LLM
-
Every Model Leaves a Trace: Multi-Resolution LLM Fingerprinting
2026-10-20 16:00:34Mohammad Reza Modarres, Armin Tourajmehr, Artin Tavasoli, Yadollah Yaghoobzadeh, Mohammad Taher PilehvarAbstract
Large language models leave model-specific traces in their generated text, but it remains unclear how finely these fingerprints can distinguish related models and which signals remain informative as obvious cues are reduced. We study this through fingerprint resolution, considering both model resolution, from different families to distilled and adapted variants, and signal resolution across behavioral, stylometric, lexical, and semantic evidence. We progressively suppress surface cues, paraphrase outputs, vary the amount of available text, and test generalization across prompt distributions and languages. Our results show that fingerprints persist beyond broad family differences, remaining detectable in distilled models and even LoRA adaptations of the same base model. Although attribution weakens as surface and lexical cues are altered, substantial model-specific signal remains. These signals transfer across prompt sources and languages, with semantic representations proving most robust across languages. These findings show that model identity is recoverable at multiple resolutions, but that recoverability depends on model similarity and the evidence available in the text.
Abstract (translated)
URL
https://arxiv.org/abs/2610.28434
PDF
-
A Digital Twin to Improve Cyber-Social Impact
2026-10-20 15:58:33Vita Santa Barletta, Danilo Caivano, Gabriel Cellammare, Andrea Dimauro, Pasquale Mesagna, Massimiliano MorgaAbstract
The increasing complexity of Industrial Control Systems (ICS) and their integration with IT infrastructures has significantly expanded the cyber threat landscape. Traditional cybersecurity models, primarily based on the Confidentiality, Integrity, and Availability (CIA triad), fail to capture the broader societal and operational impacts of cyberattacks. This paper proposes a novel framework that integrates Digital Twin technology with a Cyber-Social Security model to simulate, analyze, and quantify cyber threats in critical infrastructures. Virtualized ICS environments built with Mininet and MiniCPS are used to model attack scenarios and assess their consequences across technical and social dimensions. At the core of the framework lies the Cyber-Social Vulnerability Scoring System (CSVSS), a quantitative metric that extends CVSS with a Social Impact dimension grounded in the Entity and Place pillars, producing a single, reproducible severity score that reflects real-world consequences. Crucially, the framework does not automate risk assessment but keeps the human analyst at its center: a Large Language Model (LLM) performs a first-pass analysis, mapping vulnerabilities and drafting risk reports, while the analyst, supported by explainable AI feedback, validates the findings and retains decision authority over critical cases. Experimental results demonstrate the framework's ability to model and score cyber-social impact, while the human-in-the-loop design is presented as a conceptual principle to preserve oversight.
Abstract (translated)
URL
https://arxiv.org/abs/2610.28430
PDF
-
After the Attacker Leaves: Benchmarking Carryover Attacks on LLM Agents
2026-10-20 15:55:05Zhiyue Xu, Xukai Wang, Zhiyuan Huang, Mingjie Zhan, Jian Zhao, Xu-Yao ZhangAbstract
Modern LLM agents persist state across sessions via two channels: long-lived memory files silently prepended to the system prompt, and scheduled tasks waking the agent without user input. A one-time indirect prompt injection committed to either channel can re-fire across sessions long after the attacker leaves, threats we term carryover attacks. Existing prompt-injection benchmarks evaluate attacks only in the session where malicious content is first observed, leaving the gate governing when and how often a planted payload later re-fires unmeasured. We present CarryoverBench: 12 personal-assistant scenarios and 101 attack tasks across three trigger channels: Instant, same-turn IPI; Autonomous, heartbeat-scheduled; and Latent, memory-resurfaced. A rule-based verifier scores planting and triggering as independent stages. Across five frontier models, planted payloads re-fire without measurable decay over five rounds; conditional on successful planting, the trigger stage often provides little additional protection. Porting the Latent channel to a second runtime, Hermes, attenuates but does not eliminate the effect, suggesting that carryover defense requires runtime-layer controls in addition to prompt-level interventions.
Abstract (translated)
URL
https://arxiv.org/abs/2610.28425
PDF
-
Code2Video-Compass: A Comprehensive Benchmark for General-Purpose Code-Based Video Generation across T2V, I2V, and V2V
2026-10-20 15:51:41Zhenghan Chen, Zekai Shao, Xin Lin, Leting Cheng, Siming ChenAbstract
Code-based video generation offers fine-grained control, targeted editing, and component reuse compared with direct video generation. This approach requires multimodal large language models (MLLMs) to translate visual and temporal requirements into executable programs that produce complete videos. However, existing code-based benchmarks primarily target specialized tasks, leaving complete video creation across diverse creative categories and input modalities insufficiently evaluated. We introduce Code2Video-Compass, a general-purpose benchmark for code-based video generation across text-to-video (T2V), image-to-video (I2V), and video-to-video (V2V) settings. Code2Video-Compass comprises 2,104 tasks spanning 10 categories, 24 subtypes, and three complexity dimensions covering construction, dynamic animation, and state dependency. It combines tasks curated from textual instructions and visual references with corresponding T2V, I2V, and V2V tasks derived from complete professional and user-created videos. We propose a dual-level evaluation framework combining 12 objective metrics and five subjective metrics. Objective evaluation covers operational performance, video quality, text-video alignment, and reference fidelity, while subjective evaluation uses an MLLM judge to evaluate instruction following and holistic video quality. Extensive evaluation of 15 recent proprietary and open-weight MLLMs from nine families reveals that even leading models face challenges in reliably generating high-quality complete videos aligned with textual and visual requirements, with performance generally declining as construction, dynamic animation, and state-dependency complexity increases. We anticipate that Code2Video-Compass will advance reliable, coherent, and controllable code-based video generation and inspire broader progress in programmatic visual creation.
Abstract (translated)
URL
https://arxiv.org/abs/2610.28416
PDF
-
DIVE: Benchmarking Full-Duplex Interaction with Visual Engagement in Continuous Streams
2026-10-20 15:48:29Yuhang Chen, Guangyue Liu, Shilong LiuAbstract
Recent advances in multimodal large language models (MLLMs) have substantially improved spoken interaction and video understanding. Existing online and streaming benchmarks primarily assess causal multimodal perception through timestamped question answering, whereas real-world interaction requires models to continuously identify task-relevant multimodal cues and decide whether, when, and how to respond. This leaves open how well current models can sustain real-time interaction throughout an evolving multimodal stream. To address this gap, we introduce DIVE, a benchmark for evaluating full-duplex interaction with visual engagement under continuous and realistic multimodal streams. It organizes six tasks into three complementary interaction scenarios. (1) Proactive Intervention evaluates whether models can identify emerging visual or semantic cues and intervene at the appropriate moment. (2) Continuous Engagement evaluates sustained participation through continuous visual tracking and visually grounded query interaction. (3) Interaction Coordination evaluates how models coordinate listening and speaking under overlapping and multi-party interactions. DIVE contains 525 audio/video samples with fine-grained human annotations and precise temporal metadata. Its evaluation protocol separates response quality, participation decisions, and response timing. We evaluate representative real-time models using response quality, window response rates, and reaction latency. Experiments show that leading real-time models achieve sub-second average response latency, yet their average window response rates remain at or below 53.77%. Response quality varies substantially across models and tasks, with adapting ongoing responses to newly arriving information remaining particularly challenging. We hope DIVE facilitates progress toward more natural and capable real-time multimodal interactive systems.
Abstract (translated)
URL
https://arxiv.org/abs/2610.28405
PDF
-
MAS-Guard: Unified Guardrails from Individual Agents to Multi-Agent LLM Systems
2026-10-20 15:44:37Mohammad Kachuee, Phu Mon Htut, Sheng Li, Nikolaos Pappas, Miguel Ballesteros Martinez, Yassine Benajiba, Sujith Ravi, Dan RothAbstract
Agentic AI extends large language models (LLMs) from generating responses to planning and executing complex workflows, substantially expanding their utility while exposing applications to unsafe actions, information leakage, and adversarial manipulation. Multi-agent systems further expand these capabilities through specialization and collaboration, but distributed authority, heterogeneous roles, and opaque or potentially compromised participants make safeguarding even more challenging. Existing defenses address important individual threats but provide fragmented coverage across local execution, system-level interactions, and enforcement mechanisms. We introduce MAS-Guard, a unified framework that closes this gap for safeguarding single- and multi-agent AI applications containing both instrumentable and opaque agents. Within instrumentable agents, the proposed local Agent Guardrail (A-Guard) layer applies task-specific experts to assess agentic events. At the system level, developer-owned Agent Charters and a Control Plane combine deterministic structural checks of routes, agent capabilities, and message formats with semantic assessments of roles and message content. These complementary layers enable enforcement through a shared set of interventions: allow, reflection, confirmation, and block. Evaluated across six benchmarks covering diverse single- and multi-agent scenarios with three agent LLMs from different providers, MAS-Guard reduces the overall attack success rate by 71.5-86.3% relative to unguarded agents, while meeting the utility reference criterion on benign tasks.
Abstract (translated)
URL
https://arxiv.org/abs/2610.28397
PDF
-
IDEAgent: Agentic Quality-Diversity Search for Research Idea Generation
2026-10-20 15:42:11Varun Gumma, Navonil Majumder, Soumitra Sinhahajari, Soujanya PoriaAbstract
Large Language Models (LLMs) have significantly advanced the automation of scientific discovery. However, existing systems share a core limitation: they optimize ideas independently for either Quality or Diversity. This often results in closely related ideas or large sets of trivial, unsound, or unclear concepts. We instead argue that research ideation should be framed as a conjunction of Quality and Diversity objectives, i.e., a Quality-Diversity (QD) search. Accordingly, we introduce IDEAgent, a multi-agent framework that tracks idea evolution through lineages. We jointly drive Quality using multi-objective feedback for repair and refinement, while Diversity is achieved through lightweight sequential memory and explicit comparison against completed ideas, historical variants, and rejected proposals. To evaluate this joint objective, we develop Yield, a metric that computes the largest mutually diverse set of ideas satisfying predefined quality thresholds. Across 32 topics spanning 8 CS domains, IDEAgent outperforms the best baseline by 3.89x on Yield, while achieving Yield of at least 2 on 8x as many topics. We further analyze quality improvements, showing that repair and refinement improve logical rigor and clarity while preserving non-obviousness. Lastly, we extend IDEAgent to two algorithm discovery tasks for verifiable quality judgment and find that it still consistently outperforms all baselines, and yields the largest set of high-quality and diverse programs that beat a trivial/standard baseline. To encourage future research on QD-search-based ideation, we open-source IDEAgent at this https URL.
Abstract (translated)
URL
https://arxiv.org/abs/2607.22375
PDF
-
When Copying Is Hard: Copy-Constrained Decoding for Exact Span Reproduction
2026-10-20 15:40:54Jinghui Zhang, Lang Gao, Zongfang Liu, Ruihong Zeng, Zirui Song, Rui Yan, Kentaro Inui, Xiuying ChenAbstract
Precisely reproducing a target span verbatim within a long context is a fundamental capability test for LLMs: even when the target string is present in context, models often still drift within the span or fail to stop correctly, revealing limitations in source grounding and boundary control. This capability has broad practical relevance. For instance, code generation often requires embedding sensitive strings such as API keys exactly as they appear in context. In this work, we construct a benchmark to test exact span reproduction ability across contracts, scientific literature, code, and random sequences, and find that current models still make surprisingly frequent errors. To address this, we propose CopyGen, a lightweight copy-constrained plugin for frozen LLMs that turns copying from a byproduct of next-token prediction into a controlled process. Our key idea is to decompose copying into three distinct decisions: when to copy, how far to copy, and when to stop. Concretely, CopyGen first determines whether to enter copy mode, then predicts a tentative copy length, and finally verifies the generated span token by token. During verification, we adaptively adjust the logits of copy continuation tokens to incentivize or suppress continuation, and emit only the longest valid prefix. Across three LLM backbones and four benchmark tasks, CopyGen improves average exact match by 35% relative without model finetuning, and integrates seamlessly with SFT to further improve it by 4%, while substantially accelerating generation. These results suggest that reliable exact copying is not something LLMs obtain for free, but a decoding-time control problem that requires explicit modeling. Code is available at this https URL.
Abstract (translated)
URL
https://arxiv.org/abs/2610.28393
PDF
-
AM-Bench: A Unified Taxonomy and Evaluation Suite for Agentic Misalignment
2026-10-20 15:40:19Eric Zhang, Terry Jingchen Zhang, Chijioke Ugwuanyi, Jerick Shi, Bernhard Sch\"olkopf, Zhijing JinAbstract
As LLM-based agents are deployed with increasing autonomy in real-world environments, a critical failure mode emerges: agentic misalignment, where agents choose harmful actions over failure. However, existing safety benchmarks primarily evaluate whether agents comply with explicitly harmful instructions, leaving the conditions that cause agents to choose harmful actions poorly understood. To address this limitation, we propose AM-Bench, a unified taxonomy and 35 scenario evaluation benchmark spanning seven industries for agentic misalignment grounded in the fraud triangle, a well-established criminological framework explaining goal-directed misconduct. We adapt this framework to agentic environments and construct our evaluation suite and detection systems along three dimensions: perceived threat, perceived opportunity, and reasoning. We operationalize each dimension into structured subcategories rooted in prior theory: five threat categories derived from the instrumental convergence thesis, seven opportunity categories drawn from the Emergent Strategic Reasoning Risks (ESRR) taxonomy, and a novel four-question binary reasoning framework synthesized from prior literature. Our evaluation of 10 frontier models reveals a 62-percentage-point gap in misalignment across scenarios, with Gemini 3 Flash and Gemini 3.1 Pro exhibiting the highest levels of agentic misalignment while Claude Sonnet 4.6 displays near-zero rates across all conditions. Our reasoning framework further reveals that models may exhibit harmful behaviours without explicit situational awareness or explanation, complicating both mitigation and detection. AM-Bench scenarios and code are available at this https URL
Abstract (translated)
URL
https://arxiv.org/abs/2610.24264
PDF
-
SpecEvo: Speculative Evolution with Large Language Models for Cost-Efficient Scientific Discovery
2026-10-20 15:40:05Ha Minh Hieu, Tongyao Zhu, Do Xuan Long, Min-Yen KanAbstract
LLM-driven evolutionary search is a promising paradigm for scientific discovery, but existing methods remain costly and slow, often applying expensive reasoning uniformly while underusing evidence from failed, redundant, and saturated trials. We introduce Speculative Evolution (SpecEvo), a multi-LLM framework that coordinates lightweight and frontier models through continual consolidation of search evidence. Lightweight models power both parallel Speculators, which generate and refine candidate programs, and an Advisor, which distills accumulated successes, failures, and saturated directions into persistent guidance for subsequent exploration. At sparse checkpoints, a frontier-model Navigator inspects the evolving population and intervenes through synthesis, surgical refinement, or reframing according to search stagnation. This speculate-then-consolidate workflow delegates routine exploration and evidence consolidation to inexpensive models while reserving frontier reasoning for strategic interventions whose benefits extend across many subsequent low-cost evaluations. Across 176 tasks spanning four scientific and algorithmic discovery domains, SpecEvo matches or outperforms frontier-model baselines while reducing API cost by up to 80%, achieves up to 1.5x greater evaluation efficiency than cost-efficient baselines on equation discovery, and improves over the strongest CO-Bench baseline by 3.8%. These results support trajectory-conditioned model coordination as an effective approach to cost-efficient scientific discovery.
Abstract (translated)
URL
https://arxiv.org/abs/2610.28706
PDF
-
CAPSafe: A Benchmark for Contextual Action and Plan Safety Reasoning in Embodied Agents
2026-10-20 15:33:20Youmna Farag, Svetlana Stoyanchev, Simon Keizer, Mohan Li, Rama DoddipatlaAbstract
Reasoning about safety is essential for embodied agents powered by large language models (LLMs). We introduce the Contextual Action and Plan Safety benchmark (CAPSafe), designed to evaluate safety reasoning at two levels: (1) an action-level dataset that tests whether a single action is safe given its environmental context, and (2) a plan-level dataset that evaluates whether agents can generate safe long-horizon plans for complex hazardous household tasks. CAPSafe models safety conditions where hazards are situational (inferred from object properties), multiple hazards can co-occur within the same environment, and each is annotated with a relative severity level for fine-grained analysis. Evaluating nine closed and open-source LLMs reveals a substantial gap between hazard recognition and safe long-horizon planning. Several models achieve high accuracy in recognizing unsafe actions (exceeding 90% in the best cases), yet safe long-horizon planning remains substantially harder. Gemini-3-pro achieves the highest safe success rate (78.19%), while other models remain below 65%. These results highlight that robust safety reasoning over long action sequences remains an open challenge.
Abstract (translated)
URL
https://arxiv.org/abs/2610.28381
PDF
-
FALCA: Falsification-Aware Reinforcement Learning with Hierarchical Credit Assignment for Long-Horizon Deep Search Agents
2026-10-20 15:24:44Zijun Dou, Kun Luo, Chaofan Li, Shuqi Lu, Zheng LiuAbstract
Long-horizon deep search requires LLM agents to iteratively decompose complex information needs, gather evidence from external sources, and maintain a reliable research state under multiple constraints. A central challenge is that early mistakes can be amplified over the search trajectory: agents may anchor on an incorrect candidate, collect partially supporting evidence, overlook contradictions, and accumulate errors that misguide subsequent reasoning. We propose FALCA, a falsification-aware framework that formulates deep search as a propose-falsify process. FALCA adopts a collaborative Searcher-Falsifier architecture, where the Searcher proposes candidate answers or sub-answers, while independent Falsifiers actively seek counterevidence, check unmet constraints, identify conflicts, and produce concise falsification reports for research-state revision. To strengthen this loop, we introduce falsification-aware training, combining SFT warmup on falsification trajectories with agentic RL based on parallel rollouts and hierarchical credit assignment over research stages, state transitions, and actions. Experiments on five search benchmarks show that FALCA improves long-horizon deep search performance over backbone and harness ablations, while reducing confirmation bias and error accumulation through more effective falsification and research-state revision.
Abstract (translated)
URL
https://arxiv.org/abs/2610.28361
PDF
-
Auto-Director: Leveraging Multi-Agent Collaboration for Autonomous Video Editing
2026-10-20 15:11:55Yuxia Chen, Wenbo Zhu, Yangguang Ji, Zhenxiang Jiang, Teng Zhong, Yu ZhangAbstract
Autonomous video editing from open-world resources remains a challenging problem in multimedia content creation. Retrieval-based methods are constrained by fixed source material and lack creative flexibility, while generative models struggle with long-horizon coherence and often diverge from user intent. Both paradigms fail to replicate the structured, multi-stage reasoning that underlies professional post-production. To address these limitations, we present Auto-Director, a tool-augmented multi-agent framework that mirrors professional editorial workflows: a Planning Agent decomposes abstract topics into fact-grounded scripts via retrieval-augmented planning; a Material Agent indexes open-world footage through shot segmentation and visual tagging; an Editor Agent assembles semantically aligned clips with MLLM-driven spatial adaptation; and a Critic Agent governs an iterative Reflect-and-Revise loop that checks scene-level alignment and pacing. To rigorously evaluate this new paradigm, we introduce VEdit-Bench, a benchmark of 100 expert-curated topics with human-annotated Golden Scripts and a five-dimensional metric suite covering Narrative Logic, Factual Accuracy, Semantic Alignment, Compositional Quality, and Stylistic Cohesion. Extensive experiments demonstrate that Auto-Director significantly outperforms baseline methods in both automated metrics and human evaluation, effectively bridging the gap between amateurs and professional-grade video production.
Abstract (translated)
URL
https://arxiv.org/abs/2610.28346
PDF
-
DeclineCraft: A RAG Abstention Benchmark for Jointly Measuring Combined Retrieval and Parametric Knowledge
2026-10-20 15:07:19Atharva Thakur, Abulhair Saparov, Bruno RibeiroAbstract
Reliable retrieval-augmented generation (RAG) requires recognizing when neither retrieved documents nor parametric knowledge supports an answer. We introduce DeclineCraft, a dynamic abstention benchmark whose unanswerable questions target Minecraft feature additions after a corpus cut-off, with audits of possible parametric exposure. Later cut-offs support evaluation of future models. Across 27 LLMs, we measure Correct Abstention Rate (CAR; how often models acknowledge insufficient knowledge on unanswerable questions), Answerable Attempt Rate and Answer Accuracy. We find that (1) single-shot retrieval significantly raises CAR in 38 of 40 configurations (model and thinking setting) versus no retrieval. (2) Iterative RAG, in which the model self-assesses its context and can request up to two further retrieval rounds, significantly lowers CAR versus single-shot RAG in 30 of 40 configurations (mean 46.4%), with little change in Answer Accuracy (68.5% to 69.3%). Adding query rewriting and reranking to hybrid (dense & sparse) retrieval lowers CAR for all 12 models tested. (3) Larger models answer more accurately (Spearman $\rho = +0.60$ with total parameters), but neither model size nor release date shows a significant association with CAR. (4) Without retrieval, enabling thinking on the same weights lowers CAR significantly for 8 of 13 models (by 14% points on average), and the responses shift from acknowledged uncertainty to premise denial.
Abstract (translated)
URL
https://arxiv.org/abs/2610.28340
PDF
-
PACE: Plan, Attend, Check, and Emit for Training-Free Response Timing in Streaming Video Understanding
2026-10-20 14:58:18Siqi Guo, Yuxia Chen, Wenbo Zhu, Yangguang Ji, Zhenxiang Jiang, Teng Zhong, Yu ZhangAbstract
Streaming video understanding requires a model to decide not only what to answer but also when to answer. Evidence arrives progressively, and recent benchmarks identify premature responses, issued before the queried event has concluded, as a recurring failure mode. Existing approaches often learn response timing through additional training, while training-free alternatives typically rely on scalar gating signals that do not explicitly represent when an answer becomes final. We organize chunk-wise streaming errors into three patterns: sampling variation can make an answer unstable, a stable answer may lack support in the observed video, or an answer may be supported by the current evidence while the event remains unfinished. We propose PACE (Plan, Attend, Check, and Emit), a training-free framework designed around these three failure modes. Given a question, PACE drafts a plan specifying what to observe and what will signal completion. As the stream unfolds, it invokes an off-the-shelf video LLM at informative moments, evaluates each draft for stability, grounding in its cited frames, and completion, and records verified observations in an event ledger. Experiments on ProactiveVideoQA and OVO-Bench Forward show that PACE improves over the official protocols by up to 5.9 average PAUC points and 6.2 average accuracy points across proprietary and open-source backbones, yielding state-of-the-art performance.
Abstract (translated)
URL
https://arxiv.org/abs/2610.28323
PDF
-
ABACUS: A Benchmark for Image Editing Models to Assess Object Category and Counting Under Scene Conservation
2026-10-20 14:53:22Aymen Lassoued, Angelique Loesch, Romaric AudigierAbstract
Instruction-based image editing models can now add, remove a specified number of objects or change their category while preserving the scene. Yet these models often fail to produce the correct number, and these failures grow sharply as scenes become more complex and the requested count increases. Current benchmarks largely miss this problem: they score edits with holistic similarity metrics or with vision-language model (VLM) judges that are themselves poor at counting, so numerical errors go unnoticed or are graded incorrectly. As a result, the true numerical ability of today's editing models has never been accurately measured. We address this issue with ABACUS, a new benchmark, evaluation protocol and dedicated scoring metric for numerical image editing. Our protocol is judge-free: no VLM or MLLM scores an edit or counts objects (a VLM only confirms the category of rare ambiguous crops). Instead, it verifies each edit at the instance level against panoptic ground truth confirmed by detector consensus, checking both that the exact requested count was achieved and that the rest of the scene was preserved, across the add, remove and change operations. By evaluating leading image editors, our method reveals substantial numerical editing failures that worsen with increasing scene complexity and object count; failures that prior benchmarks often overlooked or misjudged.
Abstract (translated)
URL
https://arxiv.org/abs/2610.28316
PDF
-
Disclosure Begets Disclosure: A Self-Reinforcing Privacy Loop in LLM Agents with Persistent Memory
2026-10-20 14:52:57Masahiro Kaneko, Hiromi Arai, Timothy BaldwinAbstract
Large language model (LLM) agents increasingly use persistent memory to personalize responses, and personal attributes in this memory pose a privacy risk if exposed. We ask how consulting memory during dialogue changes what users go on to say, and thus what the memory comes to reveal. In multi-session simulations, a simulated user with 24 personal attributes talks with a support agent that does or does not consult memory, and in both settings attributes are inferred from the full accumulated record. Across four support models, memory raises response quality but also empirical leakage, the share of attribute values correctly inferred from the record, by a margin that widens over sessions. The cause is a loop in ordinary, non-adversarial assistance. Memory-conditioned responses carry earlier values forward, often as presuppositions, and users affirm them or leave them uncorrected, so that cues and the agent's own guesses become recorded values. The loop reinforces itself, and removing the agent's mentions of an attribute lowers its later disclosure about twice as much as single-turn effects predict. Users cannot leave it on their own, since correcting a value discloses the attribute, declining to discuss it shifts later predictions toward sensitive values, and incorrect values left in place become entrenched. The loop is best weakened at the agent's response. Instructing the agent not to state or presuppose attributes brings leakage to about the no-memory level while keeping most of the quality gains, whereas filtering or anonymizing what users say or what is stored gives up much of them.
Abstract (translated)
URL
https://arxiv.org/abs/2610.28315
PDF
-
Local Work, Global Progress: A Convergence Theory of AI Agent Collaboration
2026-10-20 14:37:06Ivan Ilin, Peter Richt\'arikAbstract
Teams of large language model (LLM) agents are increasingly applied to complex tasks, but their collaboration protocols are mostly designed empirically, without convergence guarantees. We formalize collaborative problem solving as block optimization: each agent improves one part of a shared solution, guided by its local objective plus a correction for how its changes affect the other parts. A coordinator checks every proposal against the global objective and merges proposals that do not interfere with each other. We prove linear convergence up to an error floor that vanishes when the corrections are exact. Interfering proposals can also be merged once their combination has been checked; the rate then depends on how well the proposals actually combine, not on worst-case interactions. With GPT-OSS-120B agents on robot routing, a three-agent team finds collision-free plans in every run, better than those of a single agent that controls all robots. On a larger routing task, a six-agent team completes its search about twice as fast as the single agent (in median). Merging the best checked combination of proposals, rather than only proposals that do not interfere, cuts the team's search time by about 30%. With constants measured on the runs, our convergence bound holds in every round and follows the observed progress. We also test the theory on query planning in SQLite and on a synthetic task.
Abstract (translated)
URL
https://arxiv.org/abs/2610.28286
PDF
-
ChartEvo: Self-Evolving Agent Harnesses for Long-Tail Chart Reasoning
2026-10-20 14:34:54Qi Wang, Peijie Wang, Jun Cao, Haoyang Tong, Xiao-Hui Li, Fei Yin, Cheng-Lin LiuAbstract
Multimodal large language models (MLLMs) have substantially advanced chart understanding and reasoning, yet their performance remains highly uneven across chart types. While modern MLLMs perform well on common charts, their accuracy drops sharply on long-tail charts with diverse visual structures and reasoning requirements. Tool-augmented agents can improve complex chart reasoning by introducing specialized visual capabilities, but manually engineering strategies and tools for the heterogeneous and evolving long tail is difficult to scale. We introduce \textbf{ChartEvo}, a recursive self-improvement framework that enables chart agents to adapt to underperforming chart types through experience-driven harness-level self-improvement. ChartEvo jointly evolves reusable chart-specific Skills and executable Tools, while learning reusable local tool-call chains to compose evolved capabilities and control the inference overhead of tool-augmented reasoning. On long-tail charts from ChartBench and ChartX, ChartEvo improves accuracy over its base MLLM, Qwen3.7-Plus, by \textbf{45.6\%} and \textbf{32.5\%}, respectively, demonstrating the effectiveness of recursive self-improvement for long-tail chart adaptation.
Abstract (translated)
URL
https://arxiv.org/abs/2610.28282
PDF
-
How Language Models Learn Taste: Brand Preference in LLMs
Abstract
Asked to name a running shoe brand, language models say Nike. Asked to recommend one, they say Brooks. This paper asks where models learn that taste. Across 102 models and 42 brand categories, the brand a model picks when recommending differs from the brand it names in 26. The field moves together to a shared recommendation. The Named brands are market leaders, and the Picked brands are mostly niche and specialist. Three open training pipelines show where this forms. Pretraining already attaches the recommended brand to advice. Supervised fine-tuning teaches the model to answer and mostly leaves the pick unchanged. Preference optimization (DPO) teaches the model which answer is better, and the pick moves in all three pipelines, by 11 to 24 points. The preference data contains almost no brands. DPO moves the model's judgments toward a second set of associations even when the preferred brand is not in the preference data, and the effect reaches beyond brands. Asked to choose a sauce, models switch from ketchup to pesto. We do not know exactly how these preferences arise. We suspect they come from the judge models that decide which answers count as better during preference training. For brands, a recommendation depends less on what a model knows about them than on whether they fit this taste.
Abstract (translated)
URL
https://arxiv.org/abs/2610.28272
PDF
