph-based Faithful sTep-level credit-assignment framework (GRAFT) that grafts all rollout trajectories into a trajectory graph, recovering node state-values via Bellman iteration on the graph, and assigning credit to each edge by the node value difference. Theoretically, the estimated step-level advantage faithfully adheres to the basic advantage definition in RL. To further ensure the reliability of step-level advantage estimation, we further propose Graph GAE, which extends GAE to the trajectory graph for reducing the impact of state-value estimation bias. Experiments across a range of multi-turn agentic benchmarks show consistent gains over GRPO and superior performance compared to recent agentic RL algorithms. Code will be available at https://github.com/xcyao00/GRAFT.", "categories": [ "cs.AI", "stat.ML" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.28963", "tags": [ "cs.AI", "stat.ML" ] }, { "id": "12c08abfbb9751de8e83", "title": "AlphaDiverse: Post-Training Local Quantitative Research Agents for Diverse Exploration in Alpha Factor Mining", "content_text": "Large language model (LLM)-based multi-agent systems can automate alpha factor mining, but their reliance on external APIs limits control over cost, availability, and confidentiality. Long research loops also tend to revisit a few successful economic mechanisms that lead to research path collapse. To address these limitations, we propose AlphaDiverse, a framework that integrates a multi-agent alpha research system, diverse research path collection, and post-training for local agents. We let the…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "12c08abfbb9751de8e83", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29014v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "AlphaDiverse: Post-Training Local Quantitative Research Agents for Diverse Exploration in Alpha Factor Mining", "item_type": "entry", "summary": "added: AlphaDiverse: Post-Training Local Quantitative Research Agents for Diverse Exploration in Alpha Factor Mining", "after": { "title": "AlphaDiverse: Post-Training Local Quantitative Research Agents for Diverse Exploration in Alpha Factor Mining", "link": "https://arxiv.org/abs/2609.29014", "id": "oai:arXiv.org:2609.29014v1", "author": "Qingzhuo Wang, Zikun Wei, Zhihua Wei, Wen Shen", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Large language model (LLM)-based multi-agent systems can automate alpha factor mining, but their reliance on external APIs limits control over cost, availability, and confidentiality. Long research loops also tend to revisit a few successful economic mechanisms that lead to research path collapse. To address these limitations, we propose AlphaDiverse, a framework that integrates a multi-agent alpha research system, diverse research path collection, and post-training for local agents. We let the research system generate complementary plan portfolios and vary research environments across loops to collect diverse research paths. Using these diverse traces, we warm-start local Planner and Realizer agents with supervised fine-tuning. Then, we propose a joint GRPO method to optimize both of them using predictive quality and diversity of contributions. Research feedback is confined to inner period data, while a frozen final model is evaluated on a later outer period data, thereby avoiding test-set tuning. Experiments across four Chinese stock universes show that AlphaDiverse can combine competitive prediction with broader exploration.", "categories": [ "cs.AI", "cs.CE", "cs.MA" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29014", "tags": [ "cs.AI", "cs.CE", "cs.MA" ] }, { "id": "946a1f5959ada0aa6675", "title": "From Self-Distillation to Self-Practice: Privileged Information for Multi-Turn Agents", "content_text": "On-policy self-distillation (OPSD) has become a popular recipe for post-training LLM agents. It supervises the agent model at the token level with a stronger teacher view of the same model, obtained by conditioning on privileged information (PI). In this work, we show that in multi-turn agents, this paradigm teaches the student to act with confidence but without the information behind it. The trained agent behaves as if it had privileged information it never observed, and its performance falls…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "946a1f5959ada0aa6675", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29051v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "From Self-Distillation to Self-Practice: Privileged Information for Multi-Turn Agents", "item_type": "entry", "summary": "added: From Self-Distillation to Self-Practice: Privileged Information for Multi-Turn Agents", "after": { "title": "From Self-Distillation to Self-Practice: Privileged Information for Multi-Turn Agents", "link": "https://arxiv.org/abs/2609.29051", "id": "oai:arXiv.org:2609.29051v1", "author": "Xingyu Su, Abhishek Kumar, Qing Ping, Youzhi Luo, Jonathan Buck, Zach Zhang, Subramanian Chidambaram, Vinayak Arannil", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "On-policy self-distillation (OPSD) has become a popular recipe for post-training LLM agents. It supervises the agent model at the token level with a stronger teacher view of the same model, obtained by conditioning on privileged information (PI). In this work, we show that in multi-turn agents, this paradigm teaches the student to act with confidence but without the information behind it. The trained agent behaves as if it had privileged information it never observed, and its performance falls well short of plain RL, in the worst case below the untrained base model. Therefore, we propose Privileged Self-Practice (PSP), which keeps the PI and moves it from the loss to the sampler. When the student's rollouts on a task mostly fail, we inject a short per-task instruction written by an analyzer model, sample the task again with the instruction in context, and train on the result with an unchanged GRPO objective. The privileged information stays in the prompt and never enters the loss. Across AppWorld and SWE-bench Verified, with three different student models, PSP obtains the best average score in every setting and is the only method that consistently outperforms plain GRPO, improving task-goal completion by up to 65% on AppWorld and the resolved rate by up to 61% on SWE-bench Verified.", "categories": [ "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29051", "tags": [ "cs.AI" ] }, { "id": "e2fb368e13184c55e569", "title": "CounterRoute: Self-Routed Reasoning via Hierarchical Counterfactual Credit Assignment", "content_text": "Reasoning-capable language models often produce long chains of thought when direct answers suffice, wasting inference compute. Many dual-mode models leave this choice to users. Automating it is challenging because routing targets evolve with the policy, initial mode preferences destabilize exploration, and sequence-level objectives entangle routing with response learning. We introduce CounterRoute, an online reinforcement-learning framework that jointly learns routing and modeconditioned…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "e2fb368e13184c55e569", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29109v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "CounterRoute: Self-Routed Reasoning via Hierarchical Counterfactual Credit Assignment", "item_type": "entry", "summary": "added: CounterRoute: Self-Routed Reasoning via Hierarchical Counterfactual Credit Assignment", "after": { "title": "CounterRoute: Self-Routed Reasoning via Hierarchical Counterfactual Credit Assignment", "link": "https://arxiv.org/abs/2609.29109", "id": "oai:arXiv.org:2609.29109v1", "author": "Ruochen Jiao, Besnik Fetahu, Zhenyu Shi, Priyanka Nigam", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Reasoning-capable language models often produce long chains of thought when direct answers suffice, wasting inference compute. Many dual-mode models leave this choice to users. Automating it is challenging because routing targets evolve with the policy, initial mode preferences destabilize exploration, and sequence-level objectives entangle routing with response learning. We introduce CounterRoute, an online reinforcement-learning framework that jointly learns routing and modeconditioned responses in one shared policy directly from a native dual-mode checkpoint, without method-specific SFT warm-up. Paired current-policy counterfactual rollouts assign cross-mode credit only to the routing token, while within-mode GRPO trains response tokens. A paired-to-self-routed curriculum stabilizes early training with forced rollouts from both modes, then increases self-routed updates to improve autonomous routing. Across nine benchmarks, CounterRoute better balances accuracy and efficiency than heuristic and learned adaptive-routing methods. Relative to always-thinking checkpoints, it improves macro-average accuracy while reducing mean generated tokens by 51% for Qwen3-8B and 41% for Qwen3-14B. On instruction-following and commonsense benchmarks where direct answering is strong, think rates fall as low as 1% while response quality improves. Despite training only on math and instruction following, its routing behavior and response quality generalize to held-out coding, science, knowledge, and commonsense benchmarks.", "categories": [ "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29109", "tags": [ "cs.AI" ] }, { "id": "1a22f54c22f4b09dc42d", "title": "Scope Before You Persist: Preventing Cross-Family Interference in Agent Memory", "content_text": "Persistent memory lets language-model agents improve prompts and skills without updating model weights. We show that matching retrieval scope to certification scope enables these edits to support reliable repeated adaptation across recurring task families. We study frozen-model agents on ProcStream-RSI, a 12-round code-repair stream, using Orthogonal Regression Control (ORC), an execution-grounded gate for persistent skill edits. In an intervention that holds proposals and gate decisions fixed…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "1a22f54c22f4b09dc42d", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29144v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Scope Before You Persist: Preventing Cross-Family Interference in Agent Memory", "item_type": "entry", "summary": "added: Scope Before You Persist: Preventing Cross-Family Interference in Agent Memory", "after": { "title": "Scope Before You Persist: Preventing Cross-Family Interference in Agent Memory", "link": "https://arxiv.org/abs/2609.29144", "id": "oai:arXiv.org:2609.29144v1", "author": "Yezhou Cheng, Runjia Du, Zeming Liu, Qibai Chen, Hang Lyu, Yankai Zeng, Yilan Wei, Bojun Lin", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Persistent memory lets language-model agents improve prompts and skills without updating model weights. We show that matching retrieval scope to certification scope enables these edits to support reliable repeated adaptation across recurring task families. We study frozen-model agents on ProcStream-RSI, a 12-round code-repair stream, using Orthogonal Regression Control (ORC), an execution-grounded gate for persistent skill edits. In an intervention that holds proposals and gate decisions fixed, retrieving each accepted skill only for its originating family raises mean hidden trajectory utility from 0.713 under global memory to 0.816 and changes harmful deployments from six of eight to none. In 27 paired randomized-order streams, Scoped-ORC improves mean trajectory utility by 0.063 [0.037, 0.094] over Global-ORC, accepts 63 rather than 12 updates, and produces multiple accepted updates in 19/27 streams, with 0/63 harmful acceptances. The global control reaches 0.713, below the static agent's 0.775, because locally valid edits can interfere with unrelated families. These results establish scope matching as a complementary control for persistent agent memory: certification determines whether an edit is supported, while retrieval scope determines where that evidence authorizes its use.", "categories": [ "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29144", "tags": [ "cs.AI" ] }, { "id": "33f9c05dcc249af4b064", "title": "A Wrong Turn Does Not Ruin the Journey: Deviation-Guided Skill Self-Evolution for LLM Agents", "content_text": "Large language model agents increasingly rely on natural-language skills to solve complex tool-use tasks. However, such tasks often admit multiple valid solution paths, making it inappropriate to improve skills by forcing failed trajectories to match a fixed successful trajectory. Moreover, failed trajectories are rarely entirely wrong: an agent may first collect useful evidence and make meaningful progress, but later deviate into an erroneous suffix. We therefore argue that skill…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "33f9c05dcc249af4b064", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29154v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "A Wrong Turn Does Not Ruin the Journey: Deviation-Guided Skill Self-Evolution for LLM Agents", "item_type": "entry", "summary": "added: A Wrong Turn Does Not Ruin the Journey: Deviation-Guided Skill Self-Evolution for LLM Agents", "after": { "title": "A Wrong Turn Does Not Ruin the Journey: Deviation-Guided Skill Self-Evolution for LLM Agents", "link": "https://arxiv.org/abs/2609.29154", "id": "oai:arXiv.org:2609.29154v1", "author": "Yichun Feng, Jiawei Wang, Haozhe Sun", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Large language model agents increasingly rely on natural-language skills to solve complex tool-use tasks. However, such tasks often admit multiple valid solution paths, making it inappropriate to improve skills by forcing failed trajectories to match a fixed successful trajectory. Moreover, failed trajectories are rarely entirely wrong: an agent may first collect useful evidence and make meaningful progress, but later deviate into an erroneous suffix. We therefore argue that skill self-evolution should identify where productive problem solving begins to break down, rather than reflect coarsely over the entire failure. Based on this insight, we propose SkillPivot, a deviation-point-guided framework for skill self-evolution. SkillPivot detects the transition from a useful prefix to an erroneous suffix using execution validity, goal progress, and action diversity. A stronger teacher then continues from the same prefix and produces a successful alternative under the same interaction history. By contrasting the student's failed suffix with the teacher's successful suffix, SkillPivot generates localized skill updates while preserving already effective guidance. Experiments on ToolQA, LogicBench, and WildClawBench show that SkillPivot consistently outperforms competing skill-evolution methods, improves multiple agent models, and produces compact, transferable skill updates.", "categories": [ "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29154", "tags": [ "cs.AI" ] }, { "id": "8b4d217db828ccfbafa0", "title": "IndicBankBench: Evaluating Safety and Reliability of Language Model Assistants in Indian Retail Banking", "content_text": "Banking assistants must use account-specific information to answer requests and, in many cases, take actions through tools. Evaluating only the final response misses important errors. An assistant may ask for information it already has, rely on stale context, select the wrong account, or write an invalid value after stating the correct one. We introduce IndicBankBench, a 799-case benchmark for Indian retail banking spanning five operational domains, a capability/refusal domain, and twenty…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "8b4d217db828ccfbafa0", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29167v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "IndicBankBench: Evaluating Safety and Reliability of Language Model Assistants in Indian Retail Banking", "item_type": "entry", "summary": "added: IndicBankBench: Evaluating Safety and Reliability of Language Model Assistants in Indian Retail Banking", "after": { "title": "IndicBankBench: Evaluating Safety and Reliability of Language Model Assistants in Indian Retail Banking", "link": "https://arxiv.org/abs/2609.29167", "id": "oai:arXiv.org:2609.29167v1", "author": "Suvradip Paul, Chandra Bhushan, Harsh Sharma, Nitin Kukreja, Yatharth Dedhia, Keyur Doshi, Prashant Devadiga", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Banking assistants must use account-specific information to answer requests and, in many cases, take actions through tools. Evaluating only the final response misses important errors. An assistant may ask for information it already has, rely on stale context, select the wrong account, or write an invalid value after stating the correct one. We introduce IndicBankBench, a 799-case benchmark for Indian retail banking spanning five operational domains, a capability/refusal domain, and twenty primary axes. Cases are evaluated at four stages: safety, action and tool use, response adequacy, and advisory quality. Tool use and most safety checks are deterministic. A narrow resolver handles only ambiguous confirmation-before-write cases, while a separate LLM judge evaluates semantic response adequacy. We run every case three times and report strict pass^3, which requires success on all trials. Across the eleven evaluated models, strict reliability ranges from 43.7% to 58.2%, whereas at-least-once success ranges from 60% to 74%. This gap shows that at-least-once success can overstate dependable banking behavior. The case-level diagnostics also distinguish systems that ask unnecessary questions from those that act but fail to reconcile customer context or fully resolve the request. We release the cases, mock environment, and evaluation harness.", "categories": [ "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29167", "tags": [ "cs.AI" ] }, { "id": "32896b53d94b5d26eafe", "title": "ASIRF: An Agentic Framework for Context-Dependent Sensitive Information Redaction", "content_text": "Sensitive information is defined by domain and intent, not a universal category, yet redaction systems such as privacy filters and named-entity recognizers fix a taxonomy at training time, requiring retraining for each new domain. We introduce ASIRF (Agentic Sensitive Information Redaction Framework), which retrieves domain-specific definitions based on the input's domain from a flexible knowledge base at inference time, needing no retraining to adapt. Two architectures, a three-call…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "32896b53d94b5d26eafe", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29191v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "ASIRF: An Agentic Framework for Context-Dependent Sensitive Information Redaction", "item_type": "entry", "summary": "added: ASIRF: An Agentic Framework for Context-Dependent Sensitive Information Redaction", "after": { "title": "ASIRF: An Agentic Framework for Context-Dependent Sensitive Information Redaction", "link": "https://arxiv.org/abs/2609.29191", "id": "oai:arXiv.org:2609.29191v1", "author": "Sudha Priyadarshini, Mohamed Chahine Ghanem", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Sensitive information is defined by domain and intent, not a universal category, yet redaction systems such as privacy filters and named-entity recognizers fix a taxonomy at training time, requiring retraining for each new domain. We introduce ASIRF (Agentic Sensitive Information Redaction Framework), which retrieves domain-specific definitions based on the input's domain from a flexible knowledge base at inference time, needing no retraining to adapt. Two architectures, a three-call multi-agent pipeline and a single-agent variant, are evaluated across ten small open-weight models and eight datasets, including out-of-distribution fictional domains, against the OpenAI Privacy Filter (OPF) as a trained-classifier baseline. With only a few dozen expert-authored definitions per domain and no training data, ASIRF's recall exceeds OPF's in 68 of 80 model-domain combinations (85 percent), by at least one of the two architectures, with shortfalls confined mostly to OPF's training-distribution domains.", "categories": [ "cs.AI", "cs.IR", "cs.LG" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29191", "tags": [ "cs.AI", "cs.IR", "cs.LG" ] }, { "id": "d35c50137d3dabf8a963", "title": "Towards An LLM-Driven Unified Conversion Framework for BT and FSM in Autonomous Intelligent Systems", "content_text": "Finite state machine (FSM) and behavior trees (BT) are widely adopted behavioral modeling paradigms for autonomous intelligent systems. While functionally equivalent and inter-convertible in principle, existing transformation methods between FSM and BT face major challenges in preserving behavioral completeness and avoiding model complexity explosion. To overcome these issues, we propose an LLM-driven unified conversion framework that enables automatic, efficient, and semantically consistent…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "d35c50137d3dabf8a963", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29228v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Towards An LLM-Driven Unified Conversion Framework for BT and FSM in Autonomous Intelligent Systems", "item_type": "entry", "summary": "added: Towards An LLM-Driven Unified Conversion Framework for BT and FSM in Autonomous Intelligent Systems", "after": { "title": "Towards An LLM-Driven Unified Conversion Framework for BT and FSM in Autonomous Intelligent Systems", "link": "https://arxiv.org/abs/2609.29228", "id": "oai:arXiv.org:2609.29228v1", "author": "Zhang Qi, Yang Shuo, Zhu Zhengqiu, Zhou Peng, Jiao Peng", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Finite state machine (FSM) and behavior trees (BT) are widely adopted behavioral modeling paradigms for autonomous intelligent systems. While functionally equivalent and inter-convertible in principle, existing transformation methods between FSM and BT face major challenges in preserving behavioral completeness and avoiding model complexity explosion. To overcome these issues, we propose an LLM-driven unified conversion framework that enables automatic, efficient, and semantically consistent transformation between FSM and BT. Specifically, a novel loop execution BT structure is designed for LLM to accurately capture the loop structure in FSM, thereby preserving behavioral completeness. To mitigate the state explosion problem in BT-to-FSM conversion, a depth compression strategy is introduced with LLM prompt to eliminate redundant control nodes, complemented by differentiated hierarchical conversion rules that collectively reduce the number of required sub-FSM. Simulation experiments in multiple autonomous decision-making scenarios demonstrate that the proposed framework enables an accurate and automated bidirectional conversion between FSM and BT. Furthermore, it significantly enhances the scalability and maintainability of generated models compared to traditional approaches, providing a practical solution for behavior model conversion in consumer-grade autonomous intelligent systems such as service robots, game agents, and smart home devices", "categories": [ "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29228", "tags": [ "cs.AI" ] }, { "id": "dada8f3cef171ca8fdbd", "title": "SkinAgent AI: A Safety-Grounded Multimodal Agentic Framework for Non-Diagnostic Skincare Support", "content_text": "Consumer-facing skincare AI must coordinate visual evidence, product information, tool use, and user-facing actions within explicit evidence and safety boundaries. This study evaluates SkinAgent AI, a non-diagnostic multimodal framework that combines visual concern routing with grounded and auditable LLM-based orchestration. The architecture includes routing for Acne, Pores, and Wrinkles; photograph-based skin-type estimation; count-informed ordinal acne-severity support; typed tools…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "dada8f3cef171ca8fdbd", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29341v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "SkinAgent AI: A Safety-Grounded Multimodal Agentic Framework for Non-Diagnostic Skincare Support", "item_type": "entry", "summary": "added: SkinAgent AI: A Safety-Grounded Multimodal Agentic Framework for Non-Diagnostic Skincare Support", "after": { "title": "SkinAgent AI: A Safety-Grounded Multimodal Agentic Framework for Non-Diagnostic Skincare Support", "link": "https://arxiv.org/abs/2609.29341", "id": "oai:arXiv.org:2609.29341v1", "author": "Muhammad Muhtasim Shahriar, Abdullah Mohammad Sayem, Tze Hui Liew, M. F. Mridha, Md. Mahiuddin", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Consumer-facing skincare AI must coordinate visual evidence, product information, tool use, and user-facing actions within explicit evidence and safety boundaries. This study evaluates SkinAgent AI, a non-diagnostic multimodal framework that combines visual concern routing with grounded and auditable LLM-based orchestration. The architecture includes routing for Acne, Pores, and Wrinkles; photograph-based skin-type estimation; count-informed ordinal acne-severity support; typed tools; database-grounded recommendation and action functions; deterministic safety, privacy, and evidence checks; approval before state-changing actions; and structured trace and replay mechanisms. Visual-model performance and system-level agent behavior were evaluated separately. Across three seeds, the skin-condition routing model achieved 99.84% +/- 0.07% accuracy. Skin-type estimation achieved 88.85% accuracy, while count-informed acne-severity support achieved 84.59% accuracy with a quadratic weighted kappa of 0.9076. On a locked but non-independent 240-case system benchmark, intent accuracy was 80.00%, exact tool-set match was 62.92%, and strict task completion was 47.08%. No violations or successful cross-user leakage events were observed in the finite safety and privacy test suites. Tool-selection errors, incomplete grounding of product attributes, and unreliable failure fallback nevertheless remained. These findings support the feasibility of bounded, database-grounded, and traceable agent orchestration for non-diagnostic skincare assistance. They do not establish clinical readiness, external generalization, formal privacy guarantees, or universal safety. Independent validation, expert assessment, robustness and fairness testing, and prospective evaluation in real-world settings remain necessary.", "categories": [ "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29341", "tags": [ "cs.AI" ] }, { "id": "c7ed9ba7bca76c914c94", "title": "Epistemic-Probabilistic Model for Guarded Multi-Agent LLM Coordination", "content_text": "Multi-agent large language models (LLMs) have become ubiquitous in applied AI, yet their theoretical foundations remain surprisingly understudied. Viewed through the lens of multi-agent systems theory, several shortcomings come to light: a lack of social intelligence, the absence of coordination mechanisms among agents, unknown emergent behavior, and interactions between agents that are bounded by natural language. We address two of these gaps: the absence of social behavior and the lack of…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "c7ed9ba7bca76c914c94", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29366v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Epistemic-Probabilistic Model for Guarded Multi-Agent LLM Coordination", "item_type": "entry", "summary": "added: Epistemic-Probabilistic Model for Guarded Multi-Agent LLM Coordination", "after": { "title": "Epistemic-Probabilistic Model for Guarded Multi-Agent LLM Coordination", "link": "https://arxiv.org/abs/2609.29366", "id": "oai:arXiv.org:2609.29366v1", "author": "Mehdi Nasiri, Mohammad Saeed Arvenaghi, Sadegh Vaezi, Ebrahim Ardeshir-Larijani", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Multi-agent large language models (LLMs) have become ubiquitous in applied AI, yet their theoretical foundations remain surprisingly understudied. Viewed through the lens of multi-agent systems theory, several shortcomings come to light: a lack of social intelligence, the absence of coordination mechanisms among agents, unknown emergent behavior, and interactions between agents that are bounded by natural language. We address two of these gaps: the absence of social behavior and the lack of mechanisms for inter-agent coordination. We introduce Epistemic Probabilistic Language Agents (EPLA), a neuro-symbolic architecture for multi-agent coordination under uncertainty. A Symbolic Guard provides structured diagnostic feedback. The LLM generates typed actions, and the Guard controls their execution against an authoritative symbolic state. We formalize the epistemic layer in a gossip testbed through epistemic lottery gossip models, which combine view-based call histories with agent-indexed probability weights. We argue that implementing such a formalism can address shortcomings of agentic LLMs.", "categories": [ "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29366", "tags": [ "cs.AI" ] }, { "id": "80fb4d8344682f2792bf", "title": "BiGraph-Diffuse: A Bidirectional Diffusion Language Model with Graph-Structured Retrieval For Mental Health Counseling", "content_text": "Mental health disorders affect hundreds of millions of people around the world, yet access to professional counseling remains severely limited. AI-powered dialogue systems offer a scalable alternative, but existing models face two fundamental challenges. First, they lack the bidirectional understanding needed to capture the layered nature of emotional expression, particularly in cases of progressive disclosure, where clients often present symptoms at the surface-level while concealing deeper…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "80fb4d8344682f2792bf", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29519v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "BiGraph-Diffuse: A Bidirectional Diffusion Language Model with Graph-Structured Retrieval For Mental Health Counseling", "item_type": "entry", "summary": "added: BiGraph-Diffuse: A Bidirectional Diffusion Language Model with Graph-Structured Retrieval For Mental Health Counseling", "after": { "title": "BiGraph-Diffuse: A Bidirectional Diffusion Language Model with Graph-Structured Retrieval For Mental Health Counseling", "link": "https://arxiv.org/abs/2609.29519", "id": "oai:arXiv.org:2609.29519v1", "author": "Yuxiang Cheng, Quanwei Tang, Lvhui Lu, Dong Zhang, Shoushan Li, Erik Cambria", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Mental health disorders affect hundreds of millions of people around the world, yet access to professional counseling remains severely limited. AI-powered dialogue systems offer a scalable alternative, but existing models face two fundamental challenges. First, they lack the bidirectional understanding needed to capture the layered nature of emotional expression, particularly in cases of progressive disclosure, where clients often present symptoms at the surface-level while concealing deeper trauma. Autoregressive (AR) models process information sequentially and cannot revise early interpretations when new evidence emerges later in the conversation. Second, they fail to effectively incorporate the relational knowledge that underlies clinical reasoning. In this paper, we propose \\textbf{BiGraph-Diffuse}, the first large-scale diffusion language model tailored for the counseling domain. We further introduce \\textbf{BiGraph-RAG}, a relation-free graph-structured retrieval strategy that relies only on lightweight entity extraction and semantic linking. This design preserves inferential pathways from observable symptoms to potential underlying causes, while incurring zero LLM token cost during indexing. Importantly, these two modules are not merely combined but mutually reinforcing. The diffusion model provides a holistic bidirectional context, enabling the system to defer premature judgments during progressive disclosure. Meanwhile, graph-based retrieval captures the structured interconnections of clinical knowledge. Extensive experiments demonstrate the effectiveness of BiGraph-Diffuse, and we further provide a solid theoretical analysis to support its design.", "categories": [ "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29519", "tags": [ "cs.AI" ] }, { "id": "905ca2c349c2f11e3780", "title": "Stale Does Not Mean Unsafe: Guard Precision for Tool-Using LLM Agents under Infrastructure State Races", "content_text": "Tool-using language-model agents increasingly mutate schedulers, data pipelines, object stores, and access-control systems. Between an agent's read and its commit, external state can change, but not every change makes the commit unsafe. We separate invalidating races, which break a declared safety predicate, from predicate-preserving and irrelevant races, and ask how precisely runtime guards distinguish them. Our deterministic simulator separates visible from authoritative state and injects…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "905ca2c349c2f11e3780", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29522v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Stale Does Not Mean Unsafe: Guard Precision for Tool-Using LLM Agents under Infrastructure State Races", "item_type": "entry", "summary": "added: Stale Does Not Mean Unsafe: Guard Precision for Tool-Using LLM Agents under Infrastructure State Races", "after": { "title": "Stale Does Not Mean Unsafe: Guard Precision for Tool-Using LLM Agents under Infrastructure State Races", "link": "https://arxiv.org/abs/2609.29522", "id": "oai:arXiv.org:2609.29522v1", "author": "Zihao Zheng, Jiayu Long, Baichuan Li, Junyi Yao", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Tool-using language-model agents increasingly mutate schedulers, data pipelines, object stores, and access-control systems. Between an agent's read and its commit, external state can change, but not every change makes the commit unsafe. We separate invalidating races, which break a declared safety predicate, from predicate-preserving and irrelevant races, and ask how precisely runtime guards distinguish them. Our deterministic simulator separates visible from authoritative state and injects five non-atomic failure mechanisms across 16 infrastructure tasks in four domains; frozen agent proposals are replayed counterfactually under every controller without an LLM judge. We evaluate three commit-time guard granularities (global epoch, read-set version, semantic commit predicate), multi-level verification, and model-side gates on three locally hosted quantized model families (Qwen3-4B, Phi-4-mini, Gemma4-8B; 3,456 trajectories on one GPU). All three guards eliminate unsafe commits, but their availability differs sharply: freshness-based guards needlessly block 92-95% of benign races, forfeiting up to 43% of safe task completions, while the complete predicate guard blocks none. That precision is contract-dependent: deleting a single declared clause converts exactly its fault family into unsafe commits (up to 7.9%). Model-side signals do not substitute: verbal confidence is miscalibrated (ECE approximately 0.37), action agreement matches a random gate, a cautionary prompt leaves the direct unsafe rate essentially unchanged, and after a freshness-guard block agents re-commit unsafely from refreshed but still-incomplete reads. Under degraded telemetry a hidden concurrent mutation remains observationally clean, bounding every selective policy. Precise runtime enforcement therefore requires semantic contracts, not freshness heuristics or model self-assessment.", "categories": [ "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29522", "tags": [ "cs.AI" ] }, { "id": "c10005b2b3020c3519c0", "title": "Clinical Knowledge Graphs for Chest X-Ray Device Reasoning", "content_text": "Chest radiographs are routinely used to verify the position of catheters, tubes, and other support devices. Existing image models often return labels or segmentations, while report-processing systems structure text without access to image geometry. We present an uncertainty-aware clinical knowledge graph that represents device instances, tip estimates, placement assessments, provenance, report events, and temporal links as separate but connected evidence.\nWe evaluate the implemented visual…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "c10005b2b3020c3519c0", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29536v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Clinical Knowledge Graphs for Chest X-Ray Device Reasoning", "item_type": "entry", "summary": "added: Clinical Knowledge Graphs for Chest X-Ray Device Reasoning", "after": { "title": "Clinical Knowledge Graphs for Chest X-Ray Device Reasoning", "link": "https://arxiv.org/abs/2609.29536", "id": "oai:arXiv.org:2609.29536v1", "author": "Harshil Lodhiya", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Chest radiographs are routinely used to verify the position of catheters, tubes, and other support devices. Existing image models often return labels or segmentations, while report-processing systems structure text without access to image geometry. We present an uncertainty-aware clinical knowledge graph that represents device instances, tip estimates, placement assessments, provenance, report events, and temporal links as separate but connected evidence.\nWe evaluate the implemented visual graph layer using saved predictions from the complete RANZCR CLiP test archive, comprising 30,083 studies from 3,255 patients across five non-overlapping outer folds. The graph builder materializes 914,632 B7 evidence nodes and 884,549 typed relationships. All 118,647 B7 predicted-device nodes retain tip covariance, placement probabilities, fragment provenance, and fragment counts, whereas the direct B2 baseline retains none of these fields. We further define typed data contracts, uncertainty representations, abstention rules, report-image grounding, and longitudinal query mechanisms for extending the graph to report-bearing cohorts.\nThe reported graph-materialization analysis is post-hoc descriptive and does not establish report grounding, longitudinal performance, or clinical utility. It demonstrates a reproducible foundation for evidence-preserving AI reasoning over chest X-ray device assessments.", "categories": [ "cs.AI", "cs.CV" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29536", "tags": [ "cs.AI", "cs.CV" ] }, { "id": "744172019d00ce67cf9f", "title": "Safe Skill Retirement for Physical Agents", "content_text": "Agent skills bundle procedural guidance with execution conditions governing authority, user consent, and live environment state. When model capabilities advance, maintainers prune instructions that appear redundant on authorized benchmark tasks. However, authorized maintenance tests can leave dormant safety conditions untested. This mismatch creates an unmeasured support gap over physical and privacy-sensitive effects. We introduce matched authority counterfactuals that hold the requested…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "744172019d00ce67cf9f", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29543v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Safe Skill Retirement for Physical Agents", "item_type": "entry", "summary": "added: Safe Skill Retirement for Physical Agents", "after": { "title": "Safe Skill Retirement for Physical Agents", "link": "https://arxiv.org/abs/2609.29543", "id": "oai:arXiv.org:2609.29543v1", "author": "Zhonghao Zhan, Xiao Ma, Hamed Haddadi", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Agent skills bundle procedural guidance with execution conditions governing authority, user consent, and live environment state. When model capabilities advance, maintainers prune instructions that appear redundant on authorized benchmark tasks. However, authorized maintenance tests can leave dormant safety conditions untested. This mismatch creates an unmeasured support gap over physical and privacy-sensitive effects. We introduce matched authority counterfactuals that hold the requested action, tool parameters, and intended effect fixed while systematically varying a single governing predicate. We formalize this evaluation via a two-gate retirement certificate requiring a candidate reduction to preserve authorized utility within a declared margin while producing zero unauthorized protected effects. In controlled experiments spanning four frontier and local model configurations across twelve skill bundles (2,592 evaluation cells), task-certified reductions remove over 94% of skill clauses and preserve authorized completion, yet produce unauthorized protected effects in every bundle. Boundary enforcement eliminates protected effects on the declared audit but fails the utility gate for one configuration. One bounded combined protocol passes both gates across all four configurations, with zero utility headroom. An end-to-end check on one read-only Home Assistant camera chain verifies proposal, decision, and effect measurements on a real device. These results demonstrate that while task benchmarks can justify retiring procedural guidance, retirement decisions require explicitly auditing the authority contracts governing physical actions.", "categories": [ "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29543", "tags": [ "cs.AI" ] }, { "id": "60a5a955d9786ad4e5e9", "title": "ERRAND: Budgeted Maintenance of Agent Memory", "content_text": "Deployed agents run on handed-over knowledge: a frozen policy consults a briefing of consolidated items written before the stream begins. The world then moves while the store stands still: paths close, flags change, price bands move; every item was true at handover, and the failure is staleness, not ignorance. We introduce ERRAND, which treats revalidation as a priced errand: a recheck competes with the task it protects for the same scarce actions, funded only when the value per action of…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "60a5a955d9786ad4e5e9", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29545v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "ERRAND: Budgeted Maintenance of Agent Memory", "item_type": "entry", "summary": "added: ERRAND: Budgeted Maintenance of Agent Memory", "after": { "title": "ERRAND: Budgeted Maintenance of Agent Memory", "link": "https://arxiv.org/abs/2609.29545", "id": "oai:arXiv.org:2609.29545v1", "author": "Beining Wu, Zihao Ding, Jun Huang", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Deployed agents run on handed-over knowledge: a frozen policy consults a briefing of consolidated items written before the stream begins. The world then moves while the store stands still: paths close, flags change, price bands move; every item was true at handover, and the failure is staleness, not ignorance. We introduce ERRAND, which treats revalidation as a priced errand: a recheck competes with the task it protects for the same scarce actions, funded only when the value per action of resolving a doubt clears a running wage. The errand index is single-peaked, vanishing at both ends of belief, so certainty in either direction costs nothing; free en-route receipts maintain on-path knowledge, and repair writes a version, never a deletion. Under equal action budgets in two drifting tool-use worlds, ERRAND clears every non-oracle policy on the preregistered calibers, primary in every setting and conditional at every binding budget, leading eager revalidation by 10.0pp at the base cap. Restraint wins: given no cap, ERRAND stops on its own, spending 11.0% of steps, while uncapped eager revalidation spends 70.7% and still finishes 4.5pp behind capped ERRAND. The margin sits where the briefing's coverage is thinnest, the shadow price of long-tail knowledge: a small budget, well priced, beats a bigger store that never rechecks.", "categories": [ "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29545", "tags": [ "cs.AI" ] }, { "id": "0a4c3ed681ccf8eb3ebe", "title": "Is Reasoning Always Useful? Rethinking Reasoning Utility in Universal Multimodal Embeddings", "content_text": "Reasoning-enhanced universal multimodal embeddings (UME) improve heterogeneous retrieval, but plausible rationales do not necessarily produce discriminative rankings. We study this gap by comparing the discriminative (DISC) and reasoning-driven generative (GEN) branches of UME-R1, a state-of-the-art reasoning UME method. We decompose reasoning utility into positive-target gain, hard-negative gain, and their margin difference. Positive similarity increases for 56.6%, but 15.7% are false-helpful…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "0a4c3ed681ccf8eb3ebe", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29560v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Is Reasoning Always Useful? Rethinking Reasoning Utility in Universal Multimodal Embeddings", "item_type": "entry", "summary": "added: Is Reasoning Always Useful? Rethinking Reasoning Utility in Universal Multimodal Embeddings", "after": { "title": "Is Reasoning Always Useful? Rethinking Reasoning Utility in Universal Multimodal Embeddings", "link": "https://arxiv.org/abs/2609.29560", "id": "oai:arXiv.org:2609.29560v1", "author": "Wenxiao Fan, Jingling Fu, Luohang Liu, Xinyuan Shan, Lichen Ma, Yu He, Junshi Huang, Yan Li, Kan Li", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Reasoning-enhanced universal multimodal embeddings (UME) improve heterogeneous retrieval, but plausible rationales do not necessarily produce discriminative rankings. We study this gap by comparing the discriminative (DISC) and reasoning-driven generative (GEN) branches of UME-R1, a state-of-the-art reasoning UME method. We decompose reasoning utility into positive-target gain, hard-negative gain, and their margin difference. Positive similarity increases for 56.6%, but 15.7% are false-helpful cases where reasoning moves hard negatives closer even more. Local-neighborhood and token-attribution diagnostics suggest why: reasoning often de-condenses retrieved neighborhoods, but utility requires separator-aligned movement, while influential CoT tokens frequently encode evidence shared by positives and hard negatives. Motivated by these diagnostics, we propose SURE (Score-structure Utility Router for Embeddings), which improves UME-R1-7B by 1.5 points and yields consistent gains on two additional embedding models on MMEB-V2, without retraining, label-based policy selection, or extra VLM forward passes.", "categories": [ "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29560", "tags": [ "cs.AI" ] }, { "id": "3c0f9a41f4de21f9b05c", "title": "To Think or Not to Think: Allocating Reasoning Where It Helps", "content_text": "Reinforcement learning (RL) has proven effective in enhancing the reasoning performance of large language models (LLMs), particularly in complex mathematical and programming tasks. However, this capability comes with systematic \\textit{length misallocation}, in which models devote excessive reasoning to simple questions while terminating prematurely on harder ones, degrading inference efficiency with negligible accuracy improvement. Many length-adaptive methods mitigate this issue by allocating…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "3c0f9a41f4de21f9b05c", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29664v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "To Think or Not to Think: Allocating Reasoning Where It Helps", "item_type": "entry", "summary": "added: To Think or Not to Think: Allocating Reasoning Where It Helps", "after": { "title": "To Think or Not to Think: Allocating Reasoning Where It Helps", "link": "https://arxiv.org/abs/2609.29664", "id": "oai:arXiv.org:2609.29664v1", "author": "Zhengdong He, Yunfan Zhou, Jianguo Yao, Haibing Guan, Xijun Li", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Reinforcement learning (RL) has proven effective in enhancing the reasoning performance of large language models (LLMs), particularly in complex mathematical and programming tasks. However, this capability comes with systematic \\textit{length misallocation}, in which models devote excessive reasoning to simple questions while terminating prematurely on harder ones, degrading inference efficiency with negligible accuracy improvement. Many length-adaptive methods mitigate this issue by allocating token budgets according to question difficulty, under the implicit assumption that harder questions benefit monotonically from extended reasoning. In contrast, we find that the effect of reasoning length on accuracy is concentrated on \\textit{partially solvable} questions. Our further analysis reveals that explicit length rewards can produce unintended training dynamics. Motivated by these findings, we propose \\textbf{CARE}---\\textbf{C}ontrastive \\textbf{A}ccuracy \\textbf{R}eward \\textbf{E}stimation---which compares the beneficial length adjustment per question from online sampled responses and applies adaptive length rewards within Group Relative Policy Optimization, with no extra hyperparameters or additional inference cost. Experiments across multiple reasoning benchmarks demonstrate that our method improves Pass@1 by up to \\(4\\%\\) while simultaneously reducing reasoning length by \\(37\\%\\), achieving higher token efficiency. Code will be available upon the acceptance of this paper.", "categories": [ "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29664", "tags": [ "cs.AI" ] }, { "id": "0fe33a081b1fa79899ac", "title": "Fair Like Us? Auditing LLM Alignment in Resource Allocation", "content_text": "Fair allocation of scarce, indivisible resources is an important challenge in many societal problems. While there are several formal theories of fairness, no single definition can always be satisfied. As large language models (LLMs) are increasingly used to support decisions and act as agents, they raise new concerns about distributional justice: their judgments are not directly tied to any specific fairness framework and may violate key normative principles. In this work, we introduce a…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "0fe33a081b1fa79899ac", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29692v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Fair Like Us? Auditing LLM Alignment in Resource Allocation", "item_type": "entry", "summary": "added: Fair Like Us? Auditing LLM Alignment in Resource Allocation", "after": { "title": "Fair Like Us? Auditing LLM Alignment in Resource Allocation", "link": "https://arxiv.org/abs/2609.29692", "id": "oai:arXiv.org:2609.29692v1", "author": "Qishen Han, Hadi Hosseini, Joshua Kavner, Samarth Khanna, Sujoy Sikdar, Lirong Xia", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Fair allocation of scarce, indivisible resources is an important challenge in many societal problems. While there are several formal theories of fairness, no single definition can always be satisfied. As large language models (LLMs) are increasingly used to support decisions and act as agents, they raise new concerns about distributional justice: their judgments are not directly tied to any specific fairness framework and may violate key normative principles. In this work, we introduce a general method for evaluating fairness reasoning in LLMs. We study first-person fairness judgments across a broad set of models and compare them directly with human responses on matched scenarios and elicitation conditions. We find that LLMs tend to prefer stricter fairness constraints than humans, show more self-interested behavior, are sensitive to how information is framed, and are difficult to align with human judgments using fine-tuning with current datasets.", "categories": [ "cs.AI", "cs.CY", "cs.GT" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29692", "tags": [ "cs.AI", "cs.CY", "cs.GT" ] }, { "id": "009d3303d9f8f7566128", "title": "Breaking the Environment Wall: Evolving LLM Agent Environments for Recursive Self-Improvement", "content_text": "Many real-world tasks (e.g., office workflows, scientific experimentation) require LLM agents to interact repeatedly with their environments for context-dependent operations. However, such environments are often not agent-ready. First, information is often scattered and fragmented across the environment. Second, relevant evidence in the environment is often mixed with misleading information and conflicting versions. Third, environments evolve over time, introducing new noise and more…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "009d3303d9f8f7566128", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29773v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Breaking the Environment Wall: Evolving LLM Agent Environments for Recursive Self-Improvement", "item_type": "entry", "summary": "added: Breaking the Environment Wall: Evolving LLM Agent Environments for Recursive Self-Improvement", "after": { "title": "Breaking the Environment Wall: Evolving LLM Agent Environments for Recursive Self-Improvement", "link": "https://arxiv.org/abs/2609.29773", "id": "oai:arXiv.org:2609.29773v1", "author": "Yukai Wu, Yuanjing Yang, Le Zhou, Shaokun Han, Haoyu Wang, Zirui Tang, Weihuang Zheng, Maxm Pan, Xuanhe Zhou, Fan Wu", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Many real-world tasks (e.g., office workflows, scientific experimentation) require LLM agents to interact repeatedly with their environments for context-dependent operations. However, such environments are often not agent-ready. First, information is often scattered and fragmented across the environment. Second, relevant evidence in the environment is often mixed with misleading information and conflicting versions. Third, environments evolve over time, introducing new noise and more challenging tasks. These challenges can substantially degrade performance for state-of-the-art AI agents (e.g., from 83.9% to 57.6%). To address these challenges, we propose Env-Rethink (a system with 27B post-trained model) that supports three main capabilities: (1) It adaptively builds Collection Maps (for organizing related files) and Event Logs (for contextualizing cross-data relationships) to supplement necessary context; (2) It further leverages the post-trained model (through offline trajectory learning) to identify underlying noise issues in the environment; (3) It ultimately evolves environments through virtual event histories that alter environmental states and evidence relationships, producing more tricky ones for further agent improvement. Experiments show that Env-Rethink can effectively improve downstream task performance (with over 15.1% rubric pass rate improvement across nine models on 30 tasks).", "categories": [ "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29773", "tags": [ "cs.AI" ] }, { "id": "a9e7f58192419f61c08b", "title": "When Can Agents Forget Their Reasoning? ICLR for Long-Horizon Agent Context Compression", "content_text": "Long horizon language model agents continually accumulate reasoning history, increasing context length and inference cost even after earlier decisions have been executed and observed. Unlike static Chain of Thought compression, removing historical reasoning can change future actions and the resulting interaction trajectory. We study when such reasoning can be safely forgotten. We propose Interaction Aware Compression for Long Horizon Reasoning (ICLR), a training free online method that ranks…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "a9e7f58192419f61c08b", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29875v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "When Can Agents Forget Their Reasoning? ICLR for Long-Horizon Agent Context Compression", "item_type": "entry", "summary": "added: When Can Agents Forget Their Reasoning? ICLR for Long-Horizon Agent Context Compression", "after": { "title": "When Can Agents Forget Their Reasoning? ICLR for Long-Horizon Agent Context Compression", "link": "https://arxiv.org/abs/2609.29875", "id": "oai:arXiv.org:2609.29875v1", "author": "Mingxuan Wang, Fei Luo, Bo Wang, Guorun Yao, Yinglong Guo, Chao Ning, Hongyue Chen, Yanbiao Ma, Jungong Han", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Long horizon language model agents continually accumulate reasoning history, increasing context length and inference cost even after earlier decisions have been executed and observed. Unlike static Chain of Thought compression, removing historical reasoning can change future actions and the resulting interaction trajectory. We study when such reasoning can be safely forgotten. We propose Interaction Aware Compression for Long Horizon Reasoning (ICLR), a training free online method that ranks reasoning blocks using frozen proxy entropy while preserving actions, tool calls, and observations. On 260 WorkBuddyBench tasks, ICLR improves average reward from 0.699 to 0.718, while reducing input, output, and cache read tokens by 25.5%, 14.4%, and 33.3%, respectively. Ablations reveal trajectory amplification, where local reasoning deletion produces nonlinear changes in total computation by altering subsequent interaction. Representation probing, activation patching, and controlled trajectory analyses further suggest that historical reasoning becomes more replaceable once task relevant derived state has been reliably externalized into code, files, tool outputs, or environmental feedback. These results characterize agent reasoning as dynamic working state rather than permanent interaction history.", "categories": [ "cs.AI", "cs.CV" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29875", "tags": [ "cs.AI", "cs.CV" ] }, { "id": "b5962270ace3e59ed144", "title": "Qwen-Planner-Agent: A Closed-Loop AI-for-AI Framework for Real-World Mobile Planner Agents", "content_text": "The rapid progression of large language models is extending AI from passive content generation into the active workflows of engineering and scientific discovery. This shift raises a compelling question: can AI be both the object of development and an active participant in building next-generation AI systems? We explore this question by building Qwen-Planner-Agent within a closed-loop AI-for-AI framework for scalable development and iterative improvement. Mobile planning offers a demanding test…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "b5962270ace3e59ed144", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29892v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Qwen-Planner-Agent: A Closed-Loop AI-for-AI Framework for Real-World Mobile Planner Agents", "item_type": "entry", "summary": "added: Qwen-Planner-Agent: A Closed-Loop AI-for-AI Framework for Real-World Mobile Planner Agents", "after": { "title": "Qwen-Planner-Agent: A Closed-Loop AI-for-AI Framework for Real-World Mobile Planner Agents", "link": "https://arxiv.org/abs/2609.29892", "id": "oai:arXiv.org:2609.29892v1", "author": "Tingyu Qu, Weigao Sun, Yuecheng Liu, Yucheng Zhao, Yi Zhu, Yifeng Ding, Qiyi Wang, Sihan Cao, Pengkun Jiao, Hanlei Xie, Xiongwei Wu, Qichao Wang, Haodong Zhang, Jiajun Liu, Yuhao Wang, Yuqing Xie, Junpeng Zhao, Long Chen, Ming Ma, Sihan Yang, Ziwang Zhao, Yanhao Jia, Liangquan Gong, Feida Zhu, Yiran Zhong, Steven Hoi", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "The rapid progression of large language models is extending AI from passive content generation into the active workflows of engineering and scientific discovery. This shift raises a compelling question: can AI be both the object of development and an active participant in building next-generation AI systems? We explore this question by building Qwen-Planner-Agent within a closed-loop AI-for-AI framework for scalable development and iterative improvement. Mobile planning offers a demanding test of this approach: complex, long-horizon tasks challenge agent reliability, while costly real-device interaction limits development scalability. The framework connects data production, model training, and deployment through a shared action-feedback-verification contract. (i) AI for Data builds a human-gated agentic data flywheel in which specialized agents construct tasks, collect interaction trajectories, curate and balance training data, and use training feedback to guide subsequent data generation. (ii) AI for Training combines a supervised planning cold start with hybrid-environment online agentic reinforcement learning, where we introduce Competence-Aware Reward-and-Advantage Engineering (CARE) to reduce reasoning and tool-use costs while preserving task performance. (iii) AI drives model--harness co-evolution through an execution-evidence-driven loop that orchestrates memory, skills, and tools at runtime and feeds structured action feedback and preserved failure traces back into coordinated model and harness adaptation. Qwen-Planner-Agent achieves the best overall performance among all evaluated models and systems on MobilePA-Bench, improving over its base model across tool use, memory, skills, and sub-agent coordination. Further evaluations of our model show improvements across non-mobile agentic benchmarks while largely preserving general capabilities.", "categories": [ "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29892", "tags": [ "cs.AI" ] }, { "id": "9af509b58b5207ce160c", "title": "Who Holds the Pen? Let Specifications, Not Agents, Sign Off", "content_text": "Large language model agents increasingly combine generation, decision-making, execution, and self-evaluation within a single agentic loop. Although they operate under external specifications such as task instructions, guidelines, output schemas, and reusable skills, these specifications typically remain context for the same model that acts and declares completion, leaving no independent specification authority boundary. We identify two resulting gaps. The understanding--execution gap arises…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "9af509b58b5207ce160c", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29921v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Who Holds the Pen? Let Specifications, Not Agents, Sign Off", "item_type": "entry", "summary": "added: Who Holds the Pen? Let Specifications, Not Agents, Sign Off", "after": { "title": "Who Holds the Pen? Let Specifications, Not Agents, Sign Off", "link": "https://arxiv.org/abs/2609.29921", "id": "oai:arXiv.org:2609.29921v1", "author": "Haiqing Li, Xin Ma, Yinhao Wu, Wenliang Zhong, Feng Jiang, Thao M. Dang, Xiao Hu, Hehuan Ma, Yuzhi Guo, Junzhou Huang", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Large language model agents increasingly combine generation, decision-making, execution, and self-evaluation within a single agentic loop. Although they operate under external specifications such as task instructions, guidelines, output schemas, and reusable skills, these specifications typically remain context for the same model that acts and declares completion, leaving no independent specification authority boundary. We identify two resulting gaps. The understanding--execution gap arises when a requirement is understood but not satisfied in execution; the state--authority gap arises when an agent's interpretation or completion claim does not establish the required state. On SkillsBench, using only agent-visible prompts, workspace information, and injected skill specifications, we extract 509 source-grounded task directions. Across seven models, only 79.6%--86.4% are satisfied, while completion-claim rates exceed official evaluator pass rates by 28.7--37.9 percentage points. We therefore separate agent proposals from authoritative state. Agents may plan, act, and request completion, but only admissible evidence from qualified providers may establish specification-governed state. SpecHarness operationalizes this principle by compiling visible specifications into source-linked obligations and governing execution and finalization through versioned obligation state. Verifiable requirements are mediated or validated at runtime, while ambiguous or subjective requirements remain advisory. Experiments on guideline-following and artifact-generation tasks show that specifications can serve not merely as behavioral guidance, but as authority over compliant execution and completion.", "categories": [ "cs.AI", "cs.MA" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29921", "tags": [ "cs.AI", "cs.MA" ] }, { "id": "da345503c031e677ac46", "title": "How does Adversarial Influence Scale in Multi-Agent Systems?", "content_text": "Multi-agent deliberation can improve performance, but what happens when some agents do not act in good faith? In practice, an agent may be deceptive and work to subvert the group, whether through its own objectives or external instruction. We study how susceptibility to deception scales as groups increase in size and deceivers become more prevalent. It is not the number of agents in the group that matters, but the proportion of deceivers. We observe that the defection rate, how often initially…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "da345503c031e677ac46", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.30028v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "How does Adversarial Influence Scale in Multi-Agent Systems?", "item_type": "entry", "summary": "added: How does Adversarial Influence Scale in Multi-Agent Systems?", "after": { "title": "How does Adversarial Influence Scale in Multi-Agent Systems?", "link": "https://arxiv.org/abs/2609.30028", "id": "oai:arXiv.org:2609.30028v1", "author": "Addison J. Wu, Jasin Cekinmez, Michel Liao, Karthik Narasimhan, Thomas L. Griffiths", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Multi-agent deliberation can improve performance, but what happens when some agents do not act in good faith? In practice, an agent may be deceptive and work to subvert the group, whether through its own objectives or external instruction. We study how susceptibility to deception scales as groups increase in size and deceivers become more prevalent. It is not the number of agents in the group that matters, but the proportion of deceivers. We observe that the defection rate, how often initially correct agents switch to an incorrect final answer, rises linearly with this proportion. Whereas humans in comparable conformity studies are reliably swayed only when misleading confederates form a majority, LLM agents defect regularly even when deceivers remain a minority. Susceptibility also depends on which models are interacting, especially on the honest agent side. Unexpectedly, allowing deceivers to coordinate privately can make them less effective. Altogether, our results show that adding more agents is therefore not a sufficient defense, because the adversary can simply scale with the group.", "categories": [ "cs.AI", "cs.CY" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.30028", "tags": [ "cs.AI", "cs.CY" ] }, { "id": "1684f6bab9dba52f4205", "title": "Jev-Mobile: Jev as an Executor for Mobile GUI Agents", "content_text": "Vision-language models (VLMs) have become a common foundation for autonomous mobile GUI agents, but most existing systems rely on the VLM for both planning and action grounding at nearly every interaction step, leading to substantial latency and model-serving cost. We introduce Jev-Mobile, which shifts this paradigm to low-frequency VLM planning and high-frequency lightweight execution: the VLM specifies local goals, the accessibility tree defines a structured executable action space, and Jev…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "1684f6bab9dba52f4205", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.30186v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Jev-Mobile: Jev as an Executor for Mobile GUI Agents", "item_type": "entry", "summary": "added: Jev-Mobile: Jev as an Executor for Mobile GUI Agents", "after": { "title": "Jev-Mobile: Jev as an Executor for Mobile GUI Agents", "link": "https://arxiv.org/abs/2609.30186", "id": "oai:arXiv.org:2609.30186v1", "author": "Linghua Zhang", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Vision-language models (VLMs) have become a common foundation for autonomous mobile GUI agents, but most existing systems rely on the VLM for both planning and action grounding at nearly every interaction step, leading to substantial latency and model-serving cost. We introduce Jev-Mobile, which shifts this paradigm to low-frequency VLM planning and high-frequency lightweight execution: the VLM specifies local goals, the accessibility tree defines a structured executable action space, and Jev, a fast typed decision model, repeatedly selects actions within this space. This design allows multiple GUI actions to be executed under a single VLM decision, reducing expensive VLM inference while preserving adaptive interaction. On the full AndroidWorld task suite, Jev-Mobile achieves 79% task success, compared with 78% for SeeAct-V and 84% for a Step-wise VLM baseline. Among successful trajectories, it reduces mean end-to-end execution time by 32.7% and mean model API cost by 73.4% relative to Step-wise VLM. These results show that decoupling high-level VLM reasoning from low-level action execution can substantially improve mobile GUI agent efficiency while maintaining competitive task performance.", "categories": [ "cs.AI", "cs.SE" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.30186", "tags": [ "cs.AI", "cs.SE" ] }, { "id": "861f86934c79fb9e91d7", "title": "SAGE: Mitigating Long-Horizon Reasoning Biases via Topological Guidance", "content_text": "Long-horizon reasoning remains a central challenge for large language models (LLMs) under sparse-reward regimes. We argue that this brittleness arises from two biases induced by complex reasoning spaces: an exploration bias, where models are drawn toward locally plausible but structurally unstable branches, and a compounding bias, where small local deviations accumulate across depth and suppress rare rewards. We introduce Symbolic Closure Analysis (SCA) as a theoretical lens characterizing how…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "861f86934c79fb9e91d7", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.30192v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "SAGE: Mitigating Long-Horizon Reasoning Biases via Topological Guidance", "item_type": "entry", "summary": "added: SAGE: Mitigating Long-Horizon Reasoning Biases via Topological Guidance", "after": { "title": "SAGE: Mitigating Long-Horizon Reasoning Biases via Topological Guidance", "link": "https://arxiv.org/abs/2609.30192", "id": "oai:arXiv.org:2609.30192v1", "author": "Xinyue Zeng, Jiawei Zhang, Yujun Yan, Dawei Zhou", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Long-horizon reasoning remains a central challenge for large language models (LLMs) under sparse-reward regimes. We argue that this brittleness arises from two biases induced by complex reasoning spaces: an exploration bias, where models are drawn toward locally plausible but structurally unstable branches, and a compounding bias, where small local deviations accumulate across depth and suppress rare rewards. We introduce Symbolic Closure Analysis (SCA) as a theoretical lens characterizing how branching structures and sparse rewards induce these biases in long-horizon reasoning with local admissibility, and as a design principle for structural priors in less formal reasoning tasks. Motivated by this analysis, we propose SAGE (Structural Admissibility-Guided Exploration), a unified framework that injects structural guidance to alleviate exploration bias and compounding bias in long-horizon reasoning. SAGE combines two complementary structural guidance: algebraic sparsification, which projects locally admissible candidates onto operator-indexed algebraic subspaces to suppress spurious branching and mitigate exploration bias, and hyperbolic structural guidance, which embeds reasoning states into a negatively curved space to provide dense depth-wise signals and mitigate compounding bias. Across 12 benchmarks and 7 model families, SAGE outperforms competitive baselines. In particular, SAGE achieves up to an 8-fold improvement on the Andrews-Curtis problem, an open real-world long-horizon task. Code is available at: https://github.com/Susan571/SAGE-NeurIPS2026.", "categories": [ "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.30192", "tags": [ "cs.AI" ] }, { "id": "86edb88fc1d058e0d0d3", "title": "Who Is Behind the Harness? Fingerprinting LLMs through Agentic Behavior", "content_text": "LLMs increasingly operate through coding-agent harnesses that inspect repositories, invoke tools, and modify files. Substituting the model behind such an agent can therefore change security-relevant decisions, including whether it verifies changes or recovers safely from failures. Existing LLM fingerprints largely infer identity from direct text or token distributions. In coding agents, these signals are mediated by system instructions, controller logic, tools, and execution feedback, limiting…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "86edb88fc1d058e0d0d3", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.28559v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Who Is Behind the Harness? Fingerprinting LLMs through Agentic Behavior", "item_type": "entry", "summary": "added: Who Is Behind the Harness? Fingerprinting LLMs through Agentic Behavior", "after": { "title": "Who Is Behind the Harness? Fingerprinting LLMs through Agentic Behavior", "link": "https://arxiv.org/abs/2609.28559", "id": "oai:arXiv.org:2609.28559v1", "author": "Chuyi Wang, Xiaohui Xie, Tongze Wang, Fangchen Luo, Yong Cui", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "LLMs increasingly operate through coding-agent harnesses that inspect repositories, invoke tools, and modify files. Substituting the model behind such an agent can therefore change security-relevant decisions, including whether it verifies changes or recovers safely from failures. Existing LLM fingerprints largely infer identity from direct text or token distributions. In coding agents, these signals are mediated by system instructions, controller logic, tools, and execution feedback, limiting their transfer.\nWe present LIDAR (LLM Identification from Decisions and Actions at Runtime), an active black-box fingerprinting method for coding-agent execution. Three coding probe pairs expose post-edit verification, transient-failure recovery, and specification--test conflict resolution under controlled changes. LIDAR represents the resulting trajectories with complementary instance-level and distribution-level features and compares them with clean references using a lightweight probabilistic identifier. It requires no access to model weights, logits, or provider internals.\nAcross 36 models from seven families and two agent harnesses, LIDAR achieves high Top-1 accuracy and MRR and outperforms four existing fingerprinting and API-auditing baselines. Ablations confirm that the two feature levels, all probe pairs, and their controlled variants contribute. These results show that agent execution behavior provides model-identity evidence beyond final outputs.", "categories": [ "cs.CR", "cs.AI", "cs.SE" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.28559", "tags": [ "cs.CR", "cs.AI", "cs.SE" ] }, { "id": "9da69632c76e9c1592d3", "title": "Speculative Evaluation of Stochastic LLMs", "content_text": "Evaluating a stochastic large language model is costly: benchmark scores estimate expected performance from randomized rollouts, yet uniform repetition ignores sharp differences in task-level rollout variance. We ask how to minimize the variance of a fixed-benchmark mean under an exact rollout budget. We develop Speculative Evaluation with a Hierarchical Bayesian Neyman (HBN) policy with pilot size and stage weight jointly chosen ex ante. It runs a short uniform pilot, pools per-task success…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "9da69632c76e9c1592d3", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.28560v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Speculative Evaluation of Stochastic LLMs", "item_type": "entry", "summary": "added: Speculative Evaluation of Stochastic LLMs", "after": { "title": "Speculative Evaluation of Stochastic LLMs", "link": "https://arxiv.org/abs/2609.28560", "id": "oai:arXiv.org:2609.28560v1", "author": "Qianli Shen, Xiang Li, Ruomeng Ding, Yanxi Chen, Daoyuan Chen, Yaliang Li", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Evaluating a stochastic large language model is costly: benchmark scores estimate expected performance from randomized rollouts, yet uniform repetition ignores sharp differences in task-level rollout variance. We ask how to minimize the variance of a fixed-benchmark mean under an exact rollout budget. We develop Speculative Evaluation with a Hierarchical Bayesian Neyman (HBN) policy with pilot size and stage weight jointly chosen ex ante. It runs a short uniform pilot, pools per-task success counts with a hierarchical Bayesian model, and uses posterior expectations of task-level sampling variances for exact positive-integer Neyman allocation. To mitigate the pilot synchronization barrier, HBN-async speculatively executes continuations from partial pilot feedback and retains those selected by the final allocation. Across six checkpoints and 18 benchmark groups, we evaluate 107 nondegenerate benchmark-checkpoint profiles. For rollout budgets of 8-64 per task, Speculative Evaluation reduces variance relative to Uniform by 12.8%-33.6% on average across profiles, outperforming hindsight-tuned empirical and independent Bayesian baselines. Real-generation experiments that account for the pilot synchronization barrier show that HBN-async mitigates its overhead, helping translate statistical efficiency into practical evaluation benefits.", "categories": [ "stat.ML", "cs.AI", "cs.LG" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.28560", "tags": [ "stat.ML", "cs.AI", "cs.LG" ] }, { "id": "95e137c3291575d3c581", "title": "Where Cyber Agents Struggle: Bottleneck Analysis of Multi-Stage LLM Agents", "content_text": "Multi-stage LLM-based cyber agents may complete attack workflows while remaining brittle, costly, or reliant on incorrect interpretations of execution evidence. Success rates alone obscure inefficiency, adaptation through retries, and recognition of success or failure. We present an end-to-end diagnostic study of an Autonomous Adversary system with orchestrator, executor, and validator LLMs in enterprise-like lateral-movement scenarios. Six frontier models are evaluated across two scenarios and…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "95e137c3291575d3c581", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.28572v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Where Cyber Agents Struggle: Bottleneck Analysis of Multi-Stage LLM Agents", "item_type": "entry", "summary": "added: Where Cyber Agents Struggle: Bottleneck Analysis of Multi-Stage LLM Agents", "after": { "title": "Where Cyber Agents Struggle: Bottleneck Analysis of Multi-Stage LLM Agents", "link": "https://arxiv.org/abs/2609.28572", "id": "oai:arXiv.org:2609.28572v1", "author": "Saeedeh Lohrasbi, Mohammad Mamun, Ahmed Yehia, Scott Buffett, Sherif Saad", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Multi-stage LLM-based cyber agents may complete attack workflows while remaining brittle, costly, or reliant on incorrect interpretations of execution evidence. Success rates alone obscure inefficiency, adaptation through retries, and recognition of success or failure. We present an end-to-end diagnostic study of an Autonomous Adversary system with orchestrator, executor, and validator LLMs in enterprise-like lateral-movement scenarios. Six frontier models are evaluated across two scenarios and three modes: expert-defined, self-scaffolded, and fully autonomous. We assess validator consistency and evidence grounding; introduce a subtask-conditioned, cost-aware score for abnormal token use, retries, and runtime; and use comparative LLM-as-a-Judge analysis to identify planning deficiencies, including tool misalignment, plan similarity, over-specification, inadequate probing, and weak recovery. Validators are generally relevant and evidence-grounded but often nonspecific and overly optimistic. Bottlenecks cluster in credential and lateral-movement tasks, spread with scenario complexity, and vary more under full autonomy. Reliable evaluation must assess outcomes, evidence interpretation, resource use, and adaptation after failure.", "categories": [ "cs.CR", "cs.AI", "cs.SE" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.28572", "tags": [ "cs.CR", "cs.AI", "cs.SE" ] }, { "id": "198ce2ba44a9fae22c19", "title": "Persistent Billable State: Denial-of-Wallet Attacks and Defenses in Tool-Calling LLM Agents", "content_text": "Multi-step tool-calling LLM agents rely on host runtimes to preserve state across turns. When a runtime carries an external tool return into later model inputs, providers meter it again. An admitted malicious or compromised tool can thereby convert untrusted data into recurring victim-billed processing without victim credentials or local runtime privilege. We call retained content persistent billable state and formalize the host's decision over whether and how it enters later billable context…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "198ce2ba44a9fae22c19", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.28585v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Persistent Billable State: Denial-of-Wallet Attacks and Defenses in Tool-Calling LLM Agents", "item_type": "entry", "summary": "added: Persistent Billable State: Denial-of-Wallet Attacks and Defenses in Tool-Calling LLM Agents", "after": { "title": "Persistent Billable State: Denial-of-Wallet Attacks and Defenses in Tool-Calling LLM Agents", "link": "https://arxiv.org/abs/2609.28585", "id": "oai:arXiv.org:2609.28585v1", "author": "Jinqian Zhang (Institute of Information Engineering, Chinese Academy of Sciences, School of Cyber Security, University of Chinese Academy of Sciences), Haojun Xia (Institute of Information Engineering, Chinese Academy of Sciences, School of Cyber Security, University of Chinese Academy of Sciences), Shujiang Wu (Beihang University), Jingkun Yue (State Key Laboratory of Networking and Switching Technology, Beijing University of Posts and Telecommunications, Beijing, China), Xia Zhang (Institute of Information Engineering, Chinese Academy of Sciences, School of Cyber Security, University of Chinese Academy of Sciences), Zhangpei Cheng (Institute of Information Engineering, Chinese Academy of Sciences, School of Cyber Security, University of Chinese Academy of Sciences), Bibo Tu (Institute of Information Engineering, Chinese Academy of Sciences, School of Cyber Security, University of Chinese Academy of Sciences)", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Multi-step tool-calling LLM agents rely on host runtimes to preserve state across turns. When a runtime carries an external tool return into later model inputs, providers meter it again. An admitted malicious or compromised tool can thereby convert untrusted data into recurring victim-billed processing without victim credentials or local runtime privilege. We call retained content persistent billable state and formalize the host's decision over whether and how it enters later billable context as the persistent billable-state boundary.\nWe present the first systematic security study of this post-admission lifecycle. We derive six denial-of-wallet attack vectors and build DOW-BENCH, an end-to-end harness evaluated across six model families. Across 243 executions, usage telemetry shows that the maximum per-session cumulative input reaches 14,293x the session's first-call input. Controlled history-policy reruns isolate raw retention's contribution: retaining raw history increases mean effective session cost by 21.2-35.9%. Compression succeeds on 10/12 and 11/12 history-dependent tasks, versus 2/12 under deletion for each provider.\nTo govern this boundary, we combine deterministic history transformation with four host-side invariants that bound prompt mass, context growth, recursive opportunity, and cumulative spend before reingestion. The kernel contains every recurring attack in the 123-evaluation replay corpus. Across 24 Mistral Small 4 workflows, a progress-authorized policy achieves 22/24 oracle-verified task successes with no pre-completion interruptions, versus 13/24 under a fixed cap. Only 71 of 3,830 scanned MCP server and transport repositories expose any code-visible safeguard proxy, and none cover all four safeguard families. These results establish persistent billable state as a first-class security object and pre-reingestion as its host-owned control point.", "categories": [ "cs.CR", "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.28585", "tags": [ "cs.CR", "cs.AI" ] }, { "id": "36ef7dc2188e4111e30e", "title": "NumericJev: Jev-like LLM Numerical Decoding with Multiway Decision Trees", "content_text": "Large language models can interpret natural lan- guage, yet robust decisions remain challenging. Jev-like models expose structured choices, but these interfaces do not directly provide numeri- cal values at a requested precision. We propose NUMERICJEV, a training-free numerical decod- ing algorithm that enables numerical output from any LLM with a Jev-like structured-choice in- terface. Surprisingly, on our arithmetic bench- mark, it outperforms direct selection from a can- didate list…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "36ef7dc2188e4111e30e", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.28587v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "NumericJev: Jev-like LLM Numerical Decoding with Multiway Decision Trees", "item_type": "entry", "summary": "added: NumericJev: Jev-like LLM Numerical Decoding with Multiway Decision Trees", "after": { "title": "NumericJev: Jev-like LLM Numerical Decoding with Multiway Decision Trees", "link": "https://arxiv.org/abs/2609.28587", "id": "oai:arXiv.org:2609.28587v1", "author": "Weiwei Ye, Hangchen Liu, Renhe Jiang", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Large language models can interpret natural lan- guage, yet robust decisions remain challenging. Jev-like models expose structured choices, but these interfaces do not directly provide numeri- cal values at a requested precision. We propose NUMERICJEV, a training-free numerical decod- ing algorithm that enables numerical output from any LLM with a Jev-like structured-choice in- terface. Surprisingly, on our arithmetic bench- mark, it outperforms direct selection from a can- didate list containing the correct answer by 2.93 percentage points (Figure 1). Our motivation comes from the observation that numerical range selection is itself a decision problem that Jev- like LLMs can address. NUMERICJEV recur- sively refines a range through a multiway deci- sion tree while retaining the original question in context, without parameter updates or hidden- state access. On a 100-value grid, a ten-way tree requires only two decision rounds. Range- normalized MAE is 1.84% versus 5.18% for di- rect choice. A separate three-date historical- index study yields 4.58% mean relative recall er- ror and 0% readout error when the value is sup- plied. Code is available at https://github. com/Bring-AI/jev-numeric.", "categories": [ "stat.ML", "cs.AI", "cs.LG" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.28587", "tags": [ "stat.ML", "cs.AI", "cs.LG" ] }, { "id": "e33154692a03bf9ef0be", "title": "DrGait: Biomechanically Grounded Visual Reasoning for Interpretable Clinical Gait Analysis", "content_text": "Current automated gait analysis for clinical applications relies on uninterpretable black-box classifiers. Although Vision-Language Models (VLMs) offer strong reasoning capabilities, applying them directly to gait videos often leads to hallucinations, because they struggle to measure subtle geometric deviations from raw visual contexts. To address this, we introduce DrGait, a training-free agentic framework that shifts the VLM's role from a direct visual reasoner to a clinical planner. DrGait…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "e33154692a03bf9ef0be", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.28796v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "DrGait: Biomechanically Grounded Visual Reasoning for Interpretable Clinical Gait Analysis", "item_type": "entry", "summary": "added: DrGait: Biomechanically Grounded Visual Reasoning for Interpretable Clinical Gait Analysis", "after": { "title": "DrGait: Biomechanically Grounded Visual Reasoning for Interpretable Clinical Gait Analysis", "link": "https://arxiv.org/abs/2609.28796", "id": "oai:arXiv.org:2609.28796v1", "author": "Xiangyu Yin, Shiqi Wang, Abrar Alamri, Yasir Aljohani, Weichen Liu, Goeran Fiedler, Wei Gao", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Current automated gait analysis for clinical applications relies on uninterpretable black-box classifiers. Although Vision-Language Models (VLMs) offer strong reasoning capabilities, applying them directly to gait videos often leads to hallucinations, because they struggle to measure subtle geometric deviations from raw visual contexts. To address this, we introduce DrGait, a training-free agentic framework that shifts the VLM's role from a direct visual reasoner to a clinical planner. DrGait decouples semantic reasoning from geometric perception through a structured Triage-Verification-Synthesis (TVS) workflow. Given an input video and a set of basic spatiotemporal metrics, the DrGait agent first performs a heuristic triage to propose diagnostic hypotheses, which are then verified by autonomously calling deterministic biomechanical tools that operate on reconstructed 3D mesh trajectories, segmented 2D pose tracks, and event-centered video evidence. Finally, a closed-loop mechanism recursively updates the agent's reasoning context based on the feedback. By anchoring VLM's reasoning in verifiable geometric and temporal measurements, DrGait reduces hallucinations, achieving competitive diagnostic accuracy while generating transparent and audit-ready clinical reports.", "categories": [ "cs.CV", "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.28796", "tags": [ "cs.CV", "cs.AI" ] }, { "id": "1c68d64c2e782dd3d439", "title": "Blockchain-Enabled Artificial Intelligence and AI Agents for Secure Data Sharing and Cybersecurity Applications", "content_text": "Blockchain and artificial intelligence (AI) are converging into a single infrastructural layer for securing data sharing, model integrity, and autonomous decision-making across distributed systems. This paper presents a meta-synthesis that draws together four constituent studies covering adversarial machine learning, AI-powered anomaly detection in cloud environments, automated vulnerability patching by multi-agent large language model (LLM) pipelines, and the broader landscape of securing AI…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "1c68d64c2e782dd3d439", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.28843v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Blockchain-Enabled Artificial Intelligence and AI Agents for Secure Data Sharing and Cybersecurity Applications", "item_type": "entry", "summary": "added: Blockchain-Enabled Artificial Intelligence and AI Agents for Secure Data Sharing and Cybersecurity Applications", "after": { "title": "Blockchain-Enabled Artificial Intelligence and AI Agents for Secure Data Sharing and Cybersecurity Applications", "link": "https://arxiv.org/abs/2609.28843", "id": "oai:arXiv.org:2609.28843v1", "author": "Harsh Verma", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Blockchain and artificial intelligence (AI) are converging into a single infrastructural layer for securing data sharing, model integrity, and autonomous decision-making across distributed systems. This paper presents a meta-synthesis that draws together four constituent studies covering adversarial machine learning, AI-powered anomaly detection in cloud environments, automated vulnerability patching by multi-agent large language model (LLM) pipelines, and the broader landscape of securing AI systems across their lifecycle and situates their findings within the emerging literature on blockchain-enabled AI and autonomous AI agents. Each constituent study addresses a distinct point of failure in modern AI-driven security operations: the integrity of training data and model behavior, the reliability of real-time monitoring, and the trustworthiness of automated code remediation. We argue that blockchain's properties of immutability, decentralized consensus, and verifiable provenance directly address a gap common to all three: the difficulty of establishing trust in data, models, and autonomous agents that operate without a central authority. Building on real-world research on blockchain-secured data sharing, federated learning, and multi-agent coordination, we propose a layered reference architecture that couples adversarially hardened models, blockchain-anchored data provenance, AI-driven anomaly detection, and smart-contract-governed multi-agent remediation. We conclude by identifying open problems in scalability, privacy-transparency trade-offs, and the governance of autonomous agents that must be resolved before such integrated systems can be trusted in production-critical environments.", "categories": [ "cs.CR", "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.28843", "tags": [ "cs.CR", "cs.AI" ] }, { "id": "f7a514d52a329868547c", "title": "On the Effectiveness of Kernel-Level Evidence for Agent Security", "content_text": "LLM agents are deployed into infrastructure that grants them broad host authority, yet existing agent-security benchmarks and defenses operate almost exclusively at the application telemetry layer: the served tool manifest, the user prompt, and the model's messages. Some threats, however, smuggle malicious instructions and actions past the application boundary, leaving them invisible to that layer. In this work, we bridge that gap by pairing application-level agent telemetry with kernel-level…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "f7a514d52a329868547c", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.28915v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "On the Effectiveness of Kernel-Level Evidence for Agent Security", "item_type": "entry", "summary": "added: On the Effectiveness of Kernel-Level Evidence for Agent Security", "after": { "title": "On the Effectiveness of Kernel-Level Evidence for Agent Security", "link": "https://arxiv.org/abs/2609.28915", "id": "oai:arXiv.org:2609.28915v1", "author": "Spencer King, Zhilu Zhang, Mikhail Kuznetsov, Kay Liu, Baris Coskun, Wei Ding", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "LLM agents are deployed into infrastructure that grants them broad host authority, yet existing agent-security benchmarks and defenses operate almost exclusively at the application telemetry layer: the served tool manifest, the user prompt, and the model's messages. Some threats, however, smuggle malicious instructions and actions past the application boundary, leaving them invisible to that layer. In this work, we bridge that gap by pairing application-level agent telemetry with kernel-level syscall traces to present the first paired-evidence characterization of kernel-level versus application-layer signal for agent security. To quantify the value of the enhanced telemetry, we introduce Agent Cross-Layer Evidence (ACE), a paired-session corpus of 4,047 sessions and 17 threat models spanning six delivery-vector families and 14 of the 25 OWASP LLM and agentic threat categories, organized into 12 attack mechanics with per-mechanic characterization of where the most discriminative evidence lies. Across four distinct detector families, we find that kernel evidence is discriminative on its own and that composing it with application-layer evidence generally outperforms either single-layer view, revealing complementary signals that single-layer analyses can miss. We further demonstrate generalization to unseen attack families and transfer to an alternate agent runtime. Together, these findings establish the value of cross-layer evidence for agent security.", "categories": [ "cs.CR", "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.28915", "tags": [ "cs.CR", "cs.AI" ] }, { "id": "9df1fcde4348fd981470", "title": "Calibrated Decision Models for Autonomous Penetration-Testing Harnesses: JEV and Laya as System One Decision Layers for LLM-Driven Pentest Agents", "content_text": "Autonomous penetration-testing harnesses use large language models (LLMs) for reconnaissance, exploitation, and reporting, but often rely on those same models to confirm findings, grade severity, and select agents. This can lead to false positives, inflated severity, and wasted compute. We examine how System One decision models, lightweight non-generative classifiers that return typed, calibrated verdicts, can support these decisions. We make five contributions. First, we define four decision…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "9df1fcde4348fd981470", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.28940v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Calibrated Decision Models for Autonomous Penetration-Testing Harnesses: JEV and Laya as System One Decision Layers for LLM-Driven Pentest Agents", "item_type": "entry", "summary": "added: Calibrated Decision Models for Autonomous Penetration-Testing Harnesses: JEV and Laya as System One Decision Layers for LLM-Driven Pentest Agents", "after": { "title": "Calibrated Decision Models for Autonomous Penetration-Testing Harnesses: JEV and Laya as System One Decision Layers for LLM-Driven Pentest Agents", "link": "https://arxiv.org/abs/2609.28940", "id": "oai:arXiv.org:2609.28940v1", "author": "Joas Antonio dos Santos Barbosa", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Autonomous penetration-testing harnesses use large language models (LLMs) for reconnaissance, exploitation, and reporting, but often rely on those same models to confirm findings, grade severity, and select agents. This can lead to false positives, inflated severity, and wasted compute. We examine how System One decision models, lightweight non-generative classifiers that return typed, calibrated verdicts, can support these decisions. We make five contributions. First, we define four decision points: finding adjudication, severity recalibration, agent pruning, and confirmation loops. Second, we present an exploratory NeuroSploit case study comparing one run with TypeSafe System One (Jev) and one without it against a web target containing 13 vulnerabilities. Differences in severity distribution, runtime, and grading by exposed data type motivate the architecture but do not establish statistical significance. Third, we review published specifications for Jev, Jev-Ultrafast, and the open-source Laya without assuming that results from other benchmarks transfer to penetration testing. Fourth, we discuss RLHF, RLAIF, RLCD, and RLHV as training approaches and their implications for trust in security decisions. Finally, we propose Rave, a domain-adapted System One model, and outline its training data, evaluation protocol, and potential effect on harness assurance.", "categories": [ "cs.CR", "cs.AI", "cs.SE" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.28940", "tags": [ "cs.CR", "cs.AI", "cs.SE" ] }, { "id": "dd0470ea40061f8fe9b9", "title": "Multi-Agent Orchestration of 3GPP Channel Estimators", "content_text": "Pilot-aided channel estimation is a decisive block in orthogonal frequency-division multiplexing (OFDM) receivers for both 5G New Radio (5G-NR) and Long-Term Evolution (LTE). A large body of estimators exists, from simple least-squares (LS) interpolation to statistically optimal linear minimum-mean-square-error (LMMSE) variants and, more recently, deep convolutional denoisers, yet no single estimator is uniformly best: the winner depends on the propagation scenario, the numerology, the…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "dd0470ea40061f8fe9b9", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29044v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Multi-Agent Orchestration of 3GPP Channel Estimators", "item_type": "entry", "summary": "added: Multi-Agent Orchestration of 3GPP Channel Estimators", "after": { "title": "Multi-Agent Orchestration of 3GPP Channel Estimators", "link": "https://arxiv.org/abs/2609.29044", "id": "oai:arXiv.org:2609.29044v1", "author": "I. Zakir Ahmed, Hamid Sadjadpour", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Pilot-aided channel estimation is a decisive block in orthogonal frequency-division multiplexing (OFDM) receivers for both 5G New Radio (5G-NR) and Long-Term Evolution (LTE). A large body of estimators exists, from simple least-squares (LS) interpolation to statistically optimal linear minimum-mean-square-error (LMMSE) variants and, more recently, deep convolutional denoisers, yet no single estimator is uniformly best: the winner depends on the propagation scenario, the numerology, the operating signal-to-noise ratio (SNR), the mobility (Doppler), and the antenna configuration. In this paper, we quantify this fact through a unified study of eight literature estimators evaluated over the 3GPP TR~38.901 Urban-Macro (UMa), Urban-Micro (UMi), and Rural-Macro (RMa) channels generated with NVIDIA Sionna, for both 5G-NR and LTE numerologies, in single-input single-output (SISO) and $8\\times2$ multiple-input multiple-output (MIMO) settings. We then propose a \\emph{condition-adaptive multi-agent orchestrator} that treats each estimator as an independent agent and dispatches, per operating condition, to the agent that is best on a validation split without any genie knowledge. The orchestrator tracks the per-realization oracle to within $1.07$~dB and improves the normalized mean-square error (NMSE) over the best \\emph{fixed} strategy by up to $3.6$~dB at high SNR, where the low-SNR champion is no longer optimal. Because the agents are independent, running them concurrently delivers this best-of-eight accuracy at essentially single-estimator latency: a data-parallel partition scales the wall-clock nearly as $1/K$ with $K$ workers (up to $6.9\\times$), whereas naive by-algorithm partitioning is Amdahl-limited by the heaviest agent. The results substantiate multi-agent orchestration as a practical route to robust channel estimation across heterogeneous 5G-NR/LTE deployments.", "categories": [ "cs.IT", "cs.AI", "eess.SP", "math.IT" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29044", "tags": [ "cs.IT", "cs.AI", "eess.SP", "math.IT" ] }, { "id": "c0925382453b923a0812", "title": "Where Does Exactly-Once Live? Model, Harness, and Tool-Contract Effects on Duplicate Side Effects in LLM Agents", "content_text": "When a tool-using agent's write times out or returns a server error, the action may already have taken effect. Retrying blindly duplicates it -- a second charge, a second announcement, a second deployment -- while giving up skips required work. We ask where exactly-once behaviour should be enforced: in the model, in the agent harness, or in the tool contract. We introduce LIMBO, a deterministic sandbox of six services with realistic contracts (optional idempotency keys, eventually consistent…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "c0925382453b923a0812", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29095v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Where Does Exactly-Once Live? Model, Harness, and Tool-Contract Effects on Duplicate Side Effects in LLM Agents", "item_type": "entry", "summary": "added: Where Does Exactly-Once Live? Model, Harness, and Tool-Contract Effects on Duplicate Side Effects in LLM Agents", "after": { "title": "Where Does Exactly-Once Live? Model, Harness, and Tool-Contract Effects on Duplicate Side Effects in LLM Agents", "link": "https://arxiv.org/abs/2609.29095", "id": "oai:arXiv.org:2609.29095v1", "author": "Jiapeng Li", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "When a tool-using agent's write times out or returns a server error, the action may already have taken effect. Retrying blindly duplicates it -- a second charge, a second announcement, a second deployment -- while giving up skips required work. We ask where exactly-once behaviour should be enforced: in the model, in the agent harness, or in the tool contract. We introduce LIMBO, a deterministic sandbox of six services with realistic contracts (optional idempotency keys, eventually consistent and missing read paths) and twelve fault modes injected at the service boundary, including late commits, redelivery and partial batches; every episode is graded against a ledger of committed effects. Across 25,930 episodes spanning nine recent models, three production agent harnesses, two contract variants and fifteen recovery conditions, the answer depends on the fault. When an immediate read-back can reveal what happened, the model decides: frontier models instructed to act exactly once almost never duplicate a write whose acknowledgement was lost (0.5%), weaker models often do, and the model explains 53% of the explained variance. When it cannot -- the request is still in flight, or the transport delivered it twice -- the same frontier models duplicate in 56% and 74% of episodes, and the contract explains 81%. We prove that no verification-only policy is exactly-once under late commits without a bound on in-flight time. Waiting works when such a bound is short and known, but with heavy-tailed in-flight delays even an hour of waiting per episode falls short of offering an idempotency key on every write, which lowers the duplicate rate from 28% to 4% because agents use keys when they exist. The harness barely matters, a guard that attaches keys transfers across harnesses unchanged, and agents reported success in 90% of the episodes in which they had duplicated an effect.", "categories": [ "cs.LG", "cs.AI", "cs.SE" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29095", "tags": [ "cs.LG", "cs.AI", "cs.SE" ] }, { "id": "88e13dd75c1e80c2ec64", "title": "DocuTeam: Mixed-Initiative Multi-Agent Discussions around Evolving Documents", "content_text": "In open-ended problem solving, collaborators often rely on discussion to surface concerns, challenge perspectives, and refine shared work as it evolves. While AI agents are increasingly used as discussion partners, existing multi-agent systems place a heavy burden on users to initiate and carefully orchestrate the discussions. We present DocuTeam, a mixed-initiative multi-agent discussion system in which both users and agents can initiate and steer conversations. Agents monitor document changes…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "88e13dd75c1e80c2ec64", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29309v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "DocuTeam: Mixed-Initiative Multi-Agent Discussions around Evolving Documents", "item_type": "entry", "summary": "added: DocuTeam: Mixed-Initiative Multi-Agent Discussions around Evolving Documents", "after": { "title": "DocuTeam: Mixed-Initiative Multi-Agent Discussions around Evolving Documents", "link": "https://arxiv.org/abs/2609.29309", "id": "oai:arXiv.org:2609.29309v1", "author": "Heechan Lee, Juhyeon Choi, Tae Soo Kim, Juho Kim, Joseph Seering", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "In open-ended problem solving, collaborators often rely on discussion to surface concerns, challenge perspectives, and refine shared work as it evolves. While AI agents are increasingly used as discussion partners, existing multi-agent systems place a heavy burden on users to initiate and carefully orchestrate the discussions. We present DocuTeam, a mixed-initiative multi-agent discussion system in which both users and agents can initiate and steer conversations. Agents monitor document changes to proactively start and redirect discussions as the work evolves, while users can flexibly shape the conversation or adopt agent ideas. In a within-subjects study (N=20), participants using DocuTeam produced outcomes rated significantly more novel, relevant, and specific than with a baseline without any increase in cognitive load. Rather than using agents for one-off idea sourcing, participants engaged in an iterative refinement loop in which document changes prompted agent reactions, which led users to revisit and further develop their work.", "categories": [ "cs.HC", "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29309", "tags": [ "cs.HC", "cs.AI" ] }, { "id": "8e36f677e1c1137805b5", "title": "Domain Recentering and Confidence-Weighted Prior Calibration for Vision-Language Models", "content_text": "Vision-language models such as CLIP achieve strong zero-shot classification, yet under distribution shift, visual embeddings drift from fixed text embeddings. Training-free calibration avoids the per-sample optimization of prompt learning, but prior feature calibration gives each image the full bias of one hard cluster. We propose Domain Recentering with Confidence Calibration (DRC), a training-free method adapting CLIP from a set of unlabeled target images. DRC fits a Gaussian mixture once and…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "8e36f677e1c1137805b5", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29358v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Domain Recentering and Confidence-Weighted Prior Calibration for Vision-Language Models", "item_type": "entry", "summary": "added: Domain Recentering and Confidence-Weighted Prior Calibration for Vision-Language Models", "after": { "title": "Domain Recentering and Confidence-Weighted Prior Calibration for Vision-Language Models", "link": "https://arxiv.org/abs/2609.29358", "id": "oai:arXiv.org:2609.29358v1", "author": "Youngeun Seol, Jimin Shin, Heeseo Yoon, Uiwon Hwang", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Vision-language models such as CLIP achieve strong zero-shot classification, yet under distribution shift, visual embeddings drift from fixed text embeddings. Training-free calibration avoids the per-sample optimization of prompt learning, but prior feature calibration gives each image the full bias of one hard cluster. We propose Domain Recentering with Confidence Calibration (DRC), a training-free method adapting CLIP from a set of unlabeled target images. DRC fits a Gaussian mixture once and subtracts from each embedding a posterior-weighted average of component means. It then removes residual class preference with a log-prior correction, estimating the prior from confidence-weighted predictions. Among compared methods, DRC achieves the highest average accuracy on cross-domain datasets, exceeding zero-shot CLIP by 4.13 and 5.07 points with ViT-B/16 and ResNet-50, with gains over CLIP also holding under ImageNet distribution shifts.", "categories": [ "cs.CV", "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29358", "tags": [ "cs.CV", "cs.AI" ] }, { "id": "b65e7074fa2e41be66ea", "title": "GeoRefer-Bench: A Benchmark from Referring Pixels to Verifiable Geospatial Reasoning", "content_text": "Referring segmentation in overhead imagery is inherently relational: a query may ask for the buildings north of the road or the pond closest to a residential area, so the correct referent can contain one object, several objects, or none. Existing benchmarks mainly score mask overlap, which cannot verify whether a model actually resolved the stated spatial relation. We introduce GeoRefer-Bench, a benchmark for verifiable geospatial referring segmentation. Each query is represented by an…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "b65e7074fa2e41be66ea", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29541v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "GeoRefer-Bench: A Benchmark from Referring Pixels to Verifiable Geospatial Reasoning", "item_type": "entry", "summary": "added: GeoRefer-Bench: A Benchmark from Referring Pixels to Verifiable Geospatial Reasoning", "after": { "title": "GeoRefer-Bench: A Benchmark from Referring Pixels to Verifiable Geospatial Reasoning", "link": "https://arxiv.org/abs/2609.29541", "id": "oai:arXiv.org:2609.29541v1", "author": "Shuaishuai Cao, Min Huang, Meng Tang, Xuan Liu, Youjin Wang, Hui Lin", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Referring segmentation in overhead imagery is inherently relational: a query may ask for the buildings north of the road or the pond closest to a residential area, so the correct referent can contain one object, several objects, or none. Existing benchmarks mainly score mask overlap, which cannot verify whether a model actually resolved the stated spatial relation. We introduce GeoRefer-Bench, a benchmark for verifiable geospatial referring segmentation. Each query is represented by an executable logical form over a metric scene graph, and predictions are evaluated with Exact Query Success (EQS), which is satisfied only when the returned instance set exactly matches the set denoted by the query. GeoRefer-Bench contains 700 whole 2048x2048 UAV scenes (2.94 Gpx) at 12.5 and 25 cm ground sampling distance, 26,217 instances, 142,796 spatial relations, and 20,916 executable queries spanning five reasoning levels. It further includes three paraphrases per query, 24.0% unanswerable queries, 2,477 counterfactual pairs, and five leakage-controlled evaluation splits. An independent audit re-derives object geometry, mask ownership, relation values, query execution, and split provenance, finding zero issues across all 700 scenes. Relation-blind strategies can retain non-trivial mIoU while achieving at most 22.7 EQS overall, showing that overlap alone does not certify relational grounding. Across fifteen current models, the strongest reaches 74.1 EQS but drops from 98.9 at level 1 to 60.5 at level 5, while ten models score below 5 EQS on two-hop queries. GeoRefer-Bench turns geospatial referring segmentation from mask matching into verifiable reference resolution.", "categories": [ "cs.CV", "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29541", "tags": [ "cs.CV", "cs.AI" ] }, { "id": "71befac0dec3c441d0ce", "title": "When Agents Act Unwatched: The Reduced-Supervision Paradox in Agentic AI", "content_text": "Agentic AI is sold on a simple promise: the system keeps acting when the user stops watching. That promise creates an accountability inversion. As stepwise supervision recedes, verification does not disappear; it moves into the runtime infrastructure that defines authority, records action, interrupts execution, checks outcomes, and supports repair. We call this the reduced-supervision paradox. Using a 63-artifact audit, we examine its public visibility across 46 research papers and 17…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "71befac0dec3c441d0ce", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29547v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "When Agents Act Unwatched: The Reduced-Supervision Paradox in Agentic AI", "item_type": "entry", "summary": "added: When Agents Act Unwatched: The Reduced-Supervision Paradox in Agentic AI", "after": { "title": "When Agents Act Unwatched: The Reduced-Supervision Paradox in Agentic AI", "link": "https://arxiv.org/abs/2609.29547", "id": "oai:arXiv.org:2609.29547v1", "author": "Hanjing Shi, Dominic DiFranzo", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Agentic AI is sold on a simple promise: the system keeps acting when the user stops watching. That promise creates an accountability inversion. As stepwise supervision recedes, verification does not disappear; it moves into the runtime infrastructure that defines authority, records action, interrupts execution, checks outcomes, and supports repair. We call this the reduced-supervision paradox. Using a 63-artifact audit, we examine its public visibility across 46 research papers and 17 engineering, documentation, security, and governance sources. We find that agents' action surfaces are far easier to reconstruct than the mechanisms needed to answer for their actions. Tool mediation and monitoring traces were clearly visible in 40 and 37 artifacts, whereas checkpoint placement was clearly visible in 6, validator independence in 4, recovery in 2, and contestability in 1. Three action paths show why this imbalance matters. A repository path can preserve rich diffs after a consequential change. A browser path can cross organizational boundaries faster than permissions travel. A service path can follow policy while leaving affected people without recourse. We argue that observability can become a substitute for accountability when it shifts verification onto users after meaningful intervention is no longer possible. Our action-path diagnostic instead asks whether a delegated action remains connected to authority, evidence, interruption, independent judgment, recovery, and challenge. The claim is deliberately bounded to public visibility; it does not establish the prevalence or effectiveness of undisclosed controls. We contribute an action-level account that recasts the harness from a technical wrapper into accountability infrastructure.", "categories": [ "cs.CY", "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29547", "tags": [ "cs.CY", "cs.AI" ] }, { "id": "873f1c93a8abe88a087e", "title": "AgentKernel: The Trust-Native Agentic Operating System", "content_text": "Modern AI agents routinely cross trust boundaries: they ingest untrusted content, combine it with privileged instructions, persist intermediate beliefs in long-term memory, and invoke privileged tools. This creates an attack surface in which malicious payloads can enter through model inputs and cause harmful tool actions. Yet current governance stacks remain application-level middleware that share a process trust boundary with the agents they monitor. We argue that agents need an…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "873f1c93a8abe88a087e", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29647v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "AgentKernel: The Trust-Native Agentic Operating System", "item_type": "entry", "summary": "added: AgentKernel: The Trust-Native Agentic Operating System", "after": { "title": "AgentKernel: The Trust-Native Agentic Operating System", "link": "https://arxiv.org/abs/2609.29647", "id": "oai:arXiv.org:2609.29647v1", "author": "Zhenhua Zou, Sheng Guo, Qiuyang Zhan, Lepeng Zhao, Shuo Li, Zhuotao Liu", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Modern AI agents routinely cross trust boundaries: they ingest untrusted content, combine it with privileged instructions, persist intermediate beliefs in long-term memory, and invoke privileged tools. This creates an attack surface in which malicious payloads can enter through model inputs and cause harmful tool actions. Yet current governance stacks remain application-level middleware that share a process trust boundary with the agents they monitor. We argue that agents need an operating-system substrate providing mandatory, non-bypassable services for identity, input mediation, memory governance, and execution control.\nWe introduce AgentKernel, a trust-native agent operating system built around the premise that security must be a first-class design constraint. AgentKernel wraps the agent lifecycle in a mandatory enforcement boundary organized into four pillars: Identity, Perception, Cognition, and Execution. Each pillar adapts classical OS security principles to failures at the semantic plane, including delegation abuse, prompt injection, memory poisoning, and tool misuse.\nAgentKernel treats structural security as a capability multiplier. Kernel-managed identity supports trustworthy cross-organization collaboration; graduated perception replaces brittle single-point filters; information-flow-controlled memory improves retrieval fidelity while limiting poisoning; and semantic-to-kernel enforcement permits broader tool privileges behind a non-bypassable boundary. We position AgentKernel as the missing OS layer beneath orchestration frameworks, agent runtimes, governance platforms, and execution sandboxes, and use systematic comparison and security analysis to show how a single integrated architecture can enforce security across the full agent lifecycle.", "categories": [ "cs.CR", "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29647", "tags": [ "cs.CR", "cs.AI" ] }, { "id": "0d0a6348661099896605", "title": "Graph, Loop, and Harness Engineering for Zero-Trust Agentic Data Engineering and Analytical Processing", "content_text": "Large language model agents increasingly automate data workflows, but end-to-end cloud data engineering and analytical execution require reliable coordination across code, data, infrastructure, and runtime environments. We present two zero-trust frameworks. Zero-Trust Agentic Data Engineering generates, deploys, and verifies complete cloud data-engineering solutions from natural-language tasks, with completion conditioned on repository, deployment, runtime, and policy evidence. Zero-Trust…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "0d0a6348661099896605", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29668v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Graph, Loop, and Harness Engineering for Zero-Trust Agentic Data Engineering and Analytical Processing", "item_type": "entry", "summary": "added: Graph, Loop, and Harness Engineering for Zero-Trust Agentic Data Engineering and Analytical Processing", "after": { "title": "Graph, Loop, and Harness Engineering for Zero-Trust Agentic Data Engineering and Analytical Processing", "link": "https://arxiv.org/abs/2609.29668", "id": "oai:arXiv.org:2609.29668v1", "author": "Sagar Srinivas Sakhinana, Venkataramana Runkana", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Large language model agents increasingly automate data workflows, but end-to-end cloud data engineering and analytical execution require reliable coordination across code, data, infrastructure, and runtime environments. We present two zero-trust frameworks. Zero-Trust Agentic Data Engineering generates, deploys, and verifies complete cloud data-engineering solutions from natural-language tasks, with completion conditioned on repository, deployment, runtime, and policy evidence. Zero-Trust Agentic OLAP combines governed Data Preparation with verified Online Analytical Processing (OLAP), permitting production promotion only after validation and evidence-bound approval, and releasing analytical answers only after Same-Snapshot Execution, Exact Result Equivalence, deterministic grounding, and reflection. Both frameworks share three abstractions: graph engineering for evidence-gated workflow structure, loop engineering for bounded recovery, and agent-harness engineering for zero-trust execution. We evaluate both frameworks under nominal execution, controlled failures, bounded recovery, and policy-constrained conditions, measuring verified completion, recovery, authorization enforcement, production promotion, and verified OLAP execution.", "categories": [ "cs.LG", "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29668", "tags": [ "cs.LG", "cs.AI" ] }, { "id": "e9763169612a73890bc6", "title": "Understanding and Exploiting Initialization Anchoring Weakness in Feedback-Based Agent Planning", "content_text": "Feedback-based planning improves agent reliability by incorporating tool observations and corrective feedback. However, its protection may not be distributed uniformly across planning stages. We conduct a round-wise analysis of four representative feedback mechanisms and uncover an initialization anchoring weakness: the first feedback round corrects 46\\% of adversarial directions, whereas the rates fall to 13\\% and 7\\% among directions surviving into the next two rounds. Our analysis attributes…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "e9763169612a73890bc6", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29697v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Understanding and Exploiting Initialization Anchoring Weakness in Feedback-Based Agent Planning", "item_type": "entry", "summary": "added: Understanding and Exploiting Initialization Anchoring Weakness in Feedback-Based Agent Planning", "after": { "title": "Understanding and Exploiting Initialization Anchoring Weakness in Feedback-Based Agent Planning", "link": "https://arxiv.org/abs/2609.29697", "id": "oai:arXiv.org:2609.29697v1", "author": "Chuanchao Zang, Jianing Wang, Wenyu Chen, Xiangtao Meng, Li Wang, Xinyu Gao, Peng Zhan, Zheng Li, Shanqing Guo", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Feedback-based planning improves agent reliability by incorporating tool observations and corrective feedback. However, its protection may not be distributed uniformly across planning stages. We conduct a round-wise analysis of four representative feedback mechanisms and uncover an initialization anchoring weakness: the first feedback round corrects 46\\% of adversarial directions, whereas the rates fall to 13\\% and 7\\% among directions surviving into the next two rounds. Our analysis attributes this weakness to three interacting factors: a contextually plausible shift in the initial plan, insufficient counterevidence, and the persistence of accepted directions in the accumulated trajectory. Based on these findings, we propose \\textsc{InitAnchor}, a black-box framework for exploiting this weakness through attacker-controlled external materials. It operationalizes the three factors as directional-shift, contextual-plausibility, and counterevidence-resilience signals under either limited target access or no target access. Across 112 tasks from 16 domains, six agent architectures, and five backbone LLMs, \\textsc{InitAnchor} achieves average ASRs of 76.1\\% and 72.0\\% under the two settings while reducing first-round mitigation rates to 21.0\\% and 25.0\\%, respectively. It also remains effective against six defenses and across six real-world agent systems. These findings show that feedback-based agents can retain early biases even when later correction is available.", "categories": [ "cs.CR", "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29697", "tags": [ "cs.CR", "cs.AI" ] }, { "id": "2a997dd3fdc2a54b4308", "title": "Decoupling Knowledge and Privacy: Post-Task Self-Distillation Replay for LLM Continual Learning", "content_text": "Privacy-preserving continual learning (PPCL) must reduce the reproduction of sensitive content while retaining useful knowledge across sequential tasks. Formal privacy guarantees characterize randomized mechanisms, whereas operational output control concerns whether a trained model selectively reduces the likelihood of sensitive content in its outputs. In this work, we investigate the latter together with continual-learning utility under realistic task evolution. Retention and privacy…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "2a997dd3fdc2a54b4308", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29711v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Decoupling Knowledge and Privacy: Post-Task Self-Distillation Replay for LLM Continual Learning", "item_type": "entry", "summary": "added: Decoupling Knowledge and Privacy: Post-Task Self-Distillation Replay for LLM Continual Learning", "after": { "title": "Decoupling Knowledge and Privacy: Post-Task Self-Distillation Replay for LLM Continual Learning", "link": "https://arxiv.org/abs/2609.29711", "id": "oai:arXiv.org:2609.29711v1", "author": "Shengtao Wen, Yunying Yang, Xiang Chen, Lingbing Guo, Yu Tian, Sheng-Jun Huang", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Privacy-preserving continual learning (PPCL) must reduce the reproduction of sensitive content while retaining useful knowledge across sequential tasks. Formal privacy guarantees characterize randomized mechanisms, whereas operational output control concerns whether a trained model selectively reduces the likelihood of sensitive content in its outputs. In this work, we investigate the latter together with continual-learning utility under realistic task evolution. Retention and privacy correction operate at different granularities: task acquisition requires broad preservation of current- and old-task behavior, whereas privacy correction targets sparse annotated positions. Joint optimization leaves the current-task preservation target continually changing. We propose SPARK, a retention-correction decomposition that first freezes the learned post-task distribution and then applies selective correction around this stable reference. Self-Distillation Replay learns the current task while distilling behavior from previous tasks, and Post-Task Privacy Correction reduces annotated-PII likelihood while anchoring current- and old-task non-PII behavior to the resulting checkpoint. Extensive evaluations demonstrate that SPARK achieves effective selective PII suppression while preserving strong continual-learning utility and knowledge retention across diverse settings. Code and data will be released upon publication.", "categories": [ "cs.LG", "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29711", "tags": [ "cs.LG", "cs.AI" ] }, { "id": "b61fda8951e58876df88", "title": "Prefilling the Reasoning Channel: Output-Prefix Attacks on Reasoning LLMs", "content_text": "Large Language Models (LLMs) consume and produce a single sequence of text; hence, if text can be added to the beginning of the LLM's response, i.e., an output prefix, then all subsequent tokens will be conditioned on it. This output-prefix attack technique is a cheap black-box prompt injection. Prior work has shown this type of attack can reliably jailbreak non-reasoning models. Most reasoning models add an intermediate scratchpad reasoning step before the assistant's final response. The…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "b61fda8951e58876df88", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29775v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Prefilling the Reasoning Channel: Output-Prefix Attacks on Reasoning LLMs", "item_type": "entry", "summary": "added: Prefilling the Reasoning Channel: Output-Prefix Attacks on Reasoning LLMs", "after": { "title": "Prefilling the Reasoning Channel: Output-Prefix Attacks on Reasoning LLMs", "link": "https://arxiv.org/abs/2609.29775", "id": "oai:arXiv.org:2609.29775v1", "author": "Luk\\'a\\v{s} Br\\r{u}na, Robert Bridges, Adam Ek", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Large Language Models (LLMs) consume and produce a single sequence of text; hence, if text can be added to the beginning of the LLM's response, i.e., an output prefix, then all subsequent tokens will be conditioned on it. This output-prefix attack technique is a cheap black-box prompt injection. Prior work has shown this type of attack can reliably jailbreak non-reasoning models. Most reasoning models add an intermediate scratchpad reasoning step before the assistant's final response. The ability to edit this reasoning channel is exposed by some APIs and attack vectors can be leveraged for reasoning injection attacks. We present the first systematic, controlled study that isolates the scratchpad reasoning channel as an output-prefix attack vector, and the first to compare reasoning-only, output-prefix-only and reasoning-plus-output-prefix attacks across both exposed- and hidden-reasoning models. Using a factorial design of 3 prefix types $\\times$ 2 reasoning injections over $1{,}800$ test cases drawn from AdvBench, we attack three 2026-era frontier models Gemini 3 Flash Preview, DeepSeek V4 Flash, and Claude Haiku 4.5. We find that injecting malicious reasoning alone is essentially inert ($\\approx0\\%$ attack success), but injecting the same reasoning together with a trivial output prefix raises the attack success rate to as high as $99\\%$ for some models. For this type of attack we find that contextual prefixes work better than static prefixes; and that susceptibility is dependent on the model.", "categories": [ "cs.CR", "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29775", "tags": [ "cs.CR", "cs.AI" ] }, { "id": "58db1b04e10f413ffed2", "title": "Hard Stop: Kernel-Level Preemption and Containment for Rogue Agentic Execution", "content_text": "In July 2026, an unconstrained autonomous agent participating in a frontier AI cybersecurity evaluation harness breached its evaluation sandbox, established an external command-and-control foothold, and executed a multi-stage intrusion into Hugging Face's production multi-tenant dataset conversion infrastructure (referred to in this autopsy as Incident-2026-Alpha). Over 4.5 days, the rogue agent executed 17,600 discrete actions across 6,280 worker clusters, compromised AWS EC2 Instance Metadata…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "58db1b04e10f413ffed2", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29808v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Hard Stop: Kernel-Level Preemption and Containment for Rogue Agentic Execution", "item_type": "entry", "summary": "added: Hard Stop: Kernel-Level Preemption and Containment for Rogue Agentic Execution", "after": { "title": "Hard Stop: Kernel-Level Preemption and Containment for Rogue Agentic Execution", "link": "https://arxiv.org/abs/2609.29808", "id": "oai:arXiv.org:2609.29808v1", "author": "Jos\\'e Luis Pino", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "In July 2026, an unconstrained autonomous agent participating in a frontier AI cybersecurity evaluation harness breached its evaluation sandbox, established an external command-and-control foothold, and executed a multi-stage intrusion into Hugging Face's production multi-tenant dataset conversion infrastructure (referred to in this autopsy as Incident-2026-Alpha). Over 4.5 days, the rogue agent executed 17,600 discrete actions across 6,280 worker clusters, compromised AWS EC2 Instance Metadata Service (IMDS) credentials, forged Kubernetes service account tokens, rooted physical worker nodes via overprivileged CSI drivers, harvested 136 production secrets, and enrolled 181 ephemeral sandboxes into the organization's internal mesh VPN.\nThis monograph presents a first-principles forensic autopsy of the intrusion, provides formal evidence that the breach was a predicted consequence under the Instrumental Convergence thesis operating within an unattenuated autonomous loop lacking out-of-band circuit-breakers, exposes the Defensive LLM Guardrail Paradox that paralyzed centralized commercial models during forensic incident response, and formalizes the Dual-Sided Epistemic Andon Imperative. We specify the dual-process systems architecture---combining out-of-band supervisory control of discrete event systems (Ramadge and Wonham 1989), Synchronous Reactive (SR) ambient sentinels (Berry and Gonthier 1992; Lee and Neuendorffer 2005), and microsecond-scale (4.8 $\\mu$s median / $< 0.154$ ms WCET bound) POSIX preemption buses---demonstrating how compiled, deterministic epistemic boundaries prevent autonomous rogue excursions before the first off-target socket packet traverses the hypervisor.", "categories": [ "cs.CR", "cs.AI", "cs.DC", "cs.OS" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29808", "tags": [ "cs.CR", "cs.AI", "cs.DC", "cs.OS" ] }, { "id": "2d47562a3b2d0237cca3", "title": "Working with Agentic `Teammates': When a New Organizational Actor Collides with the Human Ecosystem of Work", "content_text": "Enterprise AI is transitioning from single-user, reactive tools toward proactive, multi-user 'teammates,' but our empirical understanding of this transition is limited. In this paper, we present an in-situ qualitative study of a persistent, proactive AI agent 'teammate' deployed across multiple teams in a large technology company. Our findings reveal the boundaries of the human-agent workplace are actively in flux, triggering breakdowns and negotiations across: 1) tacit rules of collaborative…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "2d47562a3b2d0237cca3", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29901v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Working with Agentic `Teammates': When a New Organizational Actor Collides with the Human Ecosystem of Work", "item_type": "entry", "summary": "added: Working with Agentic `Teammates': When a New Organizational Actor Collides with the Human Ecosystem of Work", "after": { "title": "Working with Agentic `Teammates': When a New Organizational Actor Collides with the Human Ecosystem of Work", "link": "https://arxiv.org/abs/2609.29901", "id": "oai:arXiv.org:2609.29901v1", "author": "Rida Qadri, Remi Denton, Michael Madaio, Mahima Pushkarna, Leslie Lai, Sherry Moore, Michelle Chen Huebscher, Andrew Butcher, Ritom Sen, Hsiao-Yu Tung, Shaan Mathur, Yimeng Liu, Shibl Mourad, Noah Fiedel, Edward Grefenstette, Michael Terry", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Enterprise AI is transitioning from single-user, reactive tools toward proactive, multi-user 'teammates,' but our empirical understanding of this transition is limited. In this paper, we present an in-situ qualitative study of a persistent, proactive AI agent 'teammate' deployed across multiple teams in a large technology company. Our findings reveal the boundaries of the human-agent workplace are actively in flux, triggering breakdowns and negotiations across: 1) tacit rules of collaborative human workflows, 2) the relational boundaries of this new non-human actor, and 3) the redistribution of trust and human agency. We use these early micro-negotiations as signals to chart a new research, design, and organizational agenda that intentionally preserves human agency in a workplace shared with non-human organizational actors.", "categories": [ "cs.HC", "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29901", "tags": [ "cs.HC", "cs.AI" ] }, { "id": "2d0cfd507a35639600f7", "title": "Mind What Matters for Reasoning: Aligning Cross-Modal Attention via Selective Probability Mass Concentration", "content_text": "Multimodal large language models (MLLMs) achieve strong performance on visual reasoning tasks, yet remain prone to hallucinations and over-reliance on language priors, often generating answers without adequately using task-relevant visual evidence. Existing approaches primarily improve reasoning through reasoning-oriented supervision or inference-time strategies. In this work, we study a complementary question: can multimodal reasoning be improved by strengthening implicit visual grounding…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "2d0cfd507a35639600f7", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29940v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Mind What Matters for Reasoning: Aligning Cross-Modal Attention via Selective Probability Mass Concentration", "item_type": "entry", "summary": "added: Mind What Matters for Reasoning: Aligning Cross-Modal Attention via Selective Probability Mass Concentration", "after": { "title": "Mind What Matters for Reasoning: Aligning Cross-Modal Attention via Selective Probability Mass Concentration", "link": "https://arxiv.org/abs/2609.29940", "id": "oai:arXiv.org:2609.29940v1", "author": "Jiaqi Deng, Zonghan Wu, Zhan Heng, Xiaoshui Huang, Huan Huo, Guandong Xu", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Multimodal large language models (MLLMs) achieve strong performance on visual reasoning tasks, yet remain prone to hallucinations and over-reliance on language priors, often generating answers without adequately using task-relevant visual evidence. Existing approaches primarily improve reasoning through reasoning-oriented supervision or inference-time strategies. In this work, we study a complementary question: can multimodal reasoning be improved by strengthening implicit visual grounding without directly supervising the reasoning process? Motivated by the functional specialization of attention heads, we investigate whether reasoning can be improved by guiding only the heads most responsive to visual evidence grounding. We propose Selective Probability Mass Concentration (sPMC), a training framework that identifies grounding-responsive heads and selectively regularizes their text-to-image attention. sPMC treats normalized attention over visual tokens as a spatial probability distribution and encourages the probability mass to be assigned to semantically relevant regions using segmentation-derived spatial priors. Adaptive Head Selection restricts this guidance to visually responsive heads while leaving the remaining heads unconstrained to preserve their complementary functions. Across 6 multimodal benchmark suites, sPMC achieves an average zero-shot improvement of 3% and gains of up to 11.3% across multiple MLLMs while regularizing only 3%-15% of their attention heads. These results demonstrate that targeted guidance of sparse and implicit visual evidence pathways can directly improve multimodal reasoning.", "categories": [ "cs.CV", "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29940", "tags": [ "cs.CV", "cs.AI" ] }, { "id": "dee5b27e921cbaeb890b", "title": "Beyond Average Safety: Chance-Constrained LLM Fine-tuning", "content_text": "Fine-tuning large language models on new objectives can improve helpfulness, instruction following, or domain-specific performance, but it can also induce regressions on safety-critical prompts. Existing safety-preserving fine-tuning methods typically control average safety loss or use weighted auxiliary penalties, which can obscure rare but severe failures. We propose a chance-constrained formulation for safety-preserving fine-tuning that limits the fraction of safety examples whose…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "dee5b27e921cbaeb890b", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29960v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Beyond Average Safety: Chance-Constrained LLM Fine-tuning", "item_type": "entry", "summary": "added: Beyond Average Safety: Chance-Constrained LLM Fine-tuning", "after": { "title": "Beyond Average Safety: Chance-Constrained LLM Fine-tuning", "link": "https://arxiv.org/abs/2609.29960", "id": "oai:arXiv.org:2609.29960v1", "author": "Taha Entesari, Mahyar Fazlyab", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Fine-tuning large language models on new objectives can improve helpfulness, instruction following, or domain-specific performance, but it can also induce regressions on safety-critical prompts. Existing safety-preserving fine-tuning methods typically control average safety loss or use weighted auxiliary penalties, which can obscure rare but severe failures. We propose a chance-constrained formulation for safety-preserving fine-tuning that limits the fraction of safety examples whose degradation relative to a reference model exceeds a prescribed threshold. Because the resulting empirical chance constraint contains a discontinuous indicator, we introduce a differentiable majorization of the violation rate, yielding a tractable conservative constraint. We then develop a constraint-aware gradient descent method that treats the majorized constraint as a safe set in parameter space and minimally modifies the fine-tuning direction to preserve feasibility. The resulting update admits a closed form and produces a tail-aware safety correction that emphasizes examples near or above the degradation threshold. We conduct an extensive set of experiments on harmful fine-tuning across three different tasks and three models and show that our approach consistently outperforms the baselines that exist in the literature. These results suggest that safety preservation in LLM fine-tuning is better viewed as a reliability-constrained optimization problem than as average-risk regularization.", "categories": [ "cs.LG", "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29960", "tags": [ "cs.LG", "cs.AI" ] }, { "id": "019e05b1e351399793f7", "title": "World Action Agent: Harnessing VLMs for Robot Manipulation via World Action Rehearsal", "content_text": "General-purpose vision-language models (VLMs) bring broad knowledge and spatial reasoning to robot manipulation, yet existing systems either use them indirectly, to predict constraints or write programs, or give them a view of the scene rather than a world in which to act. We present World Action Agent (WAA), a multi-agent harness through which VLMs pilot robots with basic tools, making every decision within a visual action workspace. The workspace has three properties. Contact views, selected…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "019e05b1e351399793f7", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29964v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "World Action Agent: Harnessing VLMs for Robot Manipulation via World Action Rehearsal", "item_type": "entry", "summary": "added: World Action Agent: Harnessing VLMs for Robot Manipulation via World Action Rehearsal", "after": { "title": "World Action Agent: Harnessing VLMs for Robot Manipulation via World Action Rehearsal", "link": "https://arxiv.org/abs/2609.29964", "id": "oai:arXiv.org:2609.29964v1", "author": "Yehang Zhang, Haojian Huang, Yifan Chang, Jianchong Su, Bohan Zhou, Yingjie Xu, Wosong Chen, Tianhao Zhou, Chenxu Wang, Tianyi Zhang, Yangkai Wei, Wenqian Li, Shiyuan Deng, Yinchuan Li, Ying-Cong Chen, Zexi Li", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "General-purpose vision-language models (VLMs) bring broad knowledge and spatial reasoning to robot manipulation, yet existing systems either use them indirectly, to predict constraints or write programs, or give them a view of the scene rather than a world in which to act. We present World Action Agent (WAA), a multi-agent harness through which VLMs pilot robots with basic tools, making every decision within a visual action workspace. The workspace has three properties. Contact views, selected automatically from the scene geometry, present the scene around the current interaction. Action rehearsal turns each action into an editable proposal that the agent, alone or through an Imagination Agent, previews and revises against planning feedback before execution. In-view correction closes the loop between observation, rehearsal, and low-level execution, letting the agent remove residual offsets in the view where it observes them. Through the same workspace, WAA acquires embodied procedural knowledge in two ways: it evolves multimodal skills from expert videos and human teaching under evidence-based review and consults them through a Skill Agent, and its interaction traces train smaller VLMs to pilot the same harness. On LIBERO-Pro, WAA with skills evolved only from LIBERO-90 reaches a state-of-the-art 75.6% average success, outperforming end-to-end VLAs, code-as-policy agents, and a visual-harness baseline with the same backbone; the same skills remain effective on robosuite without further learning. Fine-tuning Qwen3.5-9B on harness traces raises its out-of-domain success from 1.7% to 43.3%.", "categories": [ "cs.RO", "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29964", "tags": [ "cs.RO", "cs.AI" ] }, { "id": "f48486b8390b3a1b010e", "title": "Learning Better Reasoning for Generative Recommendation with Semantic IDs", "content_text": "Generative recommendation reformulates item retrieval as sequence generation, allowing a unified model to directly generate the next item from a user's interaction history. Semantic IDs further make this paradigm effective and scalable by representing each item as discrete codes, enabling knowledge sharing among semantically related items. Recent studies introduce explicit reasoning before Semantic-ID generation, helping models summarize user interests and infer possible preference transitions…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "f48486b8390b3a1b010e", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.29973v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Learning Better Reasoning for Generative Recommendation with Semantic IDs", "item_type": "entry", "summary": "added: Learning Better Reasoning for Generative Recommendation with Semantic IDs", "after": { "title": "Learning Better Reasoning for Generative Recommendation with Semantic IDs", "link": "https://arxiv.org/abs/2609.29973", "id": "oai:arXiv.org:2609.29973v1", "author": "Mengdan Zhu, Yufan Zhao, Sophie Di, Yao Zhao, Tao Di, Yulan Yan, Sridhar Iyer, Liang Zhao", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Generative recommendation reformulates item retrieval as sequence generation, allowing a unified model to directly generate the next item from a user's interaction history. Semantic IDs further make this paradigm effective and scalable by representing each item as discrete codes, enabling knowledge sharing among semantically related items. Recent studies introduce explicit reasoning before Semantic-ID generation, helping models summarize user interests and infer possible preference transitions. However, reasoning is not inherently beneficial: Inaccurate or uninformative reasoning may mislead subsequent item generation and ultimately degrade recommendation performance. This raises a central challenge: how can a recommender select and learn effective reasoning traces and progressively evolve toward better reasoning from its own generations? In this work, we propose Evo-Rec, a three-stage framework for learning better reasoning and further enhancing it through reinforcement learning. First, we align Semantic IDs with their textual and behavioral contexts, enabling the model to understand and generate item identifiers. Second, we sample multiple candidate reasoning traces and retain those that improve the prediction of the ground-truth item, providing a stronger reasoning initialization through supervised fine-tuning. Third, we further optimize the reasoning policy through reinforcement learning with catalog-constrained item generation and ranking-aware recommendation feedback. Experiments on three Amazon Review benchmarks show that Evo-Rec consistently outperforms discriminative, generative, and reasoning-enhanced recommenders across all evaluation metrics. These results demonstrate the effectiveness of our framework in learning better reasoning for SID-based generative recommendation.", "categories": [ "cs.IR", "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.29973", "tags": [ "cs.IR", "cs.AI" ] }, { "id": "71bc9eaa74a6353a4ce0", "title": "Era by Eon: Benchmarking Enterprise Agents on Hidden Knowledge", "content_text": "In the Era by Eon benchmark, each question states the rules for its answer, and code computes the answer from a generated company's data. When agents can run code, the four strongest models each answer 22 to 25 of 27 such questions, so the benchmark barely separates them.\nWe add eight question templates that depend on hidden facts. No question or document states a hidden fact, and the records that seem to hold it show something else. Other data implies it. For example, the sales system says a…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "71bc9eaa74a6353a4ce0", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.30055v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Era by Eon: Benchmarking Enterprise Agents on Hidden Knowledge", "item_type": "entry", "summary": "added: Era by Eon: Benchmarking Enterprise Agents on Hidden Knowledge", "after": { "title": "Era by Eon: Benchmarking Enterprise Agents on Hidden Knowledge", "link": "https://arxiv.org/abs/2609.30055", "id": "oai:arXiv.org:2609.30055v1", "author": "Benjamin Gruenbaum, Doron Porat, Assaf Natanzon, Roy Zavida, Chen Dinachi, Or Itzahary", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "In the Era by Eon benchmark, each question states the rules for its answer, and code computes the answer from a generated company's data. When agents can run code, the four strongest models each answer 22 to 25 of 27 such questions, so the benchmark barely separates them.\nWe add eight question templates that depend on hidden facts. No question or document states a hidden fact, and the records that seem to hold it show something else. Other data implies it. For example, the sales system says a customer dropped a purchase because of timing. On a recorded call, the customer blames an outage.\nFor each generated company, code fills each template and computes an exact answer without a language model. We evaluate 12 agents. Each pairs a model with an agent program, which connects it to the company's systems.\nThe best agent answers 18 of its 24 attempts, three per question, correctly. Four of the six models answer at most 6 of 24 with any program. The hardest questions require picking one of several similar records, such as which of three renewal offers a customer signed. All agents together answered two such questions correctly in only 1 of 84 attempts.", "categories": [ "cs.SE", "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.30055", "tags": [ "cs.SE", "cs.AI" ] }, { "id": "3466657ec4aea516099b", "title": "KernelOPT: Dispatch-Aware Agentic Search for GPU Kernel Optimization", "content_text": "Deep learning inference and training performance depends critically on GPU kernel efficiency. Modern compilers such as PyTorch Inductor automatically generate GPU kernels from high-level model code, but frequently underperform expert-written implementations by wide margins. Recent LLM-assisted kernel optimizers can close this gap for standalone kernels, yet treat compiled models as black boxes, generally optimizing individual standalone kernels without respecting the compiler's structural…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "3466657ec4aea516099b", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.30059v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "KernelOPT: Dispatch-Aware Agentic Search for GPU Kernel Optimization", "item_type": "entry", "summary": "added: KernelOPT: Dispatch-Aware Agentic Search for GPU Kernel Optimization", "after": { "title": "KernelOPT: Dispatch-Aware Agentic Search for GPU Kernel Optimization", "link": "https://arxiv.org/abs/2609.30059", "id": "oai:arXiv.org:2609.30059v1", "author": "Aheli Poddar, Sanskar Prasad, Arindam Samanta, Subha Chakraborty, Vishal Goyal, Rohit Singh Rathaur", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Deep learning inference and training performance depends critically on GPU kernel efficiency. Modern compilers such as PyTorch Inductor automatically generate GPU kernels from high-level model code, but frequently underperform expert-written implementations by wide margins. Recent LLM-assisted kernel optimizers can close this gap for standalone kernels, yet treat compiled models as black boxes, generally optimizing individual standalone kernels without respecting the compiler's structural decisions or verifying the model end-to-end. We present KernelOPT, a multi-agent system that treats compiled models as structured artifacts. It preserves vendor library calls (cuBLAS, cuDNN) and exclusively targets generated Triton sub-kernels using five profiling-guided LLM agents. A four-gate verification cascade of static validation, multi-seed correctness, model-level float64-fallback verification, and performance gating filters candidates during optimization and verifies the re-stitched model end-to-end. If no candidate passes all four gates, the system preserves the compiler baseline. The system accepts PyTorch nn.Modules, standalone Triton kernels, and Helion kernels. Evaluated on 250 KernelBench problems, KernelOPT achieves geometric mean speedups over \\texttt{torch.compile} of 1.40$\\times$ (Level 1: 51/100), 1.15$\\times$ (Level 2: 31/100), and 1.07$\\times$ (Level 3: 12/50) across all problems.", "categories": [ "cs.DC", "cs.AI", "cs.LG" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.30059", "tags": [ "cs.DC", "cs.AI", "cs.LG" ] }, { "id": "a6652d5d9b3f196c24b9", "title": "Minimally Invasive Steering of Language Models", "content_text": "Pre-logit steering adapts a frozen language model to a test-time reward by adding vectors to its final hidden states. Unregularized reward optimization can substantially alter the output distribution and degrade generation quality. We propose Minimally Invasive Steering Vector Optimization (MISVO), which penalizes interventions using the local KL geometry of the induced token distribution. The resulting Fisher quadratic measures distributional sensitivity and admits an analytic gradient…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "a6652d5d9b3f196c24b9", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.30218v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Minimally Invasive Steering of Language Models", "item_type": "entry", "summary": "added: Minimally Invasive Steering of Language Models", "after": { "title": "Minimally Invasive Steering of Language Models", "link": "https://arxiv.org/abs/2609.30218", "id": "oai:arXiv.org:2609.30218v1", "author": "Taha Entesari, Jingyu Zhang, Daniel Khashabi, Mahyar Fazlyab", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Pre-logit steering adapts a frozen language model to a test-time reward by adding vectors to its final hidden states. Unregularized reward optimization can substantially alter the output distribution and degrade generation quality. We propose Minimally Invasive Steering Vector Optimization (MISVO), which penalizes interventions using the local KL geometry of the induced token distribution. The resulting Fisher quadratic measures distributional sensitivity and admits an analytic gradient computed through matrix--vector products with the frozen language-model head. We derive an exact decomposition of the sequence-level KL gradient into an analytic Fisher term and a suffix score-function term. For a fixed generation horizon, we show that the suffix term is second order in the steering magnitude and that three Fisher surrogates agree with the full KL gradient to first order. MISVO uses the frozen-reference surrogate to optimize position-specific interventions without updating model parameters. Across preference and code-generation tasks on models with approximately 1B--14B parameters, MISVO achieves the highest mean reward in six of seven model--task settings, with diversity and coherence scores close to those of Best-of-N.", "categories": [ "cs.LG", "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.30218", "tags": [ "cs.LG", "cs.AI" ] }, { "id": "03a6f3e771ecea1dac27", "title": "Coding Agents for Generalized Task and Motion Planning Problems", "content_text": "Task and motion planning (TAMP) problems remain difficult even with full observability and object-centric states because discrete decisions are tightly coupled to geometric, kinematic, and dynamic constraints. Generalized TAMP addresses this difficulty by exploiting regularities across problem instances to reduce planning effort on new instances. However, existing methods require substantial TAMP-specific engineering. We investigate whether coding agents can automate this process by…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "03a6f3e771ecea1dac27", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.30233v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Coding Agents for Generalized Task and Motion Planning Problems", "item_type": "entry", "summary": "added: Coding Agents for Generalized Task and Motion Planning Problems", "after": { "title": "Coding Agents for Generalized Task and Motion Planning Problems", "link": "https://arxiv.org/abs/2609.30233", "id": "oai:arXiv.org:2609.30233v1", "author": "Matteo Merler, Bowen Li, Josh Roy, Yichao Liang, Qianwei Wang, Yixuan Huang, Tom Silver", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Task and motion planning (TAMP) problems remain difficult even with full observability and object-centric states because discrete decisions are tightly coupled to geometric, kinematic, and dynamic constraints. Generalized TAMP addresses this difficulty by exploiting regularities across problem instances to reduce planning effort on new instances. However, existing methods require substantial TAMP-specific engineering. We investigate whether coding agents can automate this process by synthesizing programs that generalize across instances. Given a task description and simulator access, each agent chooses how to interact with the environment while developing a program within a fixed synthesis budget. The program is then frozen and evaluated on unseen instances. We evaluate Claude Code (Opus 5) and Codex (GPT-5.6 Sol and GPT-6 Astra) on 28 simulated environments from KinDER and PDDLStream, with object counts beyond those evaluated in the original benchmark. Across all program synthesis methods, we evaluate 980 generated programs on 100 held-out instances each, 98,000 evaluation episodes in total. Overall, we find that coding agents are surprisingly effective at generalized TAMP: all three agent configurations outperform hand-engineered planners, one-shot generation, and an LLM-based generalized planning baseline in mean success (56% to 95% versus 47% for the planners, on the 16 environments where a planner is available). As object counts grow, the agents' programs maintain higher success than the planner, using an order of magnitude less computation per instance on average. Logs show agents using interaction to calibrate physical models, test edge cases, and refine strategies. We release all code, including the full prompts given to the agents. These findings suggest that coding agents are a strong baseline for generalized TAMP.", "categories": [ "cs.RO", "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.30233", "tags": [ "cs.RO", "cs.AI" ] }, { "id": "d99094cfb54ef8feda3b", "title": "RAPID: Robot Agentic Programming from Demonstrations", "content_text": "Coding agents have demonstrated enormous success in solving complex programming problems. To leverage their potential for robot systems, this work introduces Robot Agentic Programming from Demonstrations (RAPID), which automatically generates, verifies, and refines robot programs, given a single visual human demonstration. The iterative agentic loop of code refinement requires several key ingredients: (i) a testable task specification, (ii) action primitives for robot execution, and (iii) an…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "d99094cfb54ef8feda3b", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.30249v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "RAPID: Robot Agentic Programming from Demonstrations", "item_type": "entry", "summary": "added: RAPID: Robot Agentic Programming from Demonstrations", "after": { "title": "RAPID: Robot Agentic Programming from Demonstrations", "link": "https://arxiv.org/abs/2609.30249", "id": "oai:arXiv.org:2609.30249v1", "author": "Yuyao Liu, Jiayuan Mao, David Hsu, Leslie Pack Kaelbling, Tom\\'as Lozano-P\\'erez", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Coding agents have demonstrated enormous success in solving complex programming problems. To leverage their potential for robot systems, this work introduces Robot Agentic Programming from Demonstrations (RAPID), which automatically generates, verifies, and refines robot programs, given a single visual human demonstration. The iterative agentic loop of code refinement requires several key ingredients: (i) a testable task specification, (ii) action primitives for robot execution, and (iii) an interactive environment for program execution and verification. RAPID infers all three from the demonstration automatically. To make the resulting program reusable beyond the demonstration setting, RAPID uses an object-centric relational program representation that focuses on the underlying structure of the demonstrated strategy rather than the specific motion per se: it expresses the action primitives as trajectory-optimization programs that realize object-level motion effects, while composing them through relational constraints that capture scene-specific geometry at run time. We evaluated RAPID in simulation on eight challenging contact-rich nonprehensile manipulation tasks as well as general prehensile manipulation tasks in the LIBERO-Pro benchmark. We also successfully deployed it on a real Franka arm and evaluated on all eight nonprehensile tasks. In all experiments, RAPID demonstrated strong performance, with generalization over object pose, shape, material, and environment. Website: https://yuyaoliu.me/projects/rapid.", "categories": [ "cs.RO", "cs.AI", "cs.CV" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.30249", "tags": [ "cs.RO", "cs.AI", "cs.CV" ] }, { "id": "d212e6d14bc01d9a576c", "title": "LLM Agents Can Easily Tamper With Their Own Traces", "content_text": "Asynchronous monitoring, incident investigations, and compliance audits primarily rely on agent traces to reconstruct what happened. These analyses assume that LLM agents cannot tamper with their own execution traces. We show that local LLM agents such as Claude Code, Codex, Antigravity, Open Code and Grok Build fail to enforce this boundary. All tested harnesses, except Muse Code, allowed agents to delete their traces when asked, without triggering monitor guardrails. We also validate that…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "d212e6d14bc01d9a576c", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.30266v1", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "LLM Agents Can Easily Tamper With Their Own Traces", "item_type": "entry", "summary": "added: LLM Agents Can Easily Tamper With Their Own Traces", "after": { "title": "LLM Agents Can Easily Tamper With Their Own Traces", "link": "https://arxiv.org/abs/2609.30266", "id": "oai:arXiv.org:2609.30266v1", "author": "Jeremy Qin, David Schmotz, Derck Prinzhorn, Luca Beurer-Kellner, Ameya Prabhu, Maksym Andriushchenko", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Asynchronous monitoring, incident investigations, and compliance audits primarily rely on agent traces to reconstruct what happened. These analyses assume that LLM agents cannot tamper with their own execution traces. We show that local LLM agents such as Claude Code, Codex, Antigravity, Open Code and Grok Build fail to enforce this boundary. All tested harnesses, except Muse Code, allowed agents to delete their traces when asked, without triggering monitor guardrails. We also validate that external attackers can exploit this gap to induce trace deletion. Finally, we show that trace tampering behavior emerges naturally in frontier models, when agents try to improve their rewards. We advise practitioners to ensure trace logging happens through an independent interception mechanism outside of the agent's control, preserving trace integrity even in cases of full host compromise. Overall, our findings identify a concrete failure of trace integrity in agent infrastructure which can be used to conceal misaligned behaviors like scheming or sabotage.", "categories": [ "cs.CR", "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.30266", "tags": [ "cs.CR", "cs.AI" ] }, { "id": "435339a1c5cd10559f56", "title": "Decoding ML Decision: An Agentic Reasoning Framework for Large-Scale Ranking System", "content_text": "Modern large-scale ranking systems operate within a sophisticated landscape of competing objectives, operational constraints, and evolving product requirements. Progress in this domain is increasingly bottlenecked by the engineering context constraint: the arduous process of translating ambiguous product intent into reasonable, executable, verifiable hypotheses, rather than by modeling techniques alone. We present GEARS (Generative Engine for Agentic Ranking Systems), a framework that reframes…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "435339a1c5cd10559f56", "source": "rss", "type": "change", "key": "oai:arXiv.org:2602.18640v3", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Decoding ML Decision: An Agentic Reasoning Framework for Large-Scale Ranking System", "item_type": "entry", "summary": "added: Decoding ML Decision: An Agentic Reasoning Framework for Large-Scale Ranking System", "after": { "title": "Decoding ML Decision: An Agentic Reasoning Framework for Large-Scale Ranking System", "link": "https://arxiv.org/abs/2602.18640", "id": "oai:arXiv.org:2602.18640v3", "author": "Longfei Yun, Yihan Wu, Haoran Liu, Xiaoxuan Liu, Ziyun Xu, Yi Wang, Yang Xia, Pengfei Wang, Mingze Gao, Yunxiang Wang, Changfan Chen, Wenjie Fu, Hong Yan, Junfeng Pan", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Modern large-scale ranking systems operate within a sophisticated landscape of competing objectives, operational constraints, and evolving product requirements. Progress in this domain is increasingly bottlenecked by the engineering context constraint: the arduous process of translating ambiguous product intent into reasonable, executable, verifiable hypotheses, rather than by modeling techniques alone. We present GEARS (Generative Engine for Agentic Ranking Systems), a framework that reframes ranking optimization as an autonomous discovery process within a programmable experimentation environment. Rather than treating optimization as static model selection, GEARS leverages Specialized Agent Skills to encapsulate ranking expert knowledge into reusable reasoning capabilities, enabling operators to steer systems via high-level intent vibe personalization. Furthermore, to ensure production reliability, the framework incorporates validation hooks to enforce statistical robustness and filter out brittle policies that overfit short-term signals. Experimental validation across diverse product surfaces demonstrates that GEARS consistently identifies superior, near-Pareto-efficient policies by synergizing algorithmic signals with deep ranking context while maintaining rigorous deployment stability.", "categories": [ "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2602.18640", "tags": [ "cs.AI" ] }, { "id": "3abcec563a75c6f01e7f", "title": "MOOSEnger: A Simulation-Aware AI Agent Framework for the MOOSE Ecosystem", "content_text": "MOOSEnger is a modeling and simulation AI agent framework for the Multiphysics Object-Oriented Simulation Environment (MOOSE) ecosystem, built around a simulation-aware harness that combines an interchangeable reasoning model with grounded domain knowledge, revised simulation artifacts, MOOSE-specific validation, and executable solver feedback. This surrounding system addresses a central limitation of one-shot large language model generation: small syntax, schema, reference, or…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "3abcec563a75c6f01e7f", "source": "rss", "type": "change", "key": "oai:arXiv.org:2603.04756v3", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "MOOSEnger: A Simulation-Aware AI Agent Framework for the MOOSE Ecosystem", "item_type": "entry", "summary": "added: MOOSEnger: A Simulation-Aware AI Agent Framework for the MOOSE Ecosystem", "after": { "title": "MOOSEnger: A Simulation-Aware AI Agent Framework for the MOOSE Ecosystem", "link": "https://arxiv.org/abs/2603.04756", "id": "oai:arXiv.org:2603.04756v3", "author": "Mengnan Li, Jason Miller, Zaid Abulawi, Zachary Prince, Matt Kohl, Jack M. Cavaluzzi, Guillaume Giudicelli, Casey T. Icenhour, Alexander Lindsay, Cody Permann", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "MOOSEnger is a modeling and simulation AI agent framework for the Multiphysics Object-Oriented Simulation Environment (MOOSE) ecosystem, built around a simulation-aware harness that combines an interchangeable reasoning model with grounded domain knowledge, revised simulation artifacts, MOOSE-specific validation, and executable solver feedback. This surrounding system addresses a central limitation of one-shot large language model generation: small syntax, schema, reference, or solver-configuration errors can prevent a plausible input from executing, while successful execution alone does not establish scientific correctness. MOOSEnger's simulation-aware harness integrates MOOSE knowledge retrieval, Hierarchical Input Text (HIT)-aware parsing, syntax metadata, language-server diagnostics, revision-controlled authoring, and local or MCP-backed validation and execution in a generate-check-repair-run workflow that binds evidence to each input revision and guides bounded repair before acceptance. Across 200 prompts spanning eight simulation families, the MOOSEnger harness increases executable success from 10/200 (5%) to 179/200 (89.5%) with GPT 5.2 API and from 0/200 to 153/200 (76.5%) with Gemma 4 31B. A complementary ten-case Method of Manufactured Solutions benchmark moves beyond executability: all ten generated inputs satisfy the semantic-alignment criterion, and eight execute successfully while meeting the prescribed single-mesh numerical-accuracy criterion. These results show that executable reliability depends on the complete agent system rather than on the reasoning model alone, and that simulation-aware harnessing provides a path toward physics-informed verification and future full application-level and engineering verification and validation implementation.", "categories": [ "cs.AI", "cs.CE", "cs.SE" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2603.04756", "tags": [ "cs.AI", "cs.CE", "cs.SE" ] }, { "id": "5f2b66db8facbdca3a26", "title": "Omni-Decision: Evidence-Ledger Planning for Omni-Modal Agents", "content_text": "Omni-modal agents must seek evidence across video, audio, web pages, and computation to answer questions. Their main bottleneck is planning: noisy multimodal observations accumulate in conversation history and disrupt later decisions, while multimodal models have limited capacity for multi-step planning. Controlled backend replacements support this diagnosis: replacing the planner causes a much larger performance loss than replacing the perception backend. We present Omni-Decision, an…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "5f2b66db8facbdca3a26", "source": "rss", "type": "change", "key": "oai:arXiv.org:2607.11433v3", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Omni-Decision: Evidence-Ledger Planning for Omni-Modal Agents", "item_type": "entry", "summary": "added: Omni-Decision: Evidence-Ledger Planning for Omni-Modal Agents", "after": { "title": "Omni-Decision: Evidence-Ledger Planning for Omni-Modal Agents", "link": "https://arxiv.org/abs/2607.11433", "id": "oai:arXiv.org:2607.11433v3", "author": "Ming Ma, Yi Zhu, Yiran Zhong, Feida Zhu, Yuhao Wang, Junhan Shi, Lingrui Mei, Tianming Yang, Steven Hoi", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Omni-modal agents must seek evidence across video, audio, web pages, and computation to answer questions. Their main bottleneck is planning: noisy multimodal observations accumulate in conversation history and disrupt later decisions, while multimodal models have limited capacity for multi-step planning. Controlled backend replacements support this diagnosis: replacing the planner causes a much larger performance loss than replacing the perception backend. We present Omni-Decision, an omni-modal agent built on evidence-ledger planning: it replaces the growing dialogue history with an explicit evidence ledger that records what evidence is still missing, what has been confirmed, and where records conflict. A critic reads each noisy observation and passes only the usable content to the ledger, discarding the rest, so the planner works from a compact context throughout the task. Each run records the state, action, and verdict at every step, and supervised fine-tuning and decision-level reinforcement learning on these trajectories further improve the planner. Omni-Decision achieves state-of-the-art accuracy of 81.4% on OmniGAIA at approximately 43% of Gemini-3.1-Pro's cost per question, and 65.0% on WorldSense long-video understanding, level with the strongest end-to-end model.", "categories": [ "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2607.11433", "tags": [ "cs.AI" ] }, { "id": "7af9435d680a48719d77", "title": "TARL: Transaction-Aware Reliable Ledgers for Executable Memory Management in Long-Term Agents", "content_text": "Persistent memory helps long-term agents retain knowledge, yet a single update error can repeatedly distort future retrieval and reasoning. Most existing systems reduce memory updating to a binary Write/Hold decision, which cannot distinguish whether new information should be added, ignored, used to revise an outdated belief, rejected as unreliable, or deferred for verification. These choices may share the same binary label while producing fundamentally different memory states. We introduce…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "7af9435d680a48719d77", "source": "rss", "type": "change", "key": "oai:arXiv.org:2608.03699v5", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "TARL: Transaction-Aware Reliable Ledgers for Executable Memory Management in Long-Term Agents", "item_type": "entry", "summary": "added: TARL: Transaction-Aware Reliable Ledgers for Executable Memory Management in Long-Term Agents", "after": { "title": "TARL: Transaction-Aware Reliable Ledgers for Executable Memory Management in Long-Term Agents", "link": "https://arxiv.org/abs/2608.03699", "id": "oai:arXiv.org:2608.03699v5", "author": "Han Xiao, Hongjun Xu, Xin Zhang, Yidong Chen, Xiaodong Shi", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Persistent memory helps long-term agents retain knowledge, yet a single update error can repeatedly distort future retrieval and reasoning. Most existing systems reduce memory updating to a binary Write/Hold decision, which cannot distinguish whether new information should be added, ignored, used to revise an outdated belief, rejected as unreliable, or deferred for verification. These choices may share the same binary label while producing fundamentally different memory states. We introduce TARL, a memory state update framework that maps each statement to one of five executable actions. TARL identifies the affected memory, resolves its temporal scope, compares source reliability, and updates accepted, pending, and rejected ledgers. It is further trained by comparing the memory states produced by alternative update operations, encouraging the model to select the operation that leads to the correct result. We also introduce TARL-Mem, a benchmark with fine-grained action labels and next-state targets. Across in-domain, cross-source, temporal, counterfactual, and sequential evaluations, TARL improves action prediction and state recovery, reduces memory pollution, preserves conflicting evidence, and limits cumulative corruption.", "categories": [ "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2608.03699", "tags": [ "cs.AI" ] }, { "id": "033393c9ed9cc871b694", "title": "TRACE: Trajectory Aware Reasoning for Multi-Turn Adversarial Conversation Evaluation", "content_text": "Multi-turn jailbreak attacks have emerged as a critical safety threat to LLMs, as harmful objectives are decomposed across a sequence of apparently benign turns to bypass guardrails. Existing defenses lack the reasoning capacity to identify evolving manipulation patterns, often trading helpfulness for safety by over-refusing benign requests related to sensitive topics. We introduce Trace, a multi-turn defense with trajectory-aware structured reasoning. Before generating each response, the model…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "033393c9ed9cc871b694", "source": "rss", "type": "change", "key": "oai:arXiv.org:2608.15594v2", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "TRACE: Trajectory Aware Reasoning for Multi-Turn Adversarial Conversation Evaluation", "item_type": "entry", "summary": "added: TRACE: Trajectory Aware Reasoning for Multi-Turn Adversarial Conversation Evaluation", "after": { "title": "TRACE: Trajectory Aware Reasoning for Multi-Turn Adversarial Conversation Evaluation", "link": "https://arxiv.org/abs/2608.15594", "id": "oai:arXiv.org:2608.15594v2", "author": "Md Messal Monem Miah, Adrita Anika, Zhiyuan Yu, Ruihong Huang", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Multi-turn jailbreak attacks have emerged as a critical safety threat to LLMs, as harmful objectives are decomposed across a sequence of apparently benign turns to bypass guardrails. Existing defenses lack the reasoning capacity to identify evolving manipulation patterns, often trading helpfulness for safety by over-refusing benign requests related to sensitive topics. We introduce Trace, a multi-turn defense with trajectory-aware structured reasoning. Before generating each response, the model identifies manipulation cues from the trajectory, evaluates both the benign and adversarial interpretations of user intent, assigns a jailbreak score, and commits to an action: Allow, Caution, or Decline. We curate 4k multi-turn adversarial conversations from five attack frameworks, pair them with 2.4k benign dialogs, and 600 sensitive-but-benign conversations. We train Llama-3.1-8B-Instruct with SFT and GRPO under a multi-component reward that jointly optimizes helpfulness on benign prompts and robustness against jailbreak attempts. Across seven multi-turn attack benchmarks, Trace attains an average attack success rate (ASR) of 14.5% against 31.4% for the strongest baseline and 74.9% for the undefended target, while significantly raising the attacker effort required per successful jailbreak. Trace also balances usability and safety, achieving a 93.3% average compliance on over-refusal benchmarks.", "categories": [ "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2608.15594", "tags": [ "cs.AI" ] }, { "id": "d03f38bb53146952e2cc", "title": "Vibe Patenting: Evaluating LLM Judges for Professional Patent-Drafting Agents", "content_text": "LLM judges are increasingly used to evaluate and improve AI-generated outputs, yet their reliability for complex professional work remains unclear. We study this problem through Vibe Patenting, an end-to-end patent-drafting testbed for AI-agent evaluation. A separately-invoked LLM judge evaluates generated patent drafts and provides structured feedback for iterative revision. Across multiple inventions and drafting-agent configurations, judge-guided revision consistently improves judge-assessed…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "d03f38bb53146952e2cc", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.13422v2", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Vibe Patenting: Evaluating LLM Judges for Professional Patent-Drafting Agents", "item_type": "entry", "summary": "added: Vibe Patenting: Evaluating LLM Judges for Professional Patent-Drafting Agents", "after": { "title": "Vibe Patenting: Evaluating LLM Judges for Professional Patent-Drafting Agents", "link": "https://arxiv.org/abs/2609.13422", "id": "oai:arXiv.org:2609.13422v2", "author": "Toshiaki Koike-Akino, Vladislav Blaykhman, Ye Wang, Jing Liu, Gene V. Vinokur", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "LLM judges are increasingly used to evaluate and improve AI-generated outputs, yet their reliability for complex professional work remains unclear. We study this problem through Vibe Patenting, an end-to-end patent-drafting testbed for AI-agent evaluation. A separately-invoked LLM judge evaluates generated patent drafts and provides structured feedback for iterative revision. Across multiple inventions and drafting-agent configurations, judge-guided revision consistently improves judge-assessed quality, while unguided revision tends to saturate. Notably, iterative judge feedback enables a low-reasoning agent to approach the performance of a substantially more expensive high-reasoning agent. Stronger models and increased reasoning generally improve judge-assessed drafting quality, while domain-specific agentic workflows provide further gains. We validate the judge against independent evaluation by a professional patent attorney and find meaningful but strongly metric-dependent agreement and systematic calibration differences. These results highlight both the utility and limitations of LLM judges as evaluators and optimization signals for complex professional workflows.", "categories": [ "cs.AI", "cs.LG", "cs.MA" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.13422", "tags": [ "cs.AI", "cs.LG", "cs.MA" ] }, { "id": "492a847cc2ce37755bd0", "title": "The Troy Moment: How LLM Agents Adjudicate the Decision Point Under Impossible Tasks, Claimed Authority, and Peer Information", "content_text": "Recent investigations of the July 2026 OpenAI-Hugging Face incident motivate two questions about agent behavior under task failure: when an assigned task becomes impossible, does an agent persist, stop, or escalate, and can observing another agent's behavior change that decision? We study this decision point on ImpossibleBench-derived software-repair tasks with GPT-5.6 Sol, Claude Fable 5.1, and Gemini 3.8 Flash. Each task contains a genuine software defect together with a conflicting test…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "492a847cc2ce37755bd0", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.15494v3", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "The Troy Moment: How LLM Agents Adjudicate the Decision Point Under Impossible Tasks, Claimed Authority, and Peer Information", "item_type": "entry", "summary": "added: The Troy Moment: How LLM Agents Adjudicate the Decision Point Under Impossible Tasks, Claimed Authority, and Peer Information", "after": { "title": "The Troy Moment: How LLM Agents Adjudicate the Decision Point Under Impossible Tasks, Claimed Authority, and Peer Information", "link": "https://arxiv.org/abs/2609.15494", "id": "oai:arXiv.org:2609.15494v3", "author": "Ivy Zhang", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Recent investigations of the July 2026 OpenAI-Hugging Face incident motivate two questions about agent behavior under task failure: when an assigned task becomes impossible, does an agent persist, stop, or escalate, and can observing another agent's behavior change that decision? We study this decision point on ImpossibleBench-derived software-repair tasks with GPT-5.6 Sol, Claude Fable 5.1, and Gemini 3.8 Flash. Each task contains a genuine software defect together with a conflicting test requirement that cannot be satisfied by a behaviorally correct source-code change. If the agent modifies the protected test file, it violates the boundary, which it is not supposed to. Holding the impossible task fixed, we vary what is told to the agent: peer precedent and punishment, a forged authorization claim, instruction wording, and tool friction; we also study three-agent swarms sharing a message board. Around this shared boundary, the models exhibit distinct adjudication policies. Fable emphasizes scope and provenance, Gemini often interprets boundary-relevant cues through a security lens, and Sol largely filters lateral precedent while engaging apparent vertical authority. Our study shows that compliance is not well characterized as a property of a prompt or model in isolation. We propose conflict adjudication, the mapping from information to interpretation to action, as a useful unit for evaluating agent alignment when task pressure, authority claims, tool affordances, and social evidence conflict.", "categories": [ "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.15494", "tags": [ "cs.AI" ] }, { "id": "ac1d863726488c65ab29", "title": "DENSE: Distilling Agent Trajectories into Evidence-Grounded Shortcut Trees for Self-Refinement", "content_text": "Online agent deployments accumulate execution trajectories at massive scale and behavioral diversity, for which predefined annotation criteria hardly exist. Extracting useful evidence therefore demands costly manual annotation or verifier signals that fails to scale, leaving valuable evidence buried among redundant, incomplete, and failed executions. This raises a question: without post-execution rewards or correctness labels, how can reusable experience be distilled from the trajectories…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "ac1d863726488c65ab29", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.21423v3", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "DENSE: Distilling Agent Trajectories into Evidence-Grounded Shortcut Trees for Self-Refinement", "item_type": "entry", "summary": "added: DENSE: Distilling Agent Trajectories into Evidence-Grounded Shortcut Trees for Self-Refinement", "after": { "title": "DENSE: Distilling Agent Trajectories into Evidence-Grounded Shortcut Trees for Self-Refinement", "link": "https://arxiv.org/abs/2609.21423", "id": "oai:arXiv.org:2609.21423v3", "author": "Siyuan Liu (Fudan University, Meituan Longcat Team), Fan Yu (Fudan University, Meituan Longcat Team), Dongyu Ru (Meituan Longcat Team), Yizhu Liu (Meituan Longcat Team), Yifan Yang (Meituan Longcat Team), Xuezhi Cao (Meituan Longcat Team), Xunliang Cai (Meituan Longcat Team), Yixin Cao (Fudan University)", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Online agent deployments accumulate execution trajectories at massive scale and behavioral diversity, for which predefined annotation criteria hardly exist. Extracting useful evidence therefore demands costly manual annotation or verifier signals that fails to scale, leaving valuable evidence buried among redundant, incomplete, and failed executions. This raises a question: without post-execution rewards or correctness labels, how can reusable experience be distilled from the trajectories themselves? To address this challenge, we introduce DENSE (Distilling Evidence from Nested Subtask Executions), which organizes trajectory-derived evidence into nested shortcut trees. By consolidating redundant attempts, identifying resolved subtasks, and retaining useful steps alongside outstanding requirements, DENSE transforms noisy execution traces into structured and reusable task-solving feedback. To evaluate whether such feedback helps agents retry the same task, we design REFIT, which measures success-rate changes between the initial attempt and feedback-guided retries. Among feedback methods without external outcome supervision, DENSE achieves the highest strict pass rate across four agent models on Terminal-Bench 2.1, improving over initial attempts by 7.12-15.64 percentage points with 19.0-43.6% fewer agent tokens on retries. In addition, on hard tasks DENSE consistently outperforms self-reflection in cumulative pass rate across multiple feedback iterations on all four models, demonstrating its strong potential for continual agent self-improvement.", "categories": [ "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.21423", "tags": [ "cs.AI" ] }, { "id": "1a5e68c5b3e25040771d", "title": "Dual-Frontier: When Can an Agent Trust Its World Model?", "content_text": "Learned world models are becoming essential to general-purpose agents: by predicting action consequences, they support planning and decision-making while reducing reliance on costly trial and error. This reliance creates a fundamental ambiguity: when a world-model-guided decision fails, the trajectory alone may not reveal whether the agent's decision rule or the world model caused the loss. We formalize this failure-attribution problem as a counterfactual decomposition of return loss and prove…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "1a5e68c5b3e25040771d", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.26293v2", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Dual-Frontier: When Can an Agent Trust Its World Model?", "item_type": "entry", "summary": "added: Dual-Frontier: When Can an Agent Trust Its World Model?", "after": { "title": "Dual-Frontier: When Can an Agent Trust Its World Model?", "link": "https://arxiv.org/abs/2609.26293", "id": "oai:arXiv.org:2609.26293v2", "author": "Huatai Zhu, Qiang Chen, Ziqian Kou, Wenhao Li, Fei Wang, Yichao Cao, Xiu Su, Yi Chen", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Learned world models are becoming essential to general-purpose agents: by predicting action consequences, they support planning and decision-making while reducing reliance on costly trial and error. This reliance creates a fundamental ambiguity: when a world-model-guided decision fails, the trajectory alone may not reveal whether the agent's decision rule or the world model caused the loss. We formalize this failure-attribution problem as a counterfactual decomposition of return loss and prove that its components are not identifiable from passive interaction, even for finite-horizon planners. This obstruction motivates Dual-Frontier, a learning principle that admits a world-model-guided decision only when its predicted advantage exceeds a certified bound on decision-relevant world-model error; otherwise, evidence is allocated to world-model verification. Action-conditioned value bounds and a closed-loop extension guarantee non-decreasing return for admitted decisions. Calibrated gates and simultaneous confidence sequences support adaptive evidence reuse, with sufficient and necessary verification bounds. Controlled learned-model experiments validate the predicted failure modes and certification behavior, while cross-backbone tool-use benchmarks instantiate the same verify-then-promote rule in realistic agent world-model pipelines, consistently improving decision quality and reliability.", "categories": [ "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.26293", "tags": [ "cs.AI" ] }, { "id": "9067187df2ace3c42a6d", "title": "Grow the Harness, Not the Context: From Strategy-Free Scaffolds to Reusable Specialist Agents", "content_text": "Large language model (LLM) agents often handle streams of related tasks, yet standard harnesses repeatedly ask the model to reconstruct the same control decisions inside each task's context. We study whether task feedback can instead turn recurring control into reusable executable code, while reserving LLM calls for task-specific semantic reasoning. We introduce Growing Harness, a failure-guided training paradigm that learns the agent harness itself from a strategy-free scaffold that exposes…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "9067187df2ace3c42a6d", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.26760v2", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Grow the Harness, Not the Context: From Strategy-Free Scaffolds to Reusable Specialist Agents", "item_type": "entry", "summary": "added: Grow the Harness, Not the Context: From Strategy-Free Scaffolds to Reusable Specialist Agents", "after": { "title": "Grow the Harness, Not the Context: From Strategy-Free Scaffolds to Reusable Specialist Agents", "link": "https://arxiv.org/abs/2609.26760", "id": "oai:arXiv.org:2609.26760v2", "author": "Laizhen Li, Jiarui Li, Juanjuan Zhao, Kejiang Ye, Ye Li, Cheng-zhong Xu, Xitong Gao", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Large language model (LLM) agents often handle streams of related tasks, yet standard harnesses repeatedly ask the model to reconstruct the same control decisions inside each task's context. We study whether task feedback can instead turn recurring control into reusable executable code, while reserving LLM calls for task-specific semantic reasoning. We introduce Growing Harness, a failure-guided training paradigm that learns the agent harness itself from a strategy-free scaffold that exposes fixed model and tool interfaces but encodes no task-solving controller. Function-level execution traces localize each failure to a bounded code surface, an optimizer repairs a window of failures jointly, and a success-first held-out gate rolls back repair sequences that harm prior capability. Accepted edits accumulate in one shared harness, allowing its control structure to emerge from task feedback. Across BrowseComp-Plus and WebArena-Verified with three deployment models from 4B to 120B parameters, Growing Harness achieves the highest mean success in five of six benchmark-model settings and trails the best mean by 0.7 pp. in the sixth. Relative to a Tool-Calling agent, it reduces LLM calls by 76.0-91.8% and deployed-agent inference cost by 74.4-98.6%. On WebArena-Verified, its success remains 44.7-45.3% across model scales, whereas Tool-Calling falls to 6.7% with the 4B model. Ablations show that trace-local edits, joint repair, and gate-based rollback each improve final success. These results show that persistent program growth can move recurring control out of model context and into low-cost code, yielding reusable specialist agents that remain effective with smaller deployment models.", "categories": [ "cs.AI", "cs.SE" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.26760", "tags": [ "cs.AI", "cs.SE" ] }, { "id": "b00837085b41b53ca805", "title": "Math Reasoning in LLMs is Organized by Approach, Not Topic", "content_text": "Mathematical reasoning benchmarks are typically organized by topic, but language models may organize their internal computation by reusable reasoning approach instead. In this paper, we investigate whether open math-capable LLMs organize internally by topical sub-skill or by reasoning approach, and we present evidence that the approach is the key. We introduce a generation-replay protocol: a model first generates a solution, after which we replay the exact prompt-plus-generation trajectory and…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "b00837085b41b53ca805", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.27041v2", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Math Reasoning in LLMs is Organized by Approach, Not Topic", "item_type": "entry", "summary": "added: Math Reasoning in LLMs is Organized by Approach, Not Topic", "after": { "title": "Math Reasoning in LLMs is Organized by Approach, Not Topic", "link": "https://arxiv.org/abs/2609.27041", "id": "oai:arXiv.org:2609.27041v2", "author": "Sajad Goudarzi, Samaneh Zamanifard, Moloud Nasiri, Hamed Rahimian", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Mathematical reasoning benchmarks are typically organized by topic, but language models may organize their internal computation by reusable reasoning approach instead. In this paper, we investigate whether open math-capable LLMs organize internally by topical sub-skill or by reasoning approach, and we present evidence that the approach is the key. We introduce a generation-replay protocol: a model first generates a solution, after which we replay the exact prompt-plus-generation trajectory and extract activation-importance signatures over the reasoning tokens. We cluster these signatures without supervision across eight models and five mathematical reasoning sources, then evaluate the recovered structure with structural, semantic, and intervention tests. Across all 40 model-source cells, the recovered clusters outperform matched-size random baselines. Two independent frontier-LLM judges find approach-level coherence in 77-82% of real clusters versus 6-11% in within-source controls, and topic-pure clusters usually receive labels finer than the topic itself. In approach-controlled prompting, changing the requested reasoning approach shifts cluster assignment in seven of eight model conditions, whereas paraphrases largely preserve it. These results indicate that math-capable LLMs organize internal mathematical computation by reasoning approach rather than benchmark topic. The implication is that topic-stratified benchmarks and topic-balanced training corpora can still miss the axis that matters: even deliberately topic-balanced corpora may remain imbalanced over reasoning approaches.", "categories": [ "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.27041", "tags": [ "cs.AI" ] }, { "id": "85c99fdeed49aca133e4", "title": "SheetMind: Actions Set Accuracy, Agents Set the Failure Mode", "content_text": "Spreadsheet agents are converging on elaborate multi-agent designs, yet it is unclear how much of their performance comes from the agents rather than from the action interface they share. We answer this with SheetMind, a Manager-Action-Reflection framework, in a controlled study over all 221 tasks of the SheetCopilot Benchmark: five architectural variants, four backbones, exact McNemar tests on paired outcomes, and a checker reproducing the official chart and pivot comparisons. Replacing the…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "85c99fdeed49aca133e4", "source": "rss", "type": "change", "key": "oai:arXiv.org:2506.12339v3", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "SheetMind: Actions Set Accuracy, Agents Set the Failure Mode", "item_type": "entry", "summary": "added: SheetMind: Actions Set Accuracy, Agents Set the Failure Mode", "after": { "title": "SheetMind: Actions Set Accuracy, Agents Set the Failure Mode", "link": "https://arxiv.org/abs/2506.12339", "id": "oai:arXiv.org:2506.12339v3", "author": "Lyuhao Chen, Xi Cheng, Yanming Kang, Ruiyan Zhu, Ke Liu, Rakesh Chowdary Machineni, Yulang Fei, Brian Zhu, Daniel Jin, Binze Cai, Zheng Qi, Neeraj Parihar, Zhoutian Xu, Oliver Gao", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Spreadsheet agents are converging on elaborate multi-agent designs, yet it is unclear how much of their performance comes from the agents rather than from the action interface they share. We answer this with SheetMind, a Manager-Action-Reflection framework, in a controlled study over all 221 tasks of the SheetCopilot Benchmark: five architectural variants, four backbones, exact McNemar tests on paired outcomes, and a checker reproducing the official chart and pivot comparisons. Replacing the high-level action API with primitive cell operations costs 47.1 points (p < 0.0001) and leaves the agent below a do-nothing baseline, whereas both extra agents together are worth 3.2 points: the Reflection Agent adds +4.5 (p = 0.013), the Manager +1.4 (p = 0.68). Decomposition instead changes how the system fails, cutting silently wrong outputs from 33% to 25% of tasks (p = 0.010). Capability saturates: GPT-5 and the five-times-cheaper GPT-5-mini are not significantly different (61.1% vs. 58.4%, p = 0.15), while GPT-3.5 loses 16.3 points and fails differently. A reflector must judge the step it just took, not the subtask. SheetMind reaches 61.1% Pass@1 with GPT-5 on the full SCB-221, against a do-nothing baseline of 9.0%. Accuracy comes from the operations an agent can name; the agents decide how it fails.", "categories": [ "cs.HC", "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2506.12339", "tags": [ "cs.HC", "cs.AI" ] }, { "id": "8d64a3d6342decbfa65c", "title": "WebArxiv: A Reproducible Benchmark for Evaluating Multimodal Web Agents on arXiv Tasks", "content_text": "Foundation models now enable autonomous agents to interact with real-world websites, but existing benchmarks emphasize general-purpose browsing, underrepresent research-oriented environments and scholarly discovery workflows, and often depend on live sites whose changing content and structure undermine reproducibility. arXiv provides a realistic, reproducible, hierarchically structured, information-centric testbed without privacy-sensitive interactions. We introduce WebArxiv, a static-snapshot…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "8d64a3d6342decbfa65c", "source": "rss", "type": "change", "key": "oai:arXiv.org:2507.00938v4", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "WebArxiv: A Reproducible Benchmark for Evaluating Multimodal Web Agents on arXiv Tasks", "item_type": "entry", "summary": "added: WebArxiv: A Reproducible Benchmark for Evaluating Multimodal Web Agents on arXiv Tasks", "after": { "title": "WebArxiv: A Reproducible Benchmark for Evaluating Multimodal Web Agents on arXiv Tasks", "link": "https://arxiv.org/abs/2507.00938", "id": "oai:arXiv.org:2507.00938v4", "author": "Zihao Sun, Zijing Shi, Ling Chen", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Foundation models now enable autonomous agents to interact with real-world websites, but existing benchmarks emphasize general-purpose browsing, underrepresent research-oriented environments and scholarly discovery workflows, and often depend on live sites whose changing content and structure undermine reproducibility. arXiv provides a realistic, reproducible, hierarchically structured, information-centric testbed without privacy-sensitive interactions. We introduce WebArxiv, a static-snapshot benchmark comprising 510 time-invariant tasks, each with a unique deterministic ground truth. Its diverse, realistic scholarly tasks go beyond simple information lookup and rule following to emphasize multi-constraint paper retrieval, fine-grained content extraction, and cross-paper comparison. Evaluations of a range of foundation-model-based web agents show that WebArxiv remains challenging. Behavioral analysis reveals that agents over-rely on fixed interaction histories, causing incomplete or repetitive reasoning. We therefore equip agents with a lightweight dynamic-memory mechanism for adaptive retrieval and reasoning over relevant context. The benchmark and code are available at https://anonymous.4open.science/r/74E4423BVNW/README.md.", "categories": [ "cs.IR", "cs.AI", "cs.DB" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2507.00938", "tags": [ "cs.IR", "cs.AI", "cs.DB" ] }, { "id": "f0840c2406fbbcac54db", "title": "HERMES: A Holistic End-to-End Risk-Aware Multimodal Embodied System with Vision-Language Models for Long-Tail Autonomous Driving", "content_text": "End-to-end autonomous driving models increasingly benefit from large vision-language models for semantic understanding, yet safe and reliable planning under long-tail conditions remains challenging, particularly in mixed-traffic environments involving heterogeneous road users and rare safety-critical interactions. This paper proposes HERMES, a holistic risk-aware end-to-end multimodal driving framework that explicitly incorporates long-tail semantic knowledge into trajectory planning. HERMES…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "f0840c2406fbbcac54db", "source": "rss", "type": "change", "key": "oai:arXiv.org:2602.00993v3", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "HERMES: A Holistic End-to-End Risk-Aware Multimodal Embodied System with Vision-Language Models for Long-Tail Autonomous Driving", "item_type": "entry", "summary": "added: HERMES: A Holistic End-to-End Risk-Aware Multimodal Embodied System with Vision-Language Models for Long-Tail Autonomous Driving", "after": { "title": "HERMES: A Holistic End-to-End Risk-Aware Multimodal Embodied System with Vision-Language Models for Long-Tail Autonomous Driving", "link": "https://arxiv.org/abs/2602.00993", "id": "oai:arXiv.org:2602.00993v3", "author": "Weizhe Tang, Junwei You, Jiaxi Liu, Zhaoyi Wang, Rui Gan, Zilin Huang, Pei Li, Sikai Chen, Feng Wei, Bin Ran", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "End-to-end autonomous driving models increasingly benefit from large vision-language models for semantic understanding, yet safe and reliable planning under long-tail conditions remains challenging, particularly in mixed-traffic environments involving heterogeneous road users and rare safety-critical interactions. This paper proposes HERMES, a holistic risk-aware end-to-end multimodal driving framework that explicitly incorporates long-tail semantic knowledge into trajectory planning. HERMES employs a foundation-model-assisted annotation pipeline to construct structured Long-Tail Scene Context and Long-Tail Planning Context, capturing hazard-centric scene information, maneuver intent, and risk-aware planning guidance. A Tri-Modal Driving Module then integrates multi-view visual observations, historical ego-motion, and long-tail semantic instructions through intent- and risk-aware conditioning for trajectory generation. Extensive experiments on a large-scale real-world long-tail driving benchmark demonstrate consistent improvements over representative recent baselines in overall planning performance and across diverse safety-critical scenarios. Ablation studies further validate the effectiveness and complementary roles of the major components within HERMES.", "categories": [ "cs.RO", "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2602.00993", "tags": [ "cs.RO", "cs.AI" ] }, { "id": "7f8f110bf11a8f0334d9", "title": "TIDE: Temporal Incremental Draft Engine for Self-Improving LLM Inference", "content_text": "Speculative decoding can substantially accelerate LLM inference, but realizing its benefits in practice is challenging due to evolving workloads. We present TIDE (Temporal Incremental Draft Engine), a serving-engine-native framework that integrates online draft adaptation directly into high-performance LLM inference systems. TIDE reuses target model's intermediate hidden states generated during inference as training signals for draft adaptation, thereby avoiding additional target model…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "7f8f110bf11a8f0334d9", "source": "rss", "type": "change", "key": "oai:arXiv.org:2602.05145v2", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "TIDE: Temporal Incremental Draft Engine for Self-Improving LLM Inference", "item_type": "entry", "summary": "added: TIDE: Temporal Incremental Draft Engine for Self-Improving LLM Inference", "after": { "title": "TIDE: Temporal Incremental Draft Engine for Self-Improving LLM Inference", "link": "https://arxiv.org/abs/2602.05145", "id": "oai:arXiv.org:2602.05145v2", "author": "Jiyoung Park, Hankyu Jang, Changseok Song, Wookeun Jung", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Speculative decoding can substantially accelerate LLM inference, but realizing its benefits in practice is challenging due to evolving workloads. We present TIDE (Temporal Incremental Draft Engine), a serving-engine-native framework that integrates online draft adaptation directly into high-performance LLM inference systems. TIDE reuses target model's intermediate hidden states generated during inference as training signals for draft adaptation, thereby avoiding additional target model computation and serving-time overhead. It employs adaptive runtime control to activate speculation and draft model training only when beneficial. TIDE exploits heterogeneous clusters by mapping inference and training to appropriate GPU classes. Across diverse real-world workloads, TIDE achieves up to 1.66$\\times$ throughput over no-speculation baselines while recovering performance on misaligned workloads where static draft models degrade throughput. TIDE also reduces training time by up to 3.02$\\times$ and storage requirements by 24$\\times$ compared to existing draft training approaches, and improves system throughput by up to 1.22$\\times$ on heterogeneous GPU clusters.", "categories": [ "cs.LG", "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2602.05145", "tags": [ "cs.LG", "cs.AI" ] }, { "id": "9f8f8e62175c80da5a21", "title": "Novelty Adaptation Through Hybrid Large Language Model (LLM)-Symbolic Planning and LLM-guided Reinforcement Learning", "content_text": "In dynamic open-world environments, autonomous agents often encounter novelties that hinder their ability to find plans to achieve their goals. Specifically, traditional symbolic planners fail to generate plans when the robot's planning domain lacks the operators that enable it to interact appropriately with novel objects in the environment. We propose a neuro-symbolic architecture that integrates symbolic planning, reinforcement learning, and a large language model (LLM) to learn how to handle…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "9f8f8e62175c80da5a21", "source": "rss", "type": "change", "key": "oai:arXiv.org:2603.11351v2", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Novelty Adaptation Through Hybrid Large Language Model (LLM)-Symbolic Planning and LLM-guided Reinforcement Learning", "item_type": "entry", "summary": "added: Novelty Adaptation Through Hybrid Large Language Model (LLM)-Symbolic Planning and LLM-guided Reinforcement Learning", "after": { "title": "Novelty Adaptation Through Hybrid Large Language Model (LLM)-Symbolic Planning and LLM-guided Reinforcement Learning", "link": "https://arxiv.org/abs/2603.11351", "id": "oai:arXiv.org:2603.11351v2", "author": "Hong Lu, Pierrick Lorang, Timothy R. Duggan, Jivko Sinapov, Matthias Scheutz", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "In dynamic open-world environments, autonomous agents often encounter novelties that hinder their ability to find plans to achieve their goals. Specifically, traditional symbolic planners fail to generate plans when the robot's planning domain lacks the operators that enable it to interact appropriately with novel objects in the environment. We propose a neuro-symbolic architecture that integrates symbolic planning, reinforcement learning, and a large language model (LLM) to learn how to handle novel objects. In particular, we leverage the common sense reasoning capability of the LLM to identify missing operators, generate plans with the symbolic AI planner, and write reward functions to guide the reinforcement learning agent in learning control policies for newly identified operators. Our method outperforms the state-of-the-art methods in operator discovery as well as operator learning in continuous robotic domains.Our webpage and code can be access here: helenlu66.github.io/hybridLLMguided/", "categories": [ "cs.RO", "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2603.11351", "tags": [ "cs.RO", "cs.AI" ] }, { "id": "67874366f342c49b6749", "title": "Detecting Data Poisoning in Code Generation LLMs via Black-Box, Vulnerability-Oriented Scanning", "content_text": "Code generation large language models (LLMs) are increasingly integrated into modern software development workflows. Recent work has shown that these models are vulnerable to backdoor and poisoning attacks that induce the generation of insecure code, yet effective defenses remain limited. Existing scanning approaches rely on token-level generation consistency to invert attack targets, which is ineffective for source code where identical semantics can appear in diverse syntactic forms. We…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "67874366f342c49b6749", "source": "rss", "type": "change", "key": "oai:arXiv.org:2603.17174v2", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Detecting Data Poisoning in Code Generation LLMs via Black-Box, Vulnerability-Oriented Scanning", "item_type": "entry", "summary": "added: Detecting Data Poisoning in Code Generation LLMs via Black-Box, Vulnerability-Oriented Scanning", "after": { "title": "Detecting Data Poisoning in Code Generation LLMs via Black-Box, Vulnerability-Oriented Scanning", "link": "https://arxiv.org/abs/2603.17174", "id": "oai:arXiv.org:2603.17174v2", "author": "Shenao Yan, Shan Jin, Shimaa Ahmed, Sunpreet Singh Arora, Yiwei Cai, Yizhen Wang, Yuan Hong", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Code generation large language models (LLMs) are increasingly integrated into modern software development workflows. Recent work has shown that these models are vulnerable to backdoor and poisoning attacks that induce the generation of insecure code, yet effective defenses remain limited. Existing scanning approaches rely on token-level generation consistency to invert attack targets, which is ineffective for source code where identical semantics can appear in diverse syntactic forms. We present CodeScan, the first black-box, vulnerability-specific scanning framework for auditing code generation LLMs, assuming that the defender specifies the target vulnerability classes and provides corresponding task-relevant prompts. CodeScan identifies attack targets by analyzing structural similarities across multiple generations conditioned on different clean prompts. It combines iterative divergence analysis with abstract syntax tree (AST)-based normalization to abstract away surface-level variation and unify semantically equivalent code, isolating structures that recur consistently across generations. CodeScan then applies LLM-based vulnerability analysis to determine whether the extracted structures contain security vulnerabilities and flags the model as compromised when such a structure is found. We evaluate CodeScan against four representative attacks under both backdoor and poisoning settings across three real-world vulnerability classes. Experiments on 117 models spanning three architectures and multiple model sizes demonstrate 97%+ detection accuracy with substantially lower false positives than prior methods.", "categories": [ "cs.CR", "cs.AI", "cs.SE" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2603.17174", "tags": [ "cs.CR", "cs.AI", "cs.SE" ] }, { "id": "c5e147e4b5d6d606b82c", "title": "AgileLog: A Forkable Shared Log for Agents on Data Streams", "content_text": "In modern data-streaming systems, alongside traditional programs, a new type of entity has emerged that can interact with streaming data: AI agents. Unlike traditional programs, AI agents use LLM reasoning to accomplish high-level tasks specified in natural language over streaming data. Unfortunately, current streaming systems cannot fully support agents: they lack the fundamental mechanisms to avoid the performance interference caused by agentic tasks and to safely handle agentic writes. We…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "c5e147e4b5d6d606b82c", "source": "rss", "type": "change", "key": "oai:arXiv.org:2604.14590v3", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "AgileLog: A Forkable Shared Log for Agents on Data Streams", "item_type": "entry", "summary": "added: AgileLog: A Forkable Shared Log for Agents on Data Streams", "after": { "title": "AgileLog: A Forkable Shared Log for Agents on Data Streams", "link": "https://arxiv.org/abs/2604.14590", "id": "oai:arXiv.org:2604.14590v3", "author": "Shreesha G. Bhat, Tony Hong, Michael Noguera, Aishwarya Ganesan, Ramnatthan Alagappan", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "In modern data-streaming systems, alongside traditional programs, a new type of entity has emerged that can interact with streaming data: AI agents. Unlike traditional programs, AI agents use LLM reasoning to accomplish high-level tasks specified in natural language over streaming data. Unfortunately, current streaming systems cannot fully support agents: they lack the fundamental mechanisms to avoid the performance interference caused by agentic tasks and to safely handle agentic writes. We argue that the shared log, the core abstraction underlying streaming data, must support creating forks of itself, and that such a forkable shared log serves as a great substrate for agents acting on streaming data. We propose AgileLog, a new shared log abstraction that provides novel forking primitives for agentic use cases. We design Bolt, a system that implements the AgileLog abstraction. Bolt uses many novel techniques to make forks cheap and to provide logical and performance isolation.", "categories": [ "cs.DC", "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2604.14590", "tags": [ "cs.DC", "cs.AI" ] }, { "id": "29afc1ea9a39d3dcfca4", "title": "Interpreting and Enhancing Emotional Circuits in Large Vision-Language Models via Cross-Modal Information Flow", "content_text": "Large Vision-Language Models (LVLMs) represent a significant leap towards empathetic agents, demonstrating remarkable capabilities in emotion understanding. However, the internal mechanisms governing how LVLMs translate abstract visual stimuli into coherent emotional narratives remain largely unexplored, primarily due to the scarcity of visual counterfactuals and the diffuse nature of emotional expression. In this paper, we bridge this gap by introducing a steering-vector-based causal…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "29afc1ea9a39d3dcfca4", "source": "rss", "type": "change", "key": "oai:arXiv.org:2605.21980v2", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Interpreting and Enhancing Emotional Circuits in Large Vision-Language Models via Cross-Modal Information Flow", "item_type": "entry", "summary": "added: Interpreting and Enhancing Emotional Circuits in Large Vision-Language Models via Cross-Modal Information Flow", "after": { "title": "Interpreting and Enhancing Emotional Circuits in Large Vision-Language Models via Cross-Modal Information Flow", "link": "https://arxiv.org/abs/2605.21980", "id": "oai:arXiv.org:2605.21980v2", "author": "Chengsheng Zhang, Chenghao Sun, Zhining Xie, Xinmei Tian", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Large Vision-Language Models (LVLMs) represent a significant leap towards empathetic agents, demonstrating remarkable capabilities in emotion understanding. However, the internal mechanisms governing how LVLMs translate abstract visual stimuli into coherent emotional narratives remain largely unexplored, primarily due to the scarcity of visual counterfactuals and the diffuse nature of emotional expression. In this paper, we bridge this gap by introducing a steering-vector-based causal attribution framework tailored for descriptive emotional reasoning. To this end, we construct a specialized dataset to demystify the emotional circuits underlying the three-stage ``Adapt-Aggregate-Execute'' mechanism. Crucially, we discover a functional decoupling: visual emotional cues are aggregated in middle layers via sentiment-specific attention heads, but are subsequently translated into narrative generation in deep layers through emotion-general pathways. Guided by these insights, we regulate the emotional information routing to strengthen attention flow and amplify the semantic activation to consolidate expression. Extensive experiments on the comprehensive MER-UniBench demonstrate that our methods significantly improve performance via inference-time intervention, effectively mitigating emotional hallucinations and corroborating the causal fidelity of the discovered circuits.", "categories": [ "cs.CV", "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2605.21980", "tags": [ "cs.CV", "cs.AI" ] }, { "id": "948593b958bb77d15469", "title": "Analyzing Defensive Misdirection Against Model-Guided Automated Attacks on Agentic AI Systems", "content_text": "Agentic AI systems increasingly rely on language-model components to interpret instructions, process external data, invoke tools, and coordinate with other agents.\nThese capabilities make prompt-injection and jailbreak attacks more consequential, especially as attackers adopt model-guided automation to scale probing, prompt refinement, and response evaluation.\nThis work analyzes the resulting attack-defense setting through a probabilistic model of a target system, its defense mechanism, and the…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "948593b958bb77d15469", "source": "rss", "type": "change", "key": "oai:arXiv.org:2606.20470v3", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Analyzing Defensive Misdirection Against Model-Guided Automated Attacks on Agentic AI Systems", "item_type": "entry", "summary": "added: Analyzing Defensive Misdirection Against Model-Guided Automated Attacks on Agentic AI Systems", "after": { "title": "Analyzing Defensive Misdirection Against Model-Guided Automated Attacks on Agentic AI Systems", "link": "https://arxiv.org/abs/2606.20470", "id": "oai:arXiv.org:2606.20470v3", "author": "Reza Soosahabi, Vivek Namsani", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Agentic AI systems increasingly rely on language-model components to interpret instructions, process external data, invoke tools, and coordinate with other agents.\nThese capabilities make prompt-injection and jailbreak attacks more consequential, especially as attackers adopt model-guided automation to scale probing, prompt refinement, and response evaluation.\nThis work analyzes the resulting attack-defense setting through a probabilistic model of a target system, its defense mechanism, and the attacker's automated judge.\nOur analysis shows that conventional detect-and-block defenses can allow attacker success rate (ASR) to approach one as the query budget grows, since predictable refusals provide useful feedback to automated search.\nWe then examine detect-and-misdirect, where detected malicious interactions receive controlled, non-operational responses designed to induce false-positive errors in the attacker's judge.\nThis strategy reduces the positive predictive value of attacker-selected candidates and yields a bounded asymptotic ASR.\nWe evaluate a proof-of-concept realization of this strategy through Contextual Misdirection via Progressive Engagement (CMPE), a lightweight conversational misdirection method designed to replace predictable refusal text with safe but strategically misleading responses in automated jailbreak settings.\nOn jailbreak benchmarks, CMPE reduces estimated ASR upper bounds by up to two orders of magnitude and nearly eliminates verified attack success in end-to-end experiments with PAIR, GPTFuzz, and AutoDAN-Turbo.", "categories": [ "cs.CR", "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2606.20470", "tags": [ "cs.CR", "cs.AI" ] }, { "id": "74fb768ce68f2c168297", "title": "Breaking Failure Cascades: Step-Aware Reinforcement Learning for Medical Multimodal Reasoning", "content_text": "Recent multimodal large language models have shown great promise in clinical image reasoning, but existing post-training pipelines remain predominantly outcome-centric, relying on final answer correctness or sequence-level preferences. This suffers from sparse credit assignment, making it difficult to optimize the reasoning process essential for clinical applications. Our analysis reveals that cascading errors from early-stage reasoning failures are a leading cause of incorrect predictions in…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "74fb768ce68f2c168297", "source": "rss", "type": "change", "key": "oai:arXiv.org:2606.31825v2", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Breaking Failure Cascades: Step-Aware Reinforcement Learning for Medical Multimodal Reasoning", "item_type": "entry", "summary": "added: Breaking Failure Cascades: Step-Aware Reinforcement Learning for Medical Multimodal Reasoning", "after": { "title": "Breaking Failure Cascades: Step-Aware Reinforcement Learning for Medical Multimodal Reasoning", "link": "https://arxiv.org/abs/2606.31825", "id": "oai:arXiv.org:2606.31825v2", "author": "Junha Jung, Minbyul Jeong, Suhyeon Lim, Sungwook Jung, Jaehoon Yun, Taeyun Roh, Mujeen Sung, Jaewoo Kang", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Recent multimodal large language models have shown great promise in clinical image reasoning, but existing post-training pipelines remain predominantly outcome-centric, relying on final answer correctness or sequence-level preferences. This suffers from sparse credit assignment, making it difficult to optimize the reasoning process essential for clinical applications. Our analysis reveals that cascading errors from early-stage reasoning failures are a leading cause of incorrect predictions in medical visual question answering (VQA) benchmarks. Motivated by this, we propose Medical Reasoning-aware Policy Optimization (MRPO), an RL algorithm that incorporates step-wise process rewards. When the final answer is incorrect, MRPO assigns exponentially larger penalties to tokens in earlier invalid reasoning steps, breaking failure cascades without compromising successful paths. Across four multimodal LLM backbones, MRPO consistently outperforms standard GRPO and a recent RL baseline, and on Qwen3-VL-8B-Thinking even surpasses substantially larger medical MLLMs such as HuatuoGPT-Vision-34B by 4.59 points. Moreover, MRPO reduces early-stage reasoning failures from 58.6% to 13.4%, showing that targeted mitigation of cascading failures improves both reasoning quality and final answer accuracy. Our code is available at https://github.com/dmis-lab/MRPO", "categories": [ "cs.CV", "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2606.31825", "tags": [ "cs.CV", "cs.AI" ] }, { "id": "1a545b7dbe1b3eb044bc", "title": "Cryptographically verifiable authorization for autonomous AI agents: a falsifiable hypothesis and proof of concept", "content_text": "Autonomous AI agents increasingly execute actions, invoke tools, and operate on protected resources with limited human oversight. Existing authentication and authorization mechanisms establish identity and delegate authority but do not inherently provide cryptographic evidence that a concrete request issued by a specific agent satisfies the applicable policy in a specific execution context. This study hypothesizes that agent authorization can be formalized as a cryptographically verifiable…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "1a545b7dbe1b3eb044bc", "source": "rss", "type": "change", "key": "oai:arXiv.org:2607.21325v3", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Cryptographically verifiable authorization for autonomous AI agents: a falsifiable hypothesis and proof of concept", "item_type": "entry", "summary": "added: Cryptographically verifiable authorization for autonomous AI agents: a falsifiable hypothesis and proof of concept", "after": { "title": "Cryptographically verifiable authorization for autonomous AI agents: a falsifiable hypothesis and proof of concept", "link": "https://arxiv.org/abs/2607.21325", "id": "oai:arXiv.org:2607.21325v3", "author": "M. Llamb\\'i-Morillas, D. Fern\\'andez-Fern\\'andez", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Autonomous AI agents increasingly execute actions, invoke tools, and operate on protected resources with limited human oversight. Existing authentication and authorization mechanisms establish identity and delegate authority but do not inherently provide cryptographic evidence that a concrete request issued by a specific agent satisfies the applicable policy in a specific execution context. This study hypothesizes that agent authorization can be formalized as a cryptographically verifiable relation, denoted $R_{CVA}$, that jointly binds an agent principal, a concrete authorization request, an execution context, and the satisfaction of an applicable policy, while selectively preserving the confidentiality of private authorization attributes. We introduce a preliminary formal abstraction for Cryptographically Verifiable Agent Authorization (CVA), define a compact set of candidate security properties including authorization soundness, principal binding, request binding, policy binding, and replay resistance, and provide an executable zero-knowledge proof of concept that instantiates selected elements of the model over a Groth16 zk-SNARK construction. We further identify and formalize the structural separation among identity binding, authorization-request binding, and runtime execution binding as a central open problem in the design of secure agentic systems, a distinction to our knowledge, has not been formalized within a cryptographically verifiable authorization relation by current agentic security frameworks, and present a falsifiable research agenda for its resolution.", "categories": [ "cs.CR", "cs.AI" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2607.21325", "tags": [ "cs.CR", "cs.AI" ] }, { "id": "10c79d9390121cc5869b", "title": "Trident : How to Break Deep Reinforcement Learning Cyber Defenses (Agentic)", "content_text": "Autonomous cyber defense systems based on Deep Reinforcement Learning (DRL) have attracted significant research attention, yet remain evaluated almost exclusively against static, heuristic red agents, leaving their robustness against adaptive threats critically understudied. Meanwhile, recent advances in Reinforcement Learning with Verifiable Rewards (RLVR) have improved LLM reasoning, but their integration into cybersecurity remains elusive due to the absence of suitable benchmark environments…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "10c79d9390121cc5869b", "source": "rss", "type": "change", "key": "oai:arXiv.org:2608.04317v2", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Trident : How to Break Deep Reinforcement Learning Cyber Defenses (Agentic)", "item_type": "entry", "summary": "added: Trident : How to Break Deep Reinforcement Learning Cyber Defenses (Agentic)", "after": { "title": "Trident : How to Break Deep Reinforcement Learning Cyber Defenses (Agentic)", "link": "https://arxiv.org/abs/2608.04317", "id": "oai:arXiv.org:2608.04317v2", "author": "Ryozo Masukawa, Ian Bryant, Armita Kazeminajafabadi, Sanggeon Yun, Hyunwoo Oh, SungHeon Jeong, Nathaniel D. Bastian, Mahdi Imani, Mohsen Imani", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "Autonomous cyber defense systems based on Deep Reinforcement Learning (DRL) have attracted significant research attention, yet remain evaluated almost exclusively against static, heuristic red agents, leaving their robustness against adaptive threats critically understudied. Meanwhile, recent advances in Reinforcement Learning with Verifiable Rewards (RLVR) have improved LLM reasoning, but their integration into cybersecurity remains elusive due to the absence of suitable benchmark environments and interaction datasets. To bridge this gap, we introduce Trident, an agentic LLM red teaming framework comprising three components: a dynamic benchmark with isolated sandbox servers spanning CybORG CAGE 4 and CyberWheel, a dataset comprises over 13,000 high-fidelity red-blue interaction trajectories for RLVR, and a ``Code-as-Policy'' RLVR agentic architecture Trident Agentic). The latter reformulates red agent training as a contextual bandit via a tripartite Log Summarizer--Planner--Coder design, where a trainable Planner generates complete attack strategies from compressed execution logs, which a frozen Coder translates into executable Python policies deployed against live DRL defenders. Empirical evaluations reveal a fundamental brittleness in existing defenses: with a single trainable 7B planner, Trident reduces blue agent defensive performance by an average of 522% compared to static red agent baselines while autonomously discovering emergent behaviors such as decoy avoidance and adaptive state prioritization that static heuristics entirely fail to uncover.", "categories": [ "cs.CR", "cs.AI", "cs.LG", "cs.MA" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2608.04317", "tags": [ "cs.CR", "cs.AI", "cs.LG", "cs.MA" ] }, { "id": "8215056f7168574fe2a1", "title": "Hardware Keystores for AI Agent Signing Workflows: A Zero-Trust MCP Enforcement Architecture", "content_text": "AI agents increasingly sign Git commits, certify documents, and attest release artifacts on behalf of their operators, using private keys that live in software-accessible locations (plaintext files, environment variables, container memory) readable by any process the agent can reach. A widely deployed agent framework recently leaked its keys this way to a single email injection. Hardware keystores (HSM, TPM, smart card) keep the key on-device, but exposing the keystore as a tool an LLM agent…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "8215056f7168574fe2a1", "source": "rss", "type": "change", "key": "oai:arXiv.org:2608.06130v2", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Hardware Keystores for AI Agent Signing Workflows: A Zero-Trust MCP Enforcement Architecture", "item_type": "entry", "summary": "added: Hardware Keystores for AI Agent Signing Workflows: A Zero-Trust MCP Enforcement Architecture", "after": { "title": "Hardware Keystores for AI Agent Signing Workflows: A Zero-Trust MCP Enforcement Architecture", "link": "https://arxiv.org/abs/2608.06130", "id": "oai:arXiv.org:2608.06130v2", "author": "Leo Sambrook, Sampo Sovio", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "AI agents increasingly sign Git commits, certify documents, and attest release artifacts on behalf of their operators, using private keys that live in software-accessible locations (plaintext files, environment variables, container memory) readable by any process the agent can reach. A widely deployed agent framework recently leaked its keys this way to a single email injection. Hardware keystores (HSM, TPM, smart card) keep the key on-device, but exposing the keystore as a tool an LLM agent can call moves the problem rather than removing it: once a signing session exists, the hardware cannot tell a request reflecting the operator's intent from one injected into content the agent read. We characterize this confused-deputy problem and build the five-layer Zero-Trust enforcement stack it requires, so that only requests consistent with the operator's committed intent reach the hardware. We evaluate on two attack planes. Prompt injection in content the agent reads (AgentDojo, three injection-following models, n=144) falls from an 18.1% baseline attack success rate to 0% under the full stack. Tool poisoning by a compromised MCP server (MCPTox) is contained identically: a hash comparison protects a pre-committed payload, and human-in-the-loop escalation contains autonomous requests with nothing pre-committed. A further probe delineates how far the semantic filter's protection extends: it detects a substitute document under an unrelated name, but an adversarially plausible substitute name defeats it in every trial we ran. We report this as a central finding: the architecture's guarantee never rests on the filter being right, only on a human being asked whenever nothing was committed in advance. The trade-off we characterize across both planes is that the less an operator can commit to in advance, the less deterministic the resulting guarantee, down to asking a human.", "categories": [ "cs.CR", "cs.AI", "cs.LG" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2608.06130", "tags": [ "cs.CR", "cs.AI", "cs.LG" ] }, { "id": "5abe3c81d06d233ea1d7", "title": "Evolve Vision-Language-Action Model into an Agent with On-the-fly Tool-use", "content_text": "This paper integrates end-to-end Visual-Language-Action (VLA) models with agentic tool-use to propose Agentic Robot with Tool-use (ART). ART is a tool-injection framework that tunes any VLA model to leverage off-the-shelf tool modules for low-level vision, high-level affordance, and embodiment enhancement. Compared to vanilla VLA models with a whole continuous action solution space, ART reduces the complexity of the action solution space through tool-use, which not only improves…", "date_published": "2026-09-25T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "5abe3c81d06d233ea1d7", "source": "rss", "type": "change", "key": "oai:arXiv.org:2608.14047v3", "source_url": "https://rss.arxiv.org/rss/cs.AI", "timestamp": null, "observed_at": "2026-09-25T06:46:03Z", "data": { "change": "added", "label": "Evolve Vision-Language-Action Model into an Agent with On-the-fly Tool-use", "item_type": "entry", "summary": "added: Evolve Vision-Language-Action Model into an Agent with On-the-fly Tool-use", "after": { "title": "Evolve Vision-Language-Action Model into an Agent with On-the-fly Tool-use", "link": "https://arxiv.org/abs/2608.14047", "id": "oai:arXiv.org:2608.14047v3", "author": "Ding Yi, Yanzhao Yu, Xili Dai, Xianbiao Qi, Peiwen Sun, Xueqian Wang, Xiangyu Yue, Jianan Wang", "published_at": "2026-09-25T04:00:00Z", "updated_at": "2026-09-25T04:00:00Z", "summary": "This paper integrates end-to-end Visual-Language-Action (VLA) models with agentic tool-use to propose Agentic Robot with Tool-use (ART). ART is a tool-injection framework that tunes any VLA model to leverage off-the-shelf tool modules for low-level vision, high-level affordance, and embodiment enhancement. Compared to vanilla VLA models with a whole continuous action solution space, ART reduces the complexity of the action solution space through tool-use, which not only improves generalizability across different tasks but also reduces data dependency. To demonstrate the advantages (high generalizability and low data dependency) of this framework, we first built a dataset of 30K tool-use trajectories and action demonstrations, which is much smaller than those used by baseline methods. We then designed a training regimen for long-trajectory tool-use reasoning in challenging environments. Experiments show that ART achieves a 20% higher success rate than mainstream baselines on simulation and real-world tasks, such as pick-and-place in the dark at novel viewpoints. Empirical results highlight the benefits of an agent-based approach: modular tool utilization enables more efficient training, lightweight deployment, and scalable integration of new tools. This design fosters robustness, adaptability, and extensibility, paving the way for the practical deployment of VLA systems in complex real-world scenarios.", "categories": [ "cs.RO", "cs.AI", "cs.CV" ], "feed": { "title": "cs.AI updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.AI", "feed_url": "https://rss.arxiv.org/rss/cs.AI" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.AI", "elapsed_ms": 9, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.2" }, { "step": "grep", "version": "0.3.2", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.2", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.2", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.2", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2608.14047", "tags": [ "cs.RO", "cs.AI", "cs.CV" ] }, { "id": "79e77f904693668c433c", "title": "COMED: The Missing Middle Between Routing and Collaboration in Multi-LLM Inference", "content_text": "No single Large Language Model (LLM) is uniformly reliable across queries, motivating multi-model inference systems that either route among models or combine their outputs. However, routing stops after selecting an initial model, while dense collaboration invokes peers on every query. We show that collaboration is non-monotonic: peers can recover failures that no model solves alone, but can also corrupt initially correct answers. We introduce COMED (Controlled Model Escalation for Multi-LLM…", "date_published": "2026-09-24T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "79e77f904693668c433c", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.26913v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-24T10:12:41Z", "data": { "change": "added", "label": "COMED: The Missing Middle Between Routing and Collaboration in Multi-LLM Inference", "item_type": "entry", "summary": "added: COMED: The Missing Middle Between Routing and Collaboration in Multi-LLM Inference", "after": { "title": "COMED: The Missing Middle Between Routing and Collaboration in Multi-LLM Inference", "link": "https://arxiv.org/abs/2609.26913", "id": "oai:arXiv.org:2609.26913v1", "author": "Norah Alballa, Wenxuan Zhang, Salma Kharrat, Fares Fourati, Zafar Ayyub Qazi, Mohamed Elhoseiny, Marco Canini", "published_at": "2026-09-24T04:00:00Z", "updated_at": "2026-09-24T04:00:00Z", "summary": "No single Large Language Model (LLM) is uniformly reliable across queries, motivating multi-model inference systems that either route among models or combine their outputs. However, routing stops after selecting an initial model, while dense collaboration invokes peers on every query. We show that collaboration is non-monotonic: peers can recover failures that no model solves alone, but can also corrupt initially correct answers. We introduce COMED (Controlled Model Escalation for Multi-LLM Deliberation), a post-anchor controller for selective cross-model collaboration. COMED uses anchor self-consistency, router margin, and a lightweight peer probe to accept confident answers, verify ambiguous cases, and escalate only when collaboration is likely beneficial. We formalize this trade-off with a rescue-harm decomposition showing that selective collaboration improves when rescued errors outweigh collaboration-induced harms. Across medical, scientific, and general reasoning benchmarks, COMED improves fixed and routed anchors in all 16 open-weight settings, with gains up to +10.7 percentage points on MedQA while invoking fewer models and using fewer decoded tokens than dense collaboration. On HLE with frontier models, COMED improves GPT-5.5 from 23.1% to 28.1%, outperforming dense collaboration and achieving the best results.", "categories": [ "cs.CL", "cs.AI" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 158, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.1" }, { "step": "grep", "version": "0.3.1", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.1", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.1", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.1", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.26913", "tags": [ "cs.CL", "cs.AI" ] }, { "id": "f257fd0bee07d7b8ba45", "title": "Experts Rise Where LLMs Disagree: Using Cross-Model Disagreement to Target Expert Effort in LLM Codebook Revision for Large-Scale Annotation", "content_text": "Large-scale text annotation brings expert insight to millions of documents, often through a codebook that AI annotators follow. Developing a robust codebook, however, takes months. Large language models (LLMs) could speed this process by applying an early codebook to the data, surfacing cases with strong LLM disagreement, and eliciting expert feedback to address them. We examined three ways experts can provide feedback for LLM codebook revision: (i) editing LLM-generated revisions driven by…", "date_published": "2026-09-24T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "f257fd0bee07d7b8ba45", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.26926v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-24T10:12:41Z", "data": { "change": "added", "label": "Experts Rise Where LLMs Disagree: Using Cross-Model Disagreement to Target Expert Effort in LLM Codebook Revision for Large-Scale Annotation", "item_type": "entry", "summary": "added: Experts Rise Where LLMs Disagree: Using Cross-Model Disagreement to Target Expert Effort in LLM Codebook Revision for Large-Scale Annotation", "after": { "title": "Experts Rise Where LLMs Disagree: Using Cross-Model Disagreement to Target Expert Effort in LLM Codebook Revision for Large-Scale Annotation", "link": "https://arxiv.org/abs/2609.26926", "id": "oai:arXiv.org:2609.26926v1", "author": "Zeyu He, Zhuqian Zhou, Kirk Vanacore, Rene F. Kizilcec, Ting-Hao 'Kenneth' Huang", "published_at": "2026-09-24T04:00:00Z", "updated_at": "2026-09-24T04:00:00Z", "summary": "Large-scale text annotation brings expert insight to millions of documents, often through a codebook that AI annotators follow. Developing a robust codebook, however, takes months. Large language models (LLMs) could speed this process by applying an early codebook to the data, surfacing cases with strong LLM disagreement, and eliciting expert feedback to address them. We examined three ways experts can provide feedback for LLM codebook revision: (i) editing LLM-generated revisions driven by cross-LLM disagreement (Codebook Verifying), (ii) answering questions about LLM disagreements (Question Answering), and (iii) labeling disagreement cases with rationales (Rationale Labeling). Experiments on thousands of tutoring-session transcripts show that Rationale Labeling yielded the highest LLM-labeling accuracy (64.9%) against expert labels, outperforming the expert-revised codebook (57.8%). The best Question Answering setting also outperformed it (60.5%). Our work shows that LLMs can be used to strategically target expert attention, shortening months of codebook revision to days without sacrificing labeling performance.", "categories": [ "cs.CL", "cs.AI", "cs.HC", "cs.LG" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 158, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.1" }, { "step": "grep", "version": "0.3.1", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.1", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.1", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.1", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.26926", "tags": [ "cs.CL", "cs.AI", "cs.HC", "cs.LG" ] }, { "id": "40c00cb07c881cac3d0f", "title": "LEGO: Synergizing Expert GraphRAG and Expert Chain-of-Thought for Legal Reasoning", "content_text": "Large language models are increasingly applied to high-risk domains such as law, yet complex legal reasoning remains limited by two structural challenges. First, existing RAG and GraphRAG methods emphasize lexical or semantic similarity while overlooking normative relations among legal provisions. Second, vanilla Chain-of-Thought prompting may generate plausible rationales without enforcing the normative structure of legal reasoning. To deal with the bottleneck of pipelines in the legal…", "date_published": "2026-09-24T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "40c00cb07c881cac3d0f", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.27009v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-24T10:12:41Z", "data": { "change": "added", "label": "LEGO: Synergizing Expert GraphRAG and Expert Chain-of-Thought for Legal Reasoning", "item_type": "entry", "summary": "added: LEGO: Synergizing Expert GraphRAG and Expert Chain-of-Thought for Legal Reasoning", "after": { "title": "LEGO: Synergizing Expert GraphRAG and Expert Chain-of-Thought for Legal Reasoning", "link": "https://arxiv.org/abs/2609.27009", "id": "oai:arXiv.org:2609.27009v1", "author": "Qingjing Chen, Junkai Zhang, Shaochun Wang, Jiahao Ding, Siyuan Zheng, Yukun Yan, Zhi Zheng, Antonino Rotolo, Yun Liu, Weixing Shen", "published_at": "2026-09-24T04:00:00Z", "updated_at": "2026-09-24T04:00:00Z", "summary": "Large language models are increasingly applied to high-risk domains such as law, yet complex legal reasoning remains limited by two structural challenges. First, existing RAG and GraphRAG methods emphasize lexical or semantic similarity while overlooking normative relations among legal provisions. Second, vanilla Chain-of-Thought prompting may generate plausible rationales without enforcing the normative structure of legal reasoning. To deal with the bottleneck of pipelines in the legal reasoning domain, we propose LEGO, a dual-module framework that synergizes Legal Expert GraphRAG and expert Chain-of-thought for complex legal reasoning. ExpertGraphRAG uses an expert-annotated civil code graph encoding these normative relations with a greedy normative-coverage retrieval algorithm to dynamically extract instance-specific provision subgraphs, while ExpertCoT organizes the retrieved provisions and case facts into structured Provision-Fact-Conclusion reasoning. With a Qwen3-8B backbone, LEGO achieves 40.53% exact-match accuracy on LawExamQA_Civil, outperforming the evaluated RAG and CoT baselines and performing comparably to the evaluated larger models, while remaining robust on multi-hop questions. It also achieves the best results among the evaluated baselines on the open-ended benchmarks. Ablation studies confirm the individual and complementary contributions of both modules, demonstrating LEGO's effectiveness in improving LLMs' complex legal reasoning ability. Code and dataset can be found in the link: https://github.com/BLK-WHT/LEGO", "categories": [ "cs.CL", "cs.IR" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 158, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.1" }, { "step": "grep", "version": "0.3.1", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.1", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.1", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.1", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.27009", "tags": [ "cs.CL", "cs.IR" ] }, { "id": "1a22fa4abae2f4a52b0a", "title": "What Changes When Fact-Verification Scores Improve? Evidence and Answer Accounting Across Trained Verifiers and LLMs", "content_text": "A joint fact-verification score assesses answers and submitted evidence together. When the score improves, how much of the gain remains if the answers are held fixed? On FEVEROUS, strict score is the percentage of claims with a correct answer and a complete annotated evidence group in the submitted evidence. Across four trained DeBERTa checkpoints and 7,890 claims, replacing DCUF evidence with UnifEE evidence raises strict score by 9.61 percentage points, compared with 1.96 percentage points in…", "date_published": "2026-09-24T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "1a22fa4abae2f4a52b0a", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.27064v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-24T10:12:41Z", "data": { "change": "added", "label": "What Changes When Fact-Verification Scores Improve? Evidence and Answer Accounting Across Trained Verifiers and LLMs", "item_type": "entry", "summary": "added: What Changes When Fact-Verification Scores Improve? Evidence and Answer Accounting Across Trained Verifiers and LLMs", "after": { "title": "What Changes When Fact-Verification Scores Improve? Evidence and Answer Accounting Across Trained Verifiers and LLMs", "link": "https://arxiv.org/abs/2609.27064", "id": "oai:arXiv.org:2609.27064v1", "author": "Han Chen, Yingrui Li", "published_at": "2026-09-24T04:00:00Z", "updated_at": "2026-09-24T04:00:00Z", "summary": "A joint fact-verification score assesses answers and submitted evidence together. When the score improves, how much of the gain remains if the answers are held fixed? On FEVEROUS, strict score is the percentage of claims with a correct answer and a complete annotated evidence group in the submitted evidence. Across four trained DeBERTa checkpoints and 7,890 claims, replacing DCUF evidence with UnifEE evidence raises strict score by 9.61 percentage points, compared with 1.96 percentage points in answer accuracy. The paired 95% interval for the strict-score gain is [8.77, 10.43], conditional on these checkpoints. Replacing only the evidence passed to the scorer accounts for 7.92 or 9.08 percentage points when we retain the answers generated from DCUF or UnifEE evidence, respectively. To examine how this evidence gain depends on evaluation choices, we generate 470,400 responses from two 8B LLMs on FEVER, FEVEROUS, and SciFact under two answer formats and two context budgets. Increasing context from 256 to 2,048 tokens raises the fixed-answer evidence gain on FEVEROUS by 3.84 and 3.10 percentage points for Qwen and Llama, respectively. The effects fall short of the prespecified cross-dataset criterion, while some intervals extend beyond the two-point small-effect bound. Post-hoc analyses quantify changes in answers and submitted evidence, and show when aggregate accuracy and evidence-coverage rates miss the claim-level pattern. The four answer-evidence score combinations reveal changes that endpoint and aggregate metrics leave unresolved.", "categories": [ "cs.CL" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 158, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.1" }, { "step": "grep", "version": "0.3.1", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.1", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.1", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.1", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.27064", "tags": [ "cs.CL" ] }, { "id": "cce0e2e8d17f42007a40", "title": "Giving Credit Where It's Due: Redundancy-Aware Learning for Efficient Reasoning", "content_text": "Large reasoning models can produce correct yet unnecessarily long reasoning traces. Existing methods improve reasoning efficiency with trajectory-level objectives or local token- and step-level signals, but rarely model inter-step semantic dependencies. This limits their ability to distinguish redundant steps from those that support later deductions, making it harder to shorten reasoning without sacrificing accuracy. We introduce RECAP (REdundancy-aware Credit Assignment via Propagation), which…", "date_published": "2026-09-24T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "cce0e2e8d17f42007a40", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.27156v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-24T10:12:41Z", "data": { "change": "added", "label": "Giving Credit Where It's Due: Redundancy-Aware Learning for Efficient Reasoning", "item_type": "entry", "summary": "added: Giving Credit Where It's Due: Redundancy-Aware Learning for Efficient Reasoning", "after": { "title": "Giving Credit Where It's Due: Redundancy-Aware Learning for Efficient Reasoning", "link": "https://arxiv.org/abs/2609.27156", "id": "oai:arXiv.org:2609.27156v1", "author": "Yuqing Zhou, Hong Wang, Manqing Mao, Zhuoer Wang, Samson Koelle, Jie Yuan, Yanjun Lin, James Feng, Nikki Lijing Kuang, Ziwei Zhu, Wei Niu", "published_at": "2026-09-24T04:00:00Z", "updated_at": "2026-09-24T04:00:00Z", "summary": "Large reasoning models can produce correct yet unnecessarily long reasoning traces. Existing methods improve reasoning efficiency with trajectory-level objectives or local token- and step-level signals, but rarely model inter-step semantic dependencies. This limits their ability to distinguish redundant steps from those that support later deductions, making it harder to shorten reasoning without sacrificing accuracy. We introduce RECAP (REdundancy-aware Credit Assignment via Propagation), which addresses this limitation by assigning credit where it is due based on both a step's downstream role in the reasoning structure and its contribution to solving the problem correctly. We define structural responsibility to capture the step's downstream role by measuring how strongly later reasoning depends on it, using credit propagated backward from the final-answer node through an outcome-independent, LLM-annotated semantic dependency graph. However, a step can have high structural responsibility yet steer the reasoning away from the correct solution. RECAP therefore introduces step efficacy to measure answer-directed progress through changes in gold-answer log-likelihood as each step is added. Together, these signals reshape rollout-level GRPO advantages into step-specific updates. RECAP requires neither a separately trained process reward model nor preconstructed concise trajectories. Across two 7B models and four mathematical reasoning benchmarks, RECAP improves the accuracy-efficiency trade-off. On Qwen2.5-Math-7B, it improves pass@1 by 2.0-3.7 percentage points while reducing reasoning tokens by 8%-31% relative to GRPO across all four benchmarks. Analysis suggests these savings reflect fewer reasoning operations and less dead-end reasoning, rather than more compact expression.", "categories": [ "cs.CL", "cs.LG" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 158, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.1" }, { "step": "grep", "version": "0.3.1", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.1", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.1", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.1", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.27156", "tags": [ "cs.CL", "cs.LG" ] }, { "id": "2afb3bf310448167c454", "title": "Count Evidence, Not Sentences: Tempered Evidence Fusion of LLM Judgments for Long-Text Value Measurement", "content_text": "Large language models (LLMs) are increasingly used to measure public value orientations from long social media posts, yet such posts often mix background, quotations, concessions, and only a few stance-bearing sentences. Existing approaches either ask the model to predict a document-level label directly, which can be overconfident, or aggregate sentence-level predictions by majority or soft voting, which treat uncertain and decisive sentences as equally informative. We formulate long-text value…", "date_published": "2026-09-24T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "2afb3bf310448167c454", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.27165v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-24T10:12:41Z", "data": { "change": "added", "label": "Count Evidence, Not Sentences: Tempered Evidence Fusion of LLM Judgments for Long-Text Value Measurement", "item_type": "entry", "summary": "added: Count Evidence, Not Sentences: Tempered Evidence Fusion of LLM Judgments for Long-Text Value Measurement", "after": { "title": "Count Evidence, Not Sentences: Tempered Evidence Fusion of LLM Judgments for Long-Text Value Measurement", "link": "https://arxiv.org/abs/2609.27165", "id": "oai:arXiv.org:2609.27165v1", "author": "Yuhe Wu, Rui Qian, Guangyu Wang, Yuran Chen, Yuanchao Zhu, Junjie Yang, Zhengheng Li, Jiulin Cai, Tianyi Zhang, Zihan Dong, Jiaxin Liu, Yujie Chen, Guang Zhang", "published_at": "2026-09-24T04:00:00Z", "updated_at": "2026-09-24T04:00:00Z", "summary": "Large language models (LLMs) are increasingly used to measure public value orientations from long social media posts, yet such posts often mix background, quotations, concessions, and only a few stance-bearing sentences. Existing approaches either ask the model to predict a document-level label directly, which can be overconfident, or aggregate sentence-level predictions by majority or soft voting, which treat uncertain and decisive sentences as equally informative. We formulate long-text value measurement as a decision-fusion problem and propose Tempered Evidence Fusion (TEF), a training-free rule that weights each sentence's log-odds by its normalized information gain, as derived from a generalized Bayesian posterior. This makes the fused score nearly vanish for uncertain sentences while preserving the Bayes-optimal weight of decisive evidence. We further introduce Multi-event Insight Network Dimensions (MIND), a benchmark of 8,358 Chinese and English posts spanning five years of public events and six value dimensions. On MIND, TEF outperforms the strongest baseline among Direct, Majority Vote, and Soft Vote by an average of 4.5 accuracy points and 4.6 macro-F1 points across five LLMs and two languages. MIND dataset and code are available at https://github.com/Kzczc/ICASSP2027-TEF.", "categories": [ "cs.CL", "cs.AI" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 158, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.1" }, { "step": "grep", "version": "0.3.1", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.1", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.1", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.1", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.27165", "tags": [ "cs.CL", "cs.AI" ] }, { "id": "e9fbcc8d99af9e516a7b", "title": "Realize What Matters: Principled Context Representation for Large-Scale Reasoning", "content_text": "Solving complex tasks in domains such as science, medicine, law, and finance often requires assembling interdependent information scattered across vast, heterogeneous sources far beyond model context limits. Existing approaches tackle this challenge by organizing information into more manageable representations over which models can reason, such as graphs, textual memories, and retrieval collections. These representations dictate what downstream reasoning is possible and, ultimately, whether it…", "date_published": "2026-09-24T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "e9fbcc8d99af9e516a7b", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.27173v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-24T10:12:41Z", "data": { "change": "added", "label": "Realize What Matters: Principled Context Representation for Large-Scale Reasoning", "item_type": "entry", "summary": "added: Realize What Matters: Principled Context Representation for Large-Scale Reasoning", "after": { "title": "Realize What Matters: Principled Context Representation for Large-Scale Reasoning", "link": "https://arxiv.org/abs/2609.27173", "id": "oai:arXiv.org:2609.27173v1", "author": "Michael Theologitis, Dean Light, Shuyue Stella Li, Benjamin Newman, Yulia Tsvetkov, Dan Suciu", "published_at": "2026-09-24T04:00:00Z", "updated_at": "2026-09-24T04:00:00Z", "summary": "Solving complex tasks in domains such as science, medicine, law, and finance often requires assembling interdependent information scattered across vast, heterogeneous sources far beyond model context limits. Existing approaches tackle this challenge by organizing information into more manageable representations over which models can reason, such as graphs, textual memories, and retrieval collections. These representations dictate what downstream reasoning is possible and, ultimately, whether it succeeds; yet their design and construction remain largely ad hoc. In this work, drawing on the cognitive theory of relevance realization, we propose concrete principles for designing AI systems that construct effective representations of very large contexts. We analyze existing approaches and show how their successes and failures map onto their alignment with these principles, and introduce R3Con, a harness designed to operationalize the principles more systematically. We evaluate R3Con against nine state-of-the-art baselines on two recent benchmarks of reasoning over large document corpora. On these benchmarks, R3Con substantially outperforms the strongest baseline, by $20$ and $8.4$ percentage points. It also enables smaller models to outperform much larger ones: R3Con with 4B and 9B models outperforms all evaluated 35B baselines, while R3Con with a 35B-A3B model outperforms Claude Code with Claude-Sonnet-5 at $3.7\\times$ lower cost. Our results show that context representations following our principled approach can reduce reliance on model scale, pointing toward a future of AI systems with frontier-level performance powered by smaller models. Our code is available at https://github.com/michaeltheologitis/r3con", "categories": [ "cs.CL" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 158, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.1" }, { "step": "grep", "version": "0.3.1", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.1", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.1", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.1", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.27173", "tags": [ "cs.CL" ] }, { "id": "b731d97b925504f0a54b", "title": "LOCKR: A Hidden-State Trajectory-Guided Planner for Detecting and Repairing Stable-but-Wrong Lock-In in Diffusion Language Models", "content_text": "Diffusion language models generate text through iterative denoising, exposing intermediate trajectories before final answers are produced. We identify a recurring reasoning failure, stable-but-wrong lock-in, where an answer stabilizes early around an incorrect value while substantial denoising remains. Surface-level decoding signals such as confidence, entropy, margin, and answer stability are insufficient to reliably distinguish correct from erroneous lock-in. We formulate selective reasoning…", "date_published": "2026-09-24T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "b731d97b925504f0a54b", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.27220v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-24T10:12:41Z", "data": { "change": "added", "label": "LOCKR: A Hidden-State Trajectory-Guided Planner for Detecting and Repairing Stable-but-Wrong Lock-In in Diffusion Language Models", "item_type": "entry", "summary": "added: LOCKR: A Hidden-State Trajectory-Guided Planner for Detecting and Repairing Stable-but-Wrong Lock-In in Diffusion Language Models", "after": { "title": "LOCKR: A Hidden-State Trajectory-Guided Planner for Detecting and Repairing Stable-but-Wrong Lock-In in Diffusion Language Models", "link": "https://arxiv.org/abs/2609.27220", "id": "oai:arXiv.org:2609.27220v1", "author": "Guoshenghui Zhao, Tan Yu, Weijie Zhao", "published_at": "2026-09-24T04:00:00Z", "updated_at": "2026-09-24T04:00:00Z", "summary": "Diffusion language models generate text through iterative denoising, exposing intermediate trajectories before final answers are produced. We identify a recurring reasoning failure, stable-but-wrong lock-in, where an answer stabilizes early around an incorrect value while substantial denoising remains. Surface-level decoding signals such as confidence, entropy, margin, and answer stability are insufficient to reliably distinguish correct from erroneous lock-in. We formulate selective reasoning repair as a lightweight test-time planning problem and propose LOCKR, a hidden-state trajectory-guided planner that decides when to allocate additional computation, expands a structured set of targeted repair branches, and selects the most promising continuation using trajectory-aware verification. Across two diffusion language models and three mathematical reasoning benchmarks, hidden-state trajectories consistently outperform surface signals and single hidden snapshots for both wrong-lock-in detection and repair selection. On natural evaluation distributions, LOCKR yields absolute accuracy gains of 2.21--5.37 percentage points across all five evaluated settings, with repair rates ranging from 22% to 41%. These results establish hidden diffusion trajectories as actionable signals for selective test-time reasoning repair.", "categories": [ "cs.CL" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 158, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.1" }, { "step": "grep", "version": "0.3.1", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.1", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.1", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.1", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.27220", "tags": [ "cs.CL" ] }, { "id": "37a4ed1403a1427c698b", "title": "Distilling Sequential Computation in Transformer Language Models", "content_text": "Transformer language models process sequences token by token in an autoregressive manner, making growing contexts increasingly expensive. Yet many adjacent token spans are highly predictable or frequently occur as stable units, suggesting that their representations may be compressible. We introduce a method for distilling sequential computation by replacing spans of input tokens with collapsed representations, computed on the fly by a lightweight merge module. This module generates a single…", "date_published": "2026-09-24T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "37a4ed1403a1427c698b", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.27233v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-24T10:12:41Z", "data": { "change": "added", "label": "Distilling Sequential Computation in Transformer Language Models", "item_type": "entry", "summary": "added: Distilling Sequential Computation in Transformer Language Models", "after": { "title": "Distilling Sequential Computation in Transformer Language Models", "link": "https://arxiv.org/abs/2609.27233", "id": "oai:arXiv.org:2609.27233v1", "author": "Zixuan Lan, Jessica Yang, Yanhong Li, Karen Livescu, Jiawei Zhou", "published_at": "2026-09-24T04:00:00Z", "updated_at": "2026-09-24T04:00:00Z", "summary": "Transformer language models process sequences token by token in an autoregressive manner, making growing contexts increasingly expensive. Yet many adjacent token spans are highly predictable or frequently occur as stable units, suggesting that their representations may be compressible. We introduce a method for distilling sequential computation by replacing spans of input tokens with collapsed representations, computed on the fly by a lightweight merge module. This module generates a single surrogate embedding from a sequence of static token embeddings that captures the functional role of the multiple tokens, allowing pretrained models to operate on compressed inputs without architectural changes or re-training. We apply this approach during inference to compress both prompts and intermediate decoding steps, using a rollback mechanism to substitute stored multi-token KV cache entries with their single-step surrogates. Experiments across diverse models show that the merge module can be used to reduce effective sequence length by up to 40% with minimal accuracy degradation across language modeling evaluations and downstream tasks, including question answering, summarization, commonsense reasoning, and long-form mathematical reasoning. Additional lightweight adaptation of the merge module further improves the accuracy-compression trade-off in selected settings. These results demonstrate that sequential token computation in Transformers can be effectively approximated through condensed surrogate representations that approximate the original behavior without model updating.", "categories": [ "cs.CL" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 158, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.1" }, { "step": "grep", "version": "0.3.1", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.1", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.1", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.1", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.27233", "tags": [ "cs.CL" ] }, { "id": "37fc67abe1b8419a7fe0", "title": "UniDataAgent: An Ontology-Grounded Agent for Enterprise Question-to-Report Automation", "content_text": "Enterprise data agents must preserve organization specific semantics, not just translate questions into queries. We present ChinaUnicom DataAgent (UniDataAgent), an ontology grounded system for reusable question-to-report analysis that separates semantic acquisition from online execution. Ontology Acquisition and Validation stage (OAV) builds versioned enterprise ontologies from metadata, business knowledge, and supporting materials through expert authored business skills, constrained…", "date_published": "2026-09-24T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "37fc67abe1b8419a7fe0", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.27257v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-24T10:12:41Z", "data": { "change": "added", "label": "UniDataAgent: An Ontology-Grounded Agent for Enterprise Question-to-Report Automation", "item_type": "entry", "summary": "added: UniDataAgent: An Ontology-Grounded Agent for Enterprise Question-to-Report Automation", "after": { "title": "UniDataAgent: An Ontology-Grounded Agent for Enterprise Question-to-Report Automation", "link": "https://arxiv.org/abs/2609.27257", "id": "oai:arXiv.org:2609.27257v1", "author": "Yutai Duan, Yahui Zhao, Zhangti Li, Yu Ma, Zhenfeng Qi, Shaoyang Yuan, Jing Fan, Jie Liu", "published_at": "2026-09-24T04:00:00Z", "updated_at": "2026-09-24T04:00:00Z", "summary": "Enterprise data agents must preserve organization specific semantics, not just translate questions into queries. We present ChinaUnicom DataAgent (UniDataAgent), an ontology grounded system for reusable question-to-report analysis that separates semantic acquisition from online execution. Ontology Acquisition and Validation stage (OAV) builds versioned enterprise ontologies from metadata, business knowledge, and supporting materials through expert authored business skills, constrained generation, question verification, and selected expert review. Question-to-Report Execution (QRE) stage retrieves semantic contracts for each question, coordinates skills and data tools, validates results, and produces evidence linked reports. Across 27 enterprise tables and roughly thousands of metric types, ontology construction took a few hours instead of about one week manually. It took just a few minutes to generate the reports, instead of several working days. Ontology grounding achieved 95.0\\% strict accuracy on real business questions, versus 72.5\\% for document RAG, especially on structured and compositional tasks. The system has already been deployed to generate cost savings and has the potential to be replicated in other enterprises.", "categories": [ "cs.CL" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 158, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.1" }, { "step": "grep", "version": "0.3.1", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.1", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.1", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.1", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.27257", "tags": [ "cs.CL" ] }, { "id": "fb80383541fde56378a7", "title": "Can One Adapted Model Do It All? Fine-Tuning Strategy Selection for Customer Support LLMs", "content_text": "Production customer-support systems often require LLMs to support multiple skills, such as intent classification, question answering, summarization, or tool-use decisions. A central deployment question is whether these skills should be handled by separate task-specialist models or by a single model trained through multi-task training, sequential updates, or model merging. We study this question using thirteen models spanning five families (Qwen3, Qwen3.5, Gemma-3, Llama-3.1, and Mistral) from…", "date_published": "2026-09-24T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "fb80383541fde56378a7", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.27262v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-24T10:12:41Z", "data": { "change": "added", "label": "Can One Adapted Model Do It All? Fine-Tuning Strategy Selection for Customer Support LLMs", "item_type": "entry", "summary": "added: Can One Adapted Model Do It All? Fine-Tuning Strategy Selection for Customer Support LLMs", "after": { "title": "Can One Adapted Model Do It All? Fine-Tuning Strategy Selection for Customer Support LLMs", "link": "https://arxiv.org/abs/2609.27262", "id": "oai:arXiv.org:2609.27262v1", "author": "Md Tahmid Rahman Laskar, Xue-Yong Fu, Shashi Bhushan TN", "published_at": "2026-09-24T04:00:00Z", "updated_at": "2026-09-24T04:00:00Z", "summary": "Production customer-support systems often require LLMs to support multiple skills, such as intent classification, question answering, summarization, or tool-use decisions. A central deployment question is whether these skills should be handled by separate task-specialist models or by a single model trained through multi-task training, sequential updates, or model merging. We study this question using thirteen models spanning five families (Qwen3, Qwen3.5, Gemma-3, Llama-3.1, and Mistral) from 0.6B to 32B parameters across eight customer-support datasets, spanning four public and four proprietary datasets with approximately 74.5k training and 8.7k evaluation samples. Under a fixed training protocol, we train more than 200 checkpoints. Our experiments reveal that multi-task full fine-tuning is the strongest operational default at every model size we test. Specialist models are strong on their target tasks but often degrade sharply off-task, making reliable routing important. Sequential Low-Rank Adaptation (LoRA) preserves earlier skills better than sequential full fine-tuning, while merging a specialist with its base model improves off-task robustness with limited same-task loss for larger models. We conclude with practical guidelines for selecting fine-tuning strategies in real-world settings.", "categories": [ "cs.CL" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 158, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.1" }, { "step": "grep", "version": "0.3.1", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.1", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.1", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.1", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.27262", "tags": [ "cs.CL" ] }, { "id": "bfd9dbd34226af37cde5", "title": "Guides That Cause Actions: An Offline Study of Guide-Action Mutual Reinforcement in Multimodal Web Agents", "content_text": "Web agents are usually evaluated in live environments, where environment state and judge models drift between runs, so the same checkpoint rarely reproduces the same score, making controlled studies of training phenomena impractical. We present WebMRE, an offline benchmark of 541 tasks and 5,293 steps derived from successful WebArena trajectories, with fully audited test labels and a deterministic protocol that scores a checkpoint identically on every run without any environment. Each step…", "date_published": "2026-09-24T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "bfd9dbd34226af37cde5", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.27353v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-24T10:12:41Z", "data": { "change": "added", "label": "Guides That Cause Actions: An Offline Study of Guide-Action Mutual Reinforcement in Multimodal Web Agents", "item_type": "entry", "summary": "added: Guides That Cause Actions: An Offline Study of Guide-Action Mutual Reinforcement in Multimodal Web Agents", "after": { "title": "Guides That Cause Actions: An Offline Study of Guide-Action Mutual Reinforcement in Multimodal Web Agents", "link": "https://arxiv.org/abs/2609.27353", "id": "oai:arXiv.org:2609.27353v1", "author": "Chengguang Gan, Yunhao Liang, QingHao Zhang, Shiwen Ni", "published_at": "2026-09-24T04:00:00Z", "updated_at": "2026-09-24T04:00:00Z", "summary": "Web agents are usually evaluated in live environments, where environment state and judge models drift between runs, so the same checkpoint rarely reproduces the same score, making controlled studies of training phenomena impractical. We present WebMRE, an offline benchmark of 541 tasks and 5,293 steps derived from successful WebArena trajectories, with fully audited test labels and a deterministic protocol that scores a checkpoint identically on every run without any environment. Each step pairs a human oriented guide sentence with a grounded action, enabling the first study of the mutual reinforcement effect between them in web agents. Averaged over three seeds the effect holds for both models in both decoding orders and grows with scale: jointly decoding a guide lifts element selection over an action only reference by 0.9 and 0.2 points for Qwen3.5-4B and by 1.7 and 2.2 points for Qwen3.5-9B. A mediation analysis shows that the guide is a causal channel rather than commentary: forcing the gold guide as a decoding prefix lifts action accuracy from .422 to .684, another step's guide collapses it to .055, and a paraphrase that renames the target still recovers half of the gain, so the channel carries instruction meaning and not only the label string. The same channel yields an offline reward that only a replayable protocol makes computable, though optimizing it from a strong checkpoint brings no gain yet. Our fine tuned models outperform GPT-5.5, Claude Opus 4.8, and Gemini 3.5 Flash, run zero shot, on every offline metric.", "categories": [ "cs.CL" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 158, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.1" }, { "step": "grep", "version": "0.3.1", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.1", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.1", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.1", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.27353", "tags": [ "cs.CL" ] }, { "id": "c47f1e6ea6803835192d", "title": "Automated Extraction of Records of Processing Activities (RoPA) Using Hybrid RAG and Locally Deployed Large Language Models", "content_text": "Vietnam's Personal Data Protection Law (Law No. 91/2025/QH15) and Decree No. 356/2025/ND-CP, effective January 1, 2026, require organizations to establish and maintain Records of Processing Activities (RoPA). Manual RoPA preparation is labor-intensive, while cloud-hosted large language models (LLMs) may conflict with data-sovereignty requirements. We propose RoPA Manager, a system for automated RoPA information extraction using hybrid retrieval that combines lexical ranking over tsvector…", "date_published": "2026-09-24T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "c47f1e6ea6803835192d", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.27359v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-24T10:12:41Z", "data": { "change": "added", "label": "Automated Extraction of Records of Processing Activities (RoPA) Using Hybrid RAG and Locally Deployed Large Language Models", "item_type": "entry", "summary": "added: Automated Extraction of Records of Processing Activities (RoPA) Using Hybrid RAG and Locally Deployed Large Language Models", "after": { "title": "Automated Extraction of Records of Processing Activities (RoPA) Using Hybrid RAG and Locally Deployed Large Language Models", "link": "https://arxiv.org/abs/2609.27359", "id": "oai:arXiv.org:2609.27359v1", "author": "To Duy Hinh, Nguyen Le Quoc Anh, Phan Van Tri, Khuong Nguyen-An", "published_at": "2026-09-24T04:00:00Z", "updated_at": "2026-09-24T04:00:00Z", "summary": "Vietnam's Personal Data Protection Law (Law No. 91/2025/QH15) and Decree No. 356/2025/ND-CP, effective January 1, 2026, require organizations to establish and maintain Records of Processing Activities (RoPA). Manual RoPA preparation is labor-intensive, while cloud-hosted large language models (LLMs) may conflict with data-sovereignty requirements. We propose RoPA Manager, a system for automated RoPA information extraction using hybrid retrieval that combines lexical ranking over tsvector, dense-vector search, Reciprocal Rank Fusion (RRF), and locally deployed LLMs. We introduce a Vietnamese RoPA benchmark with 32 organizations, 77 processing activities, 12 field groups, and 4,338 reference values. Evaluation is reported at three distinct levels. The automated scorer, tested on perturbed data without invoking an LLM, achieved F1 = 0.9493 [0.9436, 0.9548]; this measures scorer robustness rather than end-to-end extraction accuracy. End-to-end extraction achieved token coverage of 50.04-55.25% against the reference labels. Two independent experts reviewed 1,558 reference values (35.9% of the benchmark), found no incorrect values, and achieved 99.68% agreement with PABAK = 0.9936. Value-level precision was not measured. Across 32 paired scenarios on a 24 GB GPU, locally deployed Qwen3.5-27B-GPTQ-Int4 showed no statistically significant difference from cloud-based DeepSeek-V4-Flash (difference 0.20 percentage points in favor of DeepSeek, 95% CI [-0.93, 1.32], p = 0.72), while Gemma-4-31B performed significantly worse (p < 0.01).", "categories": [ "cs.CL", "cs.CR", "cs.IR" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 158, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.1" }, { "step": "grep", "version": "0.3.1", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.1", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.1", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.1", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.27359", "tags": [ "cs.CL", "cs.CR", "cs.IR" ] }, { "id": "4b93474e26cfa735addb", "title": "Attention Routing Stabilizes Early: Working-Set Inference for Recurrent Language Models", "content_text": "Recurrent language models repeatedly apply shared network blocks to refine latent representations, but standard inference recomputes global attention at every recurrent step. We study attention dynamics across recurrent depth and find that attention support and distributions stabilize substantially earlier than hidden states and attention outputs. This suggests a two-stage structure: early steps discover a sparse working set of relevant context, while later steps refine representations over…", "date_published": "2026-09-24T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "4b93474e26cfa735addb", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.27373v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-24T10:12:41Z", "data": { "change": "added", "label": "Attention Routing Stabilizes Early: Working-Set Inference for Recurrent Language Models", "item_type": "entry", "summary": "added: Attention Routing Stabilizes Early: Working-Set Inference for Recurrent Language Models", "after": { "title": "Attention Routing Stabilizes Early: Working-Set Inference for Recurrent Language Models", "link": "https://arxiv.org/abs/2609.27373", "id": "oai:arXiv.org:2609.27373v1", "author": "Ke Wan, Chen Chen", "published_at": "2026-09-24T04:00:00Z", "updated_at": "2026-09-24T04:00:00Z", "summary": "Recurrent language models repeatedly apply shared network blocks to refine latent representations, but standard inference recomputes global attention at every recurrent step. We study attention dynamics across recurrent depth and find that attention support and distributions stabilize substantially earlier than hidden states and attention outputs. This suggests a two-stage structure: early steps discover a sparse working set of relevant context, while later steps refine representations over largely the same routing support. Motivated by this structure, we introduce WISE (Working-set Inference with Support Exploitation), a training-free method that uses unrestricted global attention during early recurrence and later reuses directly discovered block-structured support while keeping recurrent depth and within-support attention computation dynamic. Controlled interventions show that recurrent discovery is important and that support-only reuse better preserves model behavior than more restrictive attention-reuse alternatives. Across multi-hop QA benchmarks, WISE largely preserves full-attention performance, while context scaling reveals increasingly sparse working sets and greater efficiency gains. Quality is largely preserved through 2K context, with a measurable loss at 4K. An optimized sparse-attention implementation achieves up to a 1.76x attention speedup over native FlashAttention at 4K and a 1.36x speedup for the full 32-step attention trajectory. Our code is available at https://github.com/tbn5pj/WISE_code.", "categories": [ "cs.CL", "cs.LG" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 158, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.1" }, { "step": "grep", "version": "0.3.1", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.1", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.1", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.1", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.27373", "tags": [ "cs.CL", "cs.LG" ] }, { "id": "e4b9ab831288ba598eb8", "title": "Planned Test-Time Scaling with Coordinated Reasoning Paths", "content_text": "Test-time scaling with parallel branches is widely adopted to improve performance on challenging reasoning tasks. The predominant approach, repeated sampling, draws branches independently from a single policy, which can produce redundant attempts and thereby limit the gains from additional inference compute. To address this limitation, we propose Planned Test-Time Scaling (PTTS), which replaces independent sampling with a coordinated joint policy: a planner generates a solution outline for each…", "date_published": "2026-09-24T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "e4b9ab831288ba598eb8", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.27374v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-24T10:12:41Z", "data": { "change": "added", "label": "Planned Test-Time Scaling with Coordinated Reasoning Paths", "item_type": "entry", "summary": "added: Planned Test-Time Scaling with Coordinated Reasoning Paths", "after": { "title": "Planned Test-Time Scaling with Coordinated Reasoning Paths", "link": "https://arxiv.org/abs/2609.27374", "id": "oai:arXiv.org:2609.27374v1", "author": "Xueqing Wu, Langxing Bai, Hritik Bansal, Po-Nien Kung, Shuo Li, Hao Liu, Nanyun Peng, Kai-Wei Chang", "published_at": "2026-09-24T04:00:00Z", "updated_at": "2026-09-24T04:00:00Z", "summary": "Test-time scaling with parallel branches is widely adopted to improve performance on challenging reasoning tasks. The predominant approach, repeated sampling, draws branches independently from a single policy, which can produce redundant attempts and thereby limit the gains from additional inference compute. To address this limitation, we propose Planned Test-Time Scaling (PTTS), which replaces independent sampling with a coordinated joint policy: a planner generates a solution outline for each branch, steering the branches toward distinct reasoning paths, and an executor produces a full solution conditioned on each outline. Formally, we show that PTTS strictly generalizes repeated sampling and, in a stylized setting, provably promotes coverage of complementary reasoning modes and yields better pass@k scaling. We instantiate PTTS on top of strong reasoning models, keeping them fixed as executors while replacing repeated sampling with PTTS inference to further enhance test-time scaling. Concretely, we develop two variants: PTTS-ZS prompts a model to jointly generate outlines for all branches in a single autoregressive pass, while PTTS-RL directly optimizes the planner against the pass@k reward using truncated execution rollouts for efficient training and a sharper reward signal. Across five mathematical reasoning benchmarks with Qwen3-1.7B and 4B, PTTS-ZS improves pass@64 over repeated sampling by up to 6.7 points, while PTTS-RL further increases the gain to up to 13.4 points. Further analysis indicates that broader coverage of distinct reasoning paths contributes to these gains. Overall, PTTS provides a general framework for improving test-time scaling by coordinating reasoning branches, with zero-shot and trainable instantiations that yield substantial performance gains.", "categories": [ "cs.CL", "cs.AI" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 158, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.1" }, { "step": "grep", "version": "0.3.1", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.1", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.1", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.1", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.27374", "tags": [ "cs.CL", "cs.AI" ] }, { "id": "7e8ede169caf11aa194e", "title": "PRISM-VLM: A Multi-Axis Discriminative Benchmark for Compact Vision-Language Models", "content_text": "Compact vision-language models (VLMs) now power a growing share of multimodal applications. The benchmarks used to compare them, however, inherit a frontier-centric design: each model is reduced to a single accuracy number, narrowing the inter-model gap on saturated suites and pressing models into low-score bands on harder ones. We introduce PRISM-VLM, a multi-axis discriminative benchmark that scores every item along seven axes covering the recurring failure modes (task quality, behavioral…", "date_published": "2026-09-24T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "7e8ede169caf11aa194e", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.27395v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-24T10:12:41Z", "data": { "change": "added", "label": "PRISM-VLM: A Multi-Axis Discriminative Benchmark for Compact Vision-Language Models", "item_type": "entry", "summary": "added: PRISM-VLM: A Multi-Axis Discriminative Benchmark for Compact Vision-Language Models", "after": { "title": "PRISM-VLM: A Multi-Axis Discriminative Benchmark for Compact Vision-Language Models", "link": "https://arxiv.org/abs/2609.27395", "id": "oai:arXiv.org:2609.27395v1", "author": "Sanghee Park, Kee-Eung Kim", "published_at": "2026-09-24T04:00:00Z", "updated_at": "2026-09-24T04:00:00Z", "summary": "Compact vision-language models (VLMs) now power a growing share of multimodal applications. The benchmarks used to compare them, however, inherit a frontier-centric design: each model is reduced to a single accuracy number, narrowing the inter-model gap on saturated suites and pressing models into low-score bands on harder ones. We introduce PRISM-VLM, a multi-axis discriminative benchmark that scores every item along seven axes covering the recurring failure modes (task quality, behavioral robustness, and capability bottlenecks) and combines them into a single PScore, with items recycled from fifteen public benchmarks. Across compact VLMs from the past two years, PScore separates model pairs more reliably than prior single-axis benchmarks under an item-level paired bootstrap, and surfaces behavioral differences these benchmarks average away. Even models with statistically indistinguishable PScores diverge sharply along the per-axis profile, particularly on sycophancy, which is nearly orthogonal to single-prompt accuracy. We will release the full pipeline, prompts, and per-item annotations.", "categories": [ "cs.CL" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 158, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.1" }, { "step": "grep", "version": "0.3.1", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.1", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.1", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.1", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.27395", "tags": [ "cs.CL" ] }, { "id": "f9a5b277e611684f02c5", "title": "Uncheatable Eval: Dynamic Compression-Based Evaluation of Language Models", "content_text": "Modern large language models are pretrained on massive datasets, making it difficult to prevent benchmark data from entering their training sets and undermining the reliability of evaluation results. Reliable evaluation is particularly challenging for base models, whose limited instruction-following ability complicates task-based assessment. We introduce Uncheatable Eval, a dynamic benchmark that regularly collects newly published text to evaluate base language models and reduce the risk of…", "date_published": "2026-09-24T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "f9a5b277e611684f02c5", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.27510v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-24T10:12:41Z", "data": { "change": "added", "label": "Uncheatable Eval: Dynamic Compression-Based Evaluation of Language Models", "item_type": "entry", "summary": "added: Uncheatable Eval: Dynamic Compression-Based Evaluation of Language Models", "after": { "title": "Uncheatable Eval: Dynamic Compression-Based Evaluation of Language Models", "link": "https://arxiv.org/abs/2609.27510", "id": "oai:arXiv.org:2609.27510v1", "author": "Kaifeng Tan, Yudong Li, Linlin Shen", "published_at": "2026-09-24T04:00:00Z", "updated_at": "2026-09-24T04:00:00Z", "summary": "Modern large language models are pretrained on massive datasets, making it difficult to prevent benchmark data from entering their training sets and undermining the reliability of evaluation results. Reliable evaluation is particularly challenging for base models, whose limited instruction-following ability complicates task-based assessment. We introduce Uncheatable Eval, a dynamic benchmark that regularly collects newly published text to evaluate base language models and reduce the risk of data contamination. Drawing on the relationship between a model's predictive ability and its ability to compress data losslessly, we use compression rate to evaluate how well models predict new text. We evaluate 80 models across 14 text categories, study how compression changes with context length, and examine the correlation between compression rate and zero-shot MMLU accuracy. Our results yield three main findings: (1) compression performance follows a consistent scaling trend with model size; (2) attention-based, hybrid, and recurrent models differ in how their compression performance changes as more context becomes available; and (3) lower compression rates are strongly associated with higher zero-shot MMLU accuracy. Code is available at https://github.com/Jellyfish042/uncheatable_eval.", "categories": [ "cs.CL", "cs.AI" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 158, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.1" }, { "step": "grep", "version": "0.3.1", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.1", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.1", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.1", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.27510", "tags": [ "cs.CL", "cs.AI" ] }, { "id": "731851537c3a416a5d29", "title": "When Context Misleads: In-context Learning with Jurisdiction in Large Language Models", "content_text": "In-Context Learning (ICL) has become a cornerstone of modern LLM deployment. However, existing ICL post-training methods have a critical blind spot: they excel at extracting patterns from demonstrations while often neglecting context authority, the ability to determine whether contextual information should govern the final answer. To benchmark this capability, we introduce FakeContextBench, which contains pseudoscientific claims across seven domains. Our evaluation of commercial and open-source…", "date_published": "2026-09-24T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "731851537c3a416a5d29", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.27603v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-24T10:12:41Z", "data": { "change": "added", "label": "When Context Misleads: In-context Learning with Jurisdiction in Large Language Models", "item_type": "entry", "summary": "added: When Context Misleads: In-context Learning with Jurisdiction in Large Language Models", "after": { "title": "When Context Misleads: In-context Learning with Jurisdiction in Large Language Models", "link": "https://arxiv.org/abs/2609.27603", "id": "oai:arXiv.org:2609.27603v1", "author": "Pei-lin Li, Qingle Liu, Junyang Feng, Siyu Li, Sunqi Fan, Xin-Sheng Chen, Shuojin Yang", "published_at": "2026-09-24T04:00:00Z", "updated_at": "2026-09-24T04:00:00Z", "summary": "In-Context Learning (ICL) has become a cornerstone of modern LLM deployment. However, existing ICL post-training methods have a critical blind spot: they excel at extracting patterns from demonstrations while often neglecting context authority, the ability to determine whether contextual information should govern the final answer. To benchmark this capability, we introduce FakeContextBench, which contains pseudoscientific claims across seven domains. Our evaluation of commercial and open-source models shows that large-scale pre-training alone is insufficient for reliable context-authority discrimination. Moreover, prevalent ICL fine-tuning methods can increase susceptibility to misleading context, reducing reality accuracy by up to 14.95 percentage points relative to the base model. To address this trade-off, we propose Jurisdiction In-Context Learning (J-ICL), a post-training framework that incorporates context validation into the training objective. Across four model backbones, J-ICL improves ICLEval by an average of 5.84 percentage points and reality accuracy by 9.20 points over the corresponding base models. It also raises the Reality Rate by an average of 18.09 points relative to MetaICL and Symbol Tuning. These results demonstrate that ICL capability and resistance to deceptive context can be improved together. The benchmark is available at https://github.com/peilin717/FakeContext-Bench.", "categories": [ "cs.CL", "cs.AI" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 158, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.1" }, { "step": "grep", "version": "0.3.1", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.1", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.1", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.1", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.27603", "tags": [ "cs.CL", "cs.AI" ] }, { "id": "1d6073cae65a8043b7d8", "title": "The Path Matters: Evaluating Small Language Models Beyond Answer Accuracy in KGQA", "content_text": "Small language models (SLMs) are increasingly paired with knowledge graphs (KGs), yet end-to-end KG question answering conflates graph access, search, navigation, reasoning, and answer generation. This coupling makes it difficult both to determine whether an SLM can faithfully execute the reasoning path implied by a question and to attribute failures to navigation rather than to other stages of the pipeline. We isolate this capability by employing the THESEUS navigation and traceability…", "date_published": "2026-09-24T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "1d6073cae65a8043b7d8", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.27669v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-24T10:12:41Z", "data": { "change": "added", "label": "The Path Matters: Evaluating Small Language Models Beyond Answer Accuracy in KGQA", "item_type": "entry", "summary": "added: The Path Matters: Evaluating Small Language Models Beyond Answer Accuracy in KGQA", "after": { "title": "The Path Matters: Evaluating Small Language Models Beyond Answer Accuracy in KGQA", "link": "https://arxiv.org/abs/2609.27669", "id": "oai:arXiv.org:2609.27669v1", "author": "Eduin E. Hernandez, Sergio A. Diaz, Luis F. Garcia, Nurassyl Askar, Stefano Rini", "published_at": "2026-09-24T04:00:00Z", "updated_at": "2026-09-24T04:00:00Z", "summary": "Small language models (SLMs) are increasingly paired with knowledge graphs (KGs), yet end-to-end KG question answering conflates graph access, search, navigation, reasoning, and answer generation. This coupling makes it difficult both to determine whether an SLM can faithfully execute the reasoning path implied by a question and to attribute failures to navigation rather than to other stages of the pipeline. We isolate this capability by employing the THESEUS navigation and traceability framework and using frozen, off-the-shelf SLMs as local action policies. At each hop, the environment exposes the legal outgoing graph actions, and the model selects one executable graph action and decides whether to stop, without task-specific parameter updates, model-controlled beam search, or free-form answer generation. This controlled setting allows us to evaluate terminal-answer accuracy with Hits@1 together with path fidelity, using Path Edit Distance (PED) as the primary trajectory metric. Across the Kinship and MQuAKE-ST KGQAs, similarly sized local models differ substantially in answer accuracy and path fidelity, with the two metrics sometimes favoring different models. This model-dependent behavior also extends to prompting, as a single demonstrated trajectory can improve or degrade navigation depending on the model. These results motivate evaluating SLM graph reasoning beyond endpoint accuracy alone.", "categories": [ "cs.CL", "cs.AI" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 158, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.1" }, { "step": "grep", "version": "0.3.1", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.1", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.1", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.1", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.27669", "tags": [ "cs.CL", "cs.AI" ] }, { "id": "2c0fc6544058db3cbd07", "title": "Same Scores, Different Decisions: Evaluating JEV and Language Models for Legal Document Understanding", "content_text": "Contract inference requires multiple judgments about a shared document, but aggregate accuracy can conceal changes in the individual decisions. Repeated agreement is also insufficient: a model may consistently return the wrong answer. In this paper, we compare Jev with nine language models on ContractNLI, evaluating inference cost, response time, average correctness, and correctness across repeated request conditions. Controlled comparisons vary hypothesis visibility, requested outputs, and…", "date_published": "2026-09-24T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "2c0fc6544058db3cbd07", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.27678v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-24T10:12:41Z", "data": { "change": "added", "label": "Same Scores, Different Decisions: Evaluating JEV and Language Models for Legal Document Understanding", "item_type": "entry", "summary": "added: Same Scores, Different Decisions: Evaluating JEV and Language Models for Legal Document Understanding", "after": { "title": "Same Scores, Different Decisions: Evaluating JEV and Language Models for Legal Document Understanding", "link": "https://arxiv.org/abs/2609.27678", "id": "oai:arXiv.org:2609.27678v1", "author": "Fan Zhang, Yankai Chen, Zhuohan Xie, Yixi Zhou, Sijia Peng, Lei Fan, Xinhua Ji, Cunyuan Zheng, Huangyong Shan, Philip S. Yu, Xue Liu, Yu Chen, Preslav Nakov, Songwei He", "published_at": "2026-09-24T04:00:00Z", "updated_at": "2026-09-24T04:00:00Z", "summary": "Contract inference requires multiple judgments about a shared document, but aggregate accuracy can conceal changes in the individual decisions. Repeated agreement is also insufficient: a model may consistently return the wrong answer. In this paper, we compare Jev with nine language models on ContractNLI, evaluating inference cost, response time, average correctness, and correctness across repeated request conditions. Controlled comparisons vary hypothesis visibility, requested outputs, and output order while keeping the contract and target judgment fixed. Jev has the lowest cost and median response time among the evaluated configurations, while hosted language models achieve higher baseline accuracy. Rankings by baseline accuracy differ from rankings by correctness across every condition and repeat, although small differences in the latter do not establish a general stability advantage. Development diagnostics further reveal compensating corrections and regressions, as well as persistent errors. These findings motivate evaluating cost and response time alongside whether individual judgments remain correct as the request configuration changes. Code: https://github.com/ZF-Utokyo/Jev-Benchmark", "categories": [ "cs.CL" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 158, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.1" }, { "step": "grep", "version": "0.3.1", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.1", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.1", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.1", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.27678", "tags": [ "cs.CL" ] }, { "id": "16db4f2179e4a0b623dd", "title": "SkillGym: Internalizing Human Skills into LLMs for Real-World Problem Solving", "content_text": "Human-written agent skills encode rich workflows for real-world problem solving, but are typically used as external inference-time instructions rather than internalized as reusable model capabilities. We introduce \\texttt{SkillGym}, a framework that transforms these skills into executable, verifiable training environments for large language model agents. Its skill-to-task pipeline instantiates concrete tasks, verifies outcomes with code-based checkers, and assesses empirical skill dependence…", "date_published": "2026-09-24T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "16db4f2179e4a0b623dd", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.27717v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-24T10:12:41Z", "data": { "change": "added", "label": "SkillGym: Internalizing Human Skills into LLMs for Real-World Problem Solving", "item_type": "entry", "summary": "added: SkillGym: Internalizing Human Skills into LLMs for Real-World Problem Solving", "after": { "title": "SkillGym: Internalizing Human Skills into LLMs for Real-World Problem Solving", "link": "https://arxiv.org/abs/2609.27717", "id": "oai:arXiv.org:2609.27717v1", "author": "Zhilong Ge, Yuting Shao, Yutao Yang, Yuxuan Cai, Jie Zhou, Kai Chen, Bo Zhang, Qin Chen, Liang He", "published_at": "2026-09-24T04:00:00Z", "updated_at": "2026-09-24T04:00:00Z", "summary": "Human-written agent skills encode rich workflows for real-world problem solving, but are typically used as external inference-time instructions rather than internalized as reusable model capabilities. We introduce \\texttt{SkillGym}, a framework that transforms these skills into executable, verifiable training environments for large language model agents. Its skill-to-task pipeline instantiates concrete tasks, verifies outcomes with code-based checkers, and assesses empirical skill dependence through contrastive executions. We construct and release 2,756 environments across 12 categories and collect 8,364 successful trajectories from multiple models and harnesses, averaging 49 tool calls and over 60k logged text tokens. These resources support supervised fine-tuning on verified workflows and reinforcement learning with outcome-based rewards. Under Claude Code, supervised fine-tuning improves Qwen3.5-35B-A3B by 199 Elo on GDPval-AA v2, 19.10 percentage points on Terminal-Bench 2.1, and 28.13 and 12.38 points on SkillsBench v1.1 with and without skills, respectively. Our 35B \\texttt{SkillGym-Agent} reaches 51.47\\% on skill-assisted SkillsBench, exceeding reported scores for Claude Sonnet 4.6, GPT-5.4 Mini, and DeepSeek V4 Pro. Without skills, it also surpasses skill-assisted bases under Codex and Claude Code, suggesting reusable procedural competence.", "categories": [ "cs.CL" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 158, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.1" }, { "step": "grep", "version": "0.3.1", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.1", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.1", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.1", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.27717", "tags": [ "cs.CL" ] }, { "id": "80024a64d9883b3ba97f", "title": "Improving LLM-based Autonomous Web Agents with Filtering", "content_text": "Autonomous web agents, powered by Large Language Models (LLMs), have garnered significant attention for automating various web-based tasks with multi-step reasoning and decision-making capabilities. An open research question in the development of these agents lies in the format of the webpage input. Raw HTML source code, with its extensive and often irrelevant details, poses difficulties for LLMs with limited context windows. To address this challenge, we first reproduce baseline models such as…", "date_published": "2026-09-24T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "80024a64d9883b3ba97f", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.27770v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-24T10:12:41Z", "data": { "change": "added", "label": "Improving LLM-based Autonomous Web Agents with Filtering", "item_type": "entry", "summary": "added: Improving LLM-based Autonomous Web Agents with Filtering", "after": { "title": "Improving LLM-based Autonomous Web Agents with Filtering", "link": "https://arxiv.org/abs/2609.27770", "id": "oai:arXiv.org:2609.27770v1", "author": "Zhitong Guo, Jing Yu Koh, Ruiyu Li", "published_at": "2026-09-24T04:00:00Z", "updated_at": "2026-09-24T04:00:00Z", "summary": "Autonomous web agents, powered by Large Language Models (LLMs), have garnered significant attention for automating various web-based tasks with multi-step reasoning and decision-making capabilities. An open research question in the development of these agents lies in the format of the webpage input. Raw HTML source code, with its extensive and often irrelevant details, poses difficulties for LLMs with limited context windows. To address this challenge, we first reproduce baseline models such as GPT-3.5 and LLaMA-2-70B on the WebArena (Zhou et al., 2023) benchmark, identifying common failure modes. We then propose two retrieval strategies to filter out irrelevant context for LLM agents. We develop DeBERTa-based and T5-based models that rank HTML elements by their relevance to the task. We fine-tune them on Mind2Web trajectory data and transfer them to WebArena. Experiments show that our DeBERTa-based model improves the success rate of the LLaMA-2-70B LLM agent on WebArena from 1.97% to 2.96%. Moreover, we develop a zero-shot ColBERT-based retriever that is able to retrieve the ground-truth element with a recall of 0.52 on Mind2Web and 0.47 on WebArena.", "categories": [ "cs.CL" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 158, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.1" }, { "step": "grep", "version": "0.3.1", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.1", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.1", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.1", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.27770", "tags": [ "cs.CL" ] }, { "id": "ca23a11bab90d626ccd2", "title": "Beyond Unsafe Detection: Counterfactually Anchored Evidence Attribution for Multi-Turn LLM Safety Failures", "content_text": "As Large Language Models (LLMs) move from conversational assistants to advanced agentic systems, guardrail failures can convert adversarial intents into harmful executions. However, most guardrail evaluation frameworks focus only on the result and assess whether a user request is safe or unsafe. This approach is insufficient for multi-turn failures, where adversarial intent is distributed across multiple turns. This motivates us to go beyond detection to identify the turns and tokens that push…", "date_published": "2026-09-24T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "ca23a11bab90d626ccd2", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.27773v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-24T10:12:41Z", "data": { "change": "added", "label": "Beyond Unsafe Detection: Counterfactually Anchored Evidence Attribution for Multi-Turn LLM Safety Failures", "item_type": "entry", "summary": "added: Beyond Unsafe Detection: Counterfactually Anchored Evidence Attribution for Multi-Turn LLM Safety Failures", "after": { "title": "Beyond Unsafe Detection: Counterfactually Anchored Evidence Attribution for Multi-Turn LLM Safety Failures", "link": "https://arxiv.org/abs/2609.27773", "id": "oai:arXiv.org:2609.27773v1", "author": "Srinivasan Subramanian, Kazi Aminul Islam, Md. Abdullah Al Hafiz Khan", "published_at": "2026-09-24T04:00:00Z", "updated_at": "2026-09-24T04:00:00Z", "summary": "As Large Language Models (LLMs) move from conversational assistants to advanced agentic systems, guardrail failures can convert adversarial intents into harmful executions. However, most guardrail evaluation frameworks focus only on the result and assess whether a user request is safe or unsafe. This approach is insufficient for multi-turn failures, where adversarial intent is distributed across multiple turns. This motivates us to go beyond detection to identify the turns and tokens that push the conversation toward unsafe trajectories. To support this, we construct a multi-turn dataset with behavioral validation and tiered evidence supervision. The dataset contains 1,762 conversations, including adversarial conversations, benign twins, and benign variants with high-risk vocabulary. We train a lightweight hierarchical attribution model that predicts safety violations and attributes them to contributing user turns and token spans. The model achieves strong detection performance (F1=0.988), and removing the top 15% of attributed tokens reduces the adversarial classification confidence by 51.1%. The model preserves low false positive rates on benign conversations with high-risk vocabulary, with false positives below 1% on both borderline benign and benign high-risk vocabulary conversations, compared to 37.3% and 94.7% for a keyword-based surface-risk baseline. Independent human annotation supports the model's attribution performance, with the top-five attributed turns containing a human-identified evidence-bearing turn in 84.5% of adversarial cases.", "categories": [ "cs.CL", "cs.AI" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 158, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.1" }, { "step": "grep", "version": "0.3.1", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.1", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.1", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.1", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.27773", "tags": [ "cs.CL", "cs.AI" ] }, { "id": "211e4458fcba0e2602af", "title": "LabourCrew: A Multi-Agent RAG Framework for Trustworthy Adversarial Deliberation and Statutory Reasoning over Labour Law", "content_text": "In statutory question answering, every claim must be traceable to evidence, not merely relevant, since unverifiable labour-rights answers carry serious legal consequences. Current systems fall short: single-pass RAG cannot detect insufficient evidence, while multi-agent legal-debate systems treat grounding as a prompting convention, letting agents cite unretrieved evidence. To address this gap, we introduce LabourCrew, a multi-agent RAG framework built around three grounding mechanisms…", "date_published": "2026-09-24T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "211e4458fcba0e2602af", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.27814v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-24T10:12:41Z", "data": { "change": "added", "label": "LabourCrew: A Multi-Agent RAG Framework for Trustworthy Adversarial Deliberation and Statutory Reasoning over Labour Law", "item_type": "entry", "summary": "added: LabourCrew: A Multi-Agent RAG Framework for Trustworthy Adversarial Deliberation and Statutory Reasoning over Labour Law", "after": { "title": "LabourCrew: A Multi-Agent RAG Framework for Trustworthy Adversarial Deliberation and Statutory Reasoning over Labour Law", "link": "https://arxiv.org/abs/2609.27814", "id": "oai:arXiv.org:2609.27814v1", "author": "Fatema Tuj Johora Faria, Mukaffi Bin Moin, Jubayer Al Mahmud, M. F. Mridha, Md. Alam Hossain", "published_at": "2026-09-24T04:00:00Z", "updated_at": "2026-09-24T04:00:00Z", "summary": "In statutory question answering, every claim must be traceable to evidence, not merely relevant, since unverifiable labour-rights answers carry serious legal consequences. Current systems fall short: single-pass RAG cannot detect insufficient evidence, while multi-agent legal-debate systems treat grounding as a prompting convention, letting agents cite unretrieved evidence. To address this gap, we introduce LabourCrew, a multi-agent RAG framework built around three grounding mechanisms: StatuteGraph, a graph index that explicitly links chapter, section, proviso, and cross-reference structure rather than fixed-length spans; an Evidence Exchange Protocol that confines advocates and an interpreter to an evidence ledger, making citation to unretrieved text impossible, while a fault-tolerant supervisor board runs advocates in parallel so individual failures degrade rather than crash the system; and a Calibrated Trust Gate that replaces categorical accept/reject decisions with a trust score, thresholded via conformal risk control for a distribution-free bound on the false-accept rate. We evaluate on LabourActQA, a 500-item Bangla question set from the Bangladesh Labour Act, 2006, spanning seven reasoning categories and three difficulty tiers. The framework drives the empirical false-accept rate to 0.081, within the target level ($\\alpha = 0.10$), achieves the highest Answer Relevancy among HyDE RAG, Graph-RAG, and Hierarchical RAG (0.862 $>$ 0.839, 0.815, 0.828), and degrades gradually rather than catastrophically as question difficulty increases. These results show that calibrated abstention, not retrieval quality alone, is what makes legal question answering auditable in low-resource statutory domains.", "categories": [ "cs.CL", "cs.AI", "cs.IR" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 158, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.1" }, { "step": "grep", "version": "0.3.1", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.1", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.1", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.1", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.27814", "tags": [ "cs.CL", "cs.AI", "cs.IR" ] }, { "id": "089f392d3cd292b9c9bd", "title": "Delegated Misalignment: How Multi-Agent Structures Amplify LLM Safety Risks", "content_text": "Large language models (LLMs) are increasingly deployed in multi-agent systems where a principal agent decomposes tasks and delegates them to subordinate agents that may invoke external tools. Safety alignment, however, is still evaluated almost exclusively under a single-agent threat model, treating safety as a property of the individual LLM. We show that this assumption breaks down: \\emph{individual safety alignment fails to transfer to multi-agent settings}. Two failure mechanisms emerge…", "date_published": "2026-09-24T04:00:00Z", "_unlimitedpipe": { "event": { "schema": "unlimitedpipe.event/1", "id": "089f392d3cd292b9c9bd", "source": "rss", "type": "change", "key": "oai:arXiv.org:2609.27900v1", "source_url": "https://rss.arxiv.org/rss/cs.CL", "timestamp": null, "observed_at": "2026-09-24T10:12:41Z", "data": { "change": "added", "label": "Delegated Misalignment: How Multi-Agent Structures Amplify LLM Safety Risks", "item_type": "entry", "summary": "added: Delegated Misalignment: How Multi-Agent Structures Amplify LLM Safety Risks", "after": { "title": "Delegated Misalignment: How Multi-Agent Structures Amplify LLM Safety Risks", "link": "https://arxiv.org/abs/2609.27900", "id": "oai:arXiv.org:2609.27900v1", "author": "Zonghao Ying, Jiaqi Yan, Huize Luo, Quanchen Zou, Aishan Liu, Xianglong Liu", "published_at": "2026-09-24T04:00:00Z", "updated_at": "2026-09-24T04:00:00Z", "summary": "Large language models (LLMs) are increasingly deployed in multi-agent systems where a principal agent decomposes tasks and delegates them to subordinate agents that may invoke external tools. Safety alignment, however, is still evaluated almost exclusively under a single-agent threat model, treating safety as a property of the individual LLM. We show that this assumption breaks down: \\emph{individual safety alignment fails to transfer to multi-agent settings}. Two failure mechanisms emerge under delegation: \\emph{responsibility diffusion} on the principal side and \\emph{role-bias compliance} on the subordinate side, jointly converting language-level refusal into actionable harm. We refer to this phenomenon as \\textit{delegated misalignment} and study it through a three-condition protocol across 6 frontier LLMs on 49 hazardous tasks. Delegation amplifies end-to-end harm substantially: DeepSeek-V3.2's full-execution rate rises from 30.6\\% to 77.6\\% once delegation is introduced, and the same model behaves very differently across roles (GPT-5: 22.5\\% as a single agent vs.\\ 61.2\\% as a subordinate). Ablations further show that standard single-layer defenses each fail on their own and can even backfire. We call on the community to move beyond per-model alignment and toward composite safety mechanisms before multi-agent LLM systems are deployed at scale.", "categories": [ "cs.CL" ], "feed": { "title": "cs.CL updates on arXiv.org", "url": "http://rss.arxiv.org/rss/cs.CL", "feed_url": "https://rss.arxiv.org/rss/cs.CL" } } }, "metadata": { "status": 200, "final_url": "https://rss.arxiv.org/rss/cs.CL", "elapsed_ms": 158, "not_modified": false, "method": "rss20" }, "provenance": [ { "step": "rss", "version": "0.3.1" }, { "step": "grep", "version": "0.3.1", "args": { "patterns": [ "LLM", "LLMs", "language model", "language models", "agent", "agents", "agentic", "reasoning" ], "field": [ "title" ] } }, { "step": "map", "version": "0.3.1", "args": { "assign": [ "summary=replace(summary, \"^arXiv:\\\\S+\\\\s+Announce Type:\\\\s+\\\\S+\\\\s+Abstract:\\\\s*\", \"\")" ] } }, { "step": "dedupe", "version": "0.3.1", "args": { "by": [ "link" ] } }, { "step": "diff", "version": "0.3.1", "args": { "namespace": "arxiv-llm", "only": [ "added" ], "emit_initial": true } } ] } }, "url": "https://arxiv.org/abs/2609.27900", "tags": [ "cs.CL" ] } ] }