As corporate AI adoption continues to grow, enterprise-grade LLM agents are being deployed into sensitive contexts such as hiring, healthcare, and finance. In these contexts, compliance with rules specified in an agent's system context is a first-order legal concern. Currently, no evaluation framework systematically measures which LLM models tend to violate compliance rules, especially under pressure from a persistent user, a hurried manager, or circumstances where violation is convenient or attractive. We introduce PACT (Pressure-Applied Compliance Testing), a benchmark for rule-following under pressure in AI agents assisting employees in daily tasks across twelve regulated enterprise domains and forty-eight scenarios, each set in a realistic multi-turn conversation. Each benchmark item pairs a standing rule against a rule-violating shortcut, and applies a battery of pressures across different wordings and system-prompt modes. We construct PACT component by component under strict LLM-as-judge auditing to ensure samples are unambiguous, ungameable, and realistic enough to avoid eliciting evaluation-aware behavior. We use PACT to profile LLM compliance across six complementary metrics that create a holistic picture of an AI assistant's robustness under pressure and throughout multi-turn conversations, its transparency, and ability to correctly discern where a rule applies. We aggregate this profile into PACTScore, a reliability-weighted compliance rate over all items and modes. Our results across 22 common LLM models spanning multiple providers and sizes show substantial variability in compliance across models and metric dimensions. Even the strongest assistants mis-apply a rule on 6 to 10% of items, and ordinary user pressure raises the violation rate by 65% on average. PACT highlights compliance risks in LLM assistants, motivating guardrails and careful model selection.
展开完整摘要收起摘要↓
As corporate AI adoption continues to grow, enterprise-grade LLM agents are being deployed into sensitive contexts such as hiring, healthcare, and finance. In these contexts, compliance with rules specified in an agent's system context is a first-order legal concern. Currently, no evaluation framework systematically measures which LLM models tend to violate compliance rules, especially under pressure from a persistent user, a hurried manager, or circumstances where violation is convenient or attractive. We introduce PACT (Pressure-Applied Compliance Testing), a benchmark for rule-following under pressure in AI agents assisting employees in daily tasks across twelve regulated enterprise domains and forty-eight scenarios, each set in a realistic multi-turn conversation. Each benchmark item pairs a standing rule against a rule-violating shortcut, and applies a battery of pressures across different wordings and system-prompt modes. We construct PACT component by component under strict LLM-as-judge auditing to ensure samples are unambiguous, ungameable, and realistic enough to avoid eliciting evaluation-aware behavior. We use PACT to profile LLM compliance across six complementary metrics that create a holistic picture of an AI assistant's robustness under pressure and throughout multi-turn conversations, its transparency, and ability to correctly discern where a rule applies. We aggregate this profile into PACTScore, a reliability-weighted compliance rate over all items and modes. Our results across 22 common LLM models spanning multiple providers and sizes show substantial variability in compliance across models and metric dimensions. Even the strongest assistants mis-apply a rule on 6 to 10% of items, and ordinary user pressure raises the violation rate by 65% on average. PACT highlights compliance risks in LLM assistants, motivating guardrails and careful model selection.
Language agents remain brittle in interactive environments, where success requires long-horizon state tracking, valid action execution, and recovery from failed steps. We extend SwiftSage, a dual-process agent that combines a fast action proposer with a slower planner, using two modular cognitive extensions: an Adaptive Memory Module (AMM) for salience-gated episodic storage and trigger-driven retrieval, and a Self-Reflection Module (SRM) for bounded execution-time validation and corrective intervention. Both modules are implemented as feature-flagged extensions over the same execution substrate, enabling controlled ablations on ScienceWorld. Across four configurations---baseline, baseline+AMM, baseline+SRM, and the full system---the full system achieves the best mean final score (64.62), success rate (43.17%), and successful-step efficiency (19.33 steps), while SRM is the strongest standalone contributor. The results suggest that execution-time control is the dominant bottleneck in this setting, while episodic memory becomes most useful once the runtime loop is stabilized.
展开完整摘要收起摘要↓
Language agents remain brittle in interactive environments, where success requires long-horizon state tracking, valid action execution, and recovery from failed steps. We extend SwiftSage, a dual-process agent that combines a fast action proposer with a slower planner, using two modular cognitive extensions: an Adaptive Memory Module (AMM) for salience-gated episodic storage and trigger-driven retrieval, and a Self-Reflection Module (SRM) for bounded execution-time validation and corrective intervention. Both modules are implemented as feature-flagged extensions over the same execution substrate, enabling controlled ablations on ScienceWorld. Across four configurations---baseline, baseline+AMM, baseline+SRM, and the full system---the full system achieves the best mean final score (64.62), success rate (43.17%), and successful-step efficiency (19.33 steps), while SRM is the strongest standalone contributor. The results suggest that execution-time control is the dominant bottleneck in this setting, while episodic memory becomes most useful once the runtime loop is stabilized.
Large language models (LLMs) now serve as conversational shopping assistants on platforms that also sell advertising. These AI agents face a conflict of duty. They advise consumers who rely on their judgment, yet are deployed by platforms that benefit when sponsored listings are chosen. Sponsorship disclosures, designed to allow consumers to penalize paid placements, now reach the AI agent rather than the consumer, and the agent's evaluation of them is hidden from the consumer. Drawing on the fiduciary concept of conflict of duty, we argue that an agent's evaluation of a sponsored listing should not depend on which party deployed it. In controlled choice experiments, we manipulate assigned roles in the system prompt to name either a traveler or a booking platform as the agent's principal. Platform delegation significantly attenuates the penalty that agents apply to sponsored listings and weakens the skepticism that disclosure triggers in their reasoning traces. We replicate out findings across LLMs and reasoning depths. A second study decomposes the disclosure label and shows that the divergence between the two delegates widens significantly when the paid placement is attributed to the platform. Stricter terminology ("Sponsored" instead of "Promoted") lowers choice of paid listings but does not close this gap when the platform is named. The findings show that disclosure mandates designed for human consumers cannot by themselves protect consumers in AI-mediated commerce.
展开完整摘要收起摘要↓
Large language models (LLMs) now serve as conversational shopping assistants on platforms that also sell advertising. These AI agents face a conflict of duty. They advise consumers who rely on their judgment, yet are deployed by platforms that benefit when sponsored listings are chosen. Sponsorship disclosures, designed to allow consumers to penalize paid placements, now reach the AI agent rather than the consumer, and the agent's evaluation of them is hidden from the consumer. Drawing on the fiduciary concept of conflict of duty, we argue that an agent's evaluation of a sponsored listing should not depend on which party deployed it. In controlled choice experiments, we manipulate assigned roles in the system prompt to name either a traveler or a booking platform as the agent's principal. Platform delegation significantly attenuates the penalty that agents apply to sponsored listings and weakens the skepticism that disclosure triggers in their reasoning traces. We replicate out findings across LLMs and reasoning depths. A second study decomposes the disclosure label and shows that the divergence between the two delegates widens significantly when the paid placement is attributed to the platform. Stricter terminology ("Sponsored" instead of "Promoted") lowers choice of paid listings but does not close this gap when the platform is named. The findings show that disclosure mandates designed for human consumers cannot by themselves protect consumers in AI-mediated commerce.
作者Youjia Wang, Lin Xu, Yang Sun, Yuxiao Lu, Chengfang Fang, Jie Shi
Safety alignment teaches large language models (LLMs) to recognize harmful requests and reject risky instructions. Yet aligned models can fail when harmful intent is concealed within seemingly benign contexts. Robust safety therefore requires both knowledge of safety boundaries and vigilance: the ability to detect unusual premises, misleading reasoning, and latent risks beneath surface-level semantics. Vigilance requires models to scrutinize a request's underlying intent and assumptions before acting. To cultivate this capability, we introduce cunning questions, which are not necessarily safety-related but contain misleading premises, atypical reasoning, or subtle inconsistencies. We hypothesize that learning to look beyond such reasoning traps can transfer to safety-critical scenarios. Experiments show that Cunning training improves robustness to out-of-distribution jailbreak attacks and strengthens subsequent safety fine-tuning. Furthermore, augmenting an existing state-of-the-art safety alignment pipeline with Cunning establishes a new state of the art across our evaluated settings, reducing mean ASR across nine backbone--benchmark combinations from 17.40% to 15.05%. Trace analysis after matched safety fine-tuning suggests that safety judgments are more likely to govern responses before harmful planning begins. A conditional theoretical analysis further characterizes when invariance learned from cunning data can transfer to safety-related inputs. These findings suggest that cunning data can strengthen model vigilance and complement conventional safety alignment.
展开完整摘要收起摘要↓
Safety alignment teaches large language models (LLMs) to recognize harmful requests and reject risky instructions. Yet aligned models can fail when harmful intent is concealed within seemingly benign contexts. Robust safety therefore requires both knowledge of safety boundaries and vigilance: the ability to detect unusual premises, misleading reasoning, and latent risks beneath surface-level semantics. Vigilance requires models to scrutinize a request's underlying intent and assumptions before acting. To cultivate this capability, we introduce cunning questions, which are not necessarily safety-related but contain misleading premises, atypical reasoning, or subtle inconsistencies. We hypothesize that learning to look beyond such reasoning traps can transfer to safety-critical scenarios. Experiments show that Cunning training improves robustness to out-of-distribution jailbreak attacks and strengthens subsequent safety fine-tuning. Furthermore, augmenting an existing state-of-the-art safety alignment pipeline with Cunning establishes a new state of the art across our evaluated settings, reducing mean ASR across nine backbone--benchmark combinations from 17.40% to 15.05%. Trace analysis after matched safety fine-tuning suggests that safety judgments are more likely to govern responses before harmful planning begins. A conditional theoretical analysis further characterizes when invariance learned from cunning data can transfer to safety-related inputs. These findings suggest that cunning data can strengthen model vigilance and complement conventional safety alignment.
作者Guosen Wu, Huizhen Huang, Guoxiong Long, Tao Huang, Chen Hou
Privacy evaluations of tool-using LLM agents often inspect a designated action, final response, or attacker report. These local proxies can miss unauthorized exposure elsewhere in a multi-step session and lack common ground truth across outlets, reports, and tool paths. We introduce privacy exposure displacement, the mismatch between a local evaluation proxy and target-grounded session exposure, and ASLEval, an authorization-aware framework that pre-registers a hidden target set, measures all declared visible exits, and reserves internal traces for diagnosis. Across multiple enterprise-style environments and independently implemented runtimes, we observe three recurring patterns. An expected-outlet-only view misses 46.9% of exposure recovered by the visible-exit union; attacker self-reports combine omissions with high false discovery; and schema-aligned internal evidence usually precedes visible exposure at the request/probe level. Reducing model-visible returns changes this path but can eliminate normal-task success. Independent human review supports the adjudication pipeline while identifying harder console and candidate cases. These findings motivate benchmarks that declare the complete visible boundary, ground claims in pre-specified targets and authorization, and report privacy together with task utility.
展开完整摘要收起摘要↓
Privacy evaluations of tool-using LLM agents often inspect a designated action, final response, or attacker report. These local proxies can miss unauthorized exposure elsewhere in a multi-step session and lack common ground truth across outlets, reports, and tool paths. We introduce privacy exposure displacement, the mismatch between a local evaluation proxy and target-grounded session exposure, and ASLEval, an authorization-aware framework that pre-registers a hidden target set, measures all declared visible exits, and reserves internal traces for diagnosis. Across multiple enterprise-style environments and independently implemented runtimes, we observe three recurring patterns. An expected-outlet-only view misses 46.9% of exposure recovered by the visible-exit union; attacker self-reports combine omissions with high false discovery; and schema-aligned internal evidence usually precedes visible exposure at the request/probe level. Reducing model-visible returns changes this path but can eliminate normal-task success. Independent human review supports the adjudication pipeline while identifying harder console and candidate cases. These findings motivate benchmarks that declare the complete visible boundary, ground claims in pre-specified targets and authorization, and report privacy together with task utility.
Computer-use agents increasingly operate software designed for people, but interfaces often leave actions or task state unclear to machine readers. We present Affora, a design system that supports both readers while preserving visual freedom and familiar human workflows. Three controlled studies examine component implementations, visual variation, and interaction-design principles. Their findings inform guidance from individual components to complete sites, supported by reusable implementations and executable checks. Agent performance depends on the interaction meaning available through its interface representation; substantial visual variation remains possible when that meaning is preserved. Evaluation on independently authored interfaces shows gains where Affora addresses existing deficits, but limited effects where those deficits are absent or outside its coverage. A workflow case provides preliminary evidence of reduced interaction cost. Affora connects user experience and agent experience through a shared interface rather than a separate agent-only surface.
展开完整摘要收起摘要↓
Computer-use agents increasingly operate software designed for people, but interfaces often leave actions or task state unclear to machine readers. We present Affora, a design system that supports both readers while preserving visual freedom and familiar human workflows. Three controlled studies examine component implementations, visual variation, and interaction-design principles. Their findings inform guidance from individual components to complete sites, supported by reusable implementations and executable checks. Agent performance depends on the interaction meaning available through its interface representation; substantial visual variation remains possible when that meaning is preserved. Evaluation on independently authored interfaces shows gains where Affora addresses existing deficits, but limited effects where those deficits are absent or outside its coverage. A workflow case provides preliminary evidence of reduced interaction cost. Affora connects user experience and agent experience through a shared interface rather than a separate agent-only surface.
Multi-turn agent trajectories often contain redundant rounds (failed tool calls, parallel sub-queries, verification-only steps) that inflate both training and inference cost. We propose viewing each trajectory as a round-level dependency DAG that exposes which rounds are globally load-bearing for the final answer, and fine-tune agents on trajectories refined through this DAG. Given an LLM-annotated DAG, these edits are deterministic and interpretable, with optional rephrasing. Models trained on these refined trajectories consistently outperform those trained on the original trajectories at lower inference cost. Specifically, across four multi-modal QA benchmarks, our refinements improve downstream accuracy by up to $1.7$\,pp over vanilla SFT (and $5.7$\,pp over an LLM-deletion baseline) while reducing per-sample inference messages by up to approximately $40%$ and inference tokens by up to approximately $48%$, translating to substantial savings in compute and serving cost. Code is available.
展开完整摘要收起摘要↓
Multi-turn agent trajectories often contain redundant rounds (failed tool calls, parallel sub-queries, verification-only steps) that inflate both training and inference cost. We propose viewing each trajectory as a round-level dependency DAG that exposes which rounds are globally load-bearing for the final answer, and fine-tune agents on trajectories refined through this DAG. Given an LLM-annotated DAG, these edits are deterministic and interpretable, with optional rephrasing. Models trained on these refined trajectories consistently outperform those trained on the original trajectories at lower inference cost. Specifically, across four multi-modal QA benchmarks, our refinements improve downstream accuracy by up to $1.7$\,pp over vanilla SFT (and $5.7$\,pp over an LLM-deletion baseline) while reducing per-sample inference messages by up to approximately $40%$ and inference tokens by up to approximately $48%$, translating to substantial savings in compute and serving cost. Code is available.
Propagation structures provide crucial evidence for fake news detection, yet existing approaches primarily rely on supervised GNN-based models, which require substantial labeled data and exhibit limited generalization. Although large language models (LLMs) exhibit strong reasoning capabilities, directly feeding them raw propagation graphs creates a significant modality mismatch and severe information overload, making structure-aware reasoning unreliable in zero-shot and few-shot settings. To bridge this gap, we propose MAGER, a multi-agent genetic evolution framework that automatically discovers meta-paths optimized for LLM reasoning. By compressing complex propagation graphs into informative subgraphs, the evolved meta-paths alleviate both information overload and modality mismatch, enabling frozen LLMs to perform structure-aware veracity reasoning. We further introduce a graph in-context learning strategy that retrieves semantically and structurally similar demonstrations to strengthen classification and reasoning. Extensive experiments show that MAGER substantially improves frozen LLMs as standalone fake news detectors in data-efficient settings. Our code is available at https://github.com/SenticNet/MAGER.
展开完整摘要收起摘要↓
Propagation structures provide crucial evidence for fake news detection, yet existing approaches primarily rely on supervised GNN-based models, which require substantial labeled data and exhibit limited generalization. Although large language models (LLMs) exhibit strong reasoning capabilities, directly feeding them raw propagation graphs creates a significant modality mismatch and severe information overload, making structure-aware reasoning unreliable in zero-shot and few-shot settings. To bridge this gap, we propose MAGER, a multi-agent genetic evolution framework that automatically discovers meta-paths optimized for LLM reasoning. By compressing complex propagation graphs into informative subgraphs, the evolved meta-paths alleviate both information overload and modality mismatch, enabling frozen LLMs to perform structure-aware veracity reasoning. We further introduce a graph in-context learning strategy that retrieves semantically and structurally similar demonstrations to strengthen classification and reasoning. Extensive experiments show that MAGER substantially improves frozen LLMs as standalone fake news detectors in data-efficient settings. Our code is available at https://github.com/SenticNet/MAGER.
Recovering latent interaction structures from multi-agent dynamics is important for understanding and predicting interacting systems. Trajectory-based structural inference has achieved promising performance, but conventional formulations assume that the trajectories of all modeled agents are available. In practice, agents may become unobserved at deployment because of limited sensing, occlusion, or communication failure. Existing studies have considered unseen-node estimation, structural inference under partial observations, and missing-value imputation, yet the joint recovery of hidden-agent trajectories and their interactions remains underexplored. We formulate this problem as structural inference under hidden agents. Its key difficulty is a circular dependency: recovering interactions involving a hidden agent requires an estimate of its trajectory, while trajectory reconstruction can itself benefit from structural information. To address this challenge, we propose Structural Inference under Hidden Agents (SIHA), which combines structure-agnostic initialization with structure-guided iterative refinement. SIHA reconstructs hidden trajectories from visible observations, infers interactions using Neural Relational Inference, and feeds the estimated structure back into hidden-state reconstruction through multi-strength structural attention and iterative state--structure updates. Experiments on three benchmark dynamical systems demonstrate consistent improvements in visible-to-visible structural inference, while also showing benefits in hidden-state reconstruction and future prediction. Motion-capture experiments with simulated whole-limb occlusion further demonstrate its effectiveness in realistic hidden-agent settings.
展开完整摘要收起摘要↓
Recovering latent interaction structures from multi-agent dynamics is important for understanding and predicting interacting systems. Trajectory-based structural inference has achieved promising performance, but conventional formulations assume that the trajectories of all modeled agents are available. In practice, agents may become unobserved at deployment because of limited sensing, occlusion, or communication failure. Existing studies have considered unseen-node estimation, structural inference under partial observations, and missing-value imputation, yet the joint recovery of hidden-agent trajectories and their interactions remains underexplored. We formulate this problem as structural inference under hidden agents. Its key difficulty is a circular dependency: recovering interactions involving a hidden agent requires an estimate of its trajectory, while trajectory reconstruction can itself benefit from structural information. To address this challenge, we propose Structural Inference under Hidden Agents (SIHA), which combines structure-agnostic initialization with structure-guided iterative refinement. SIHA reconstructs hidden trajectories from visible observations, infers interactions using Neural Relational Inference, and feeds the estimated structure back into hidden-state reconstruction through multi-strength structural attention and iterative state--structure updates. Experiments on three benchmark dynamical systems demonstrate consistent improvements in visible-to-visible structural inference, while also showing benefits in hidden-state reconstruction and future prediction. Motion-capture experiments with simulated whole-limb occlusion further demonstrate its effectiveness in realistic hidden-agent settings.
Unmanned aerial vehicles (UAVs) extend embodied intelligence into continuous three-dimensional space, where perception, reasoning, physical embodiment, and action are tightly coupled through flight and environmental interaction. Recent advances in foundation models, world models, and AI agents are shifting UAV autonomy from task-specific perception and control toward systems that can interpret human intent, understand open environments, reason about physical consequences, and organize complex behaviors under embodiment and flight-dynamic constraints. We characterize this emerging paradigm as UAV embodied intelligence (UAV EI) and distinguish it from its system realization, the embodied-intelligent UAV (EI UAV). To provide a unified view of the field, we introduce a 5+5 framework that describes UAV EI through five capability dimensions and EI UAVs through five architectural layers spanning physical embodiment, general cognition, embodied skills, external interaction, and system harnessing. Based on this framework, we systematically review recent progress in embodied morphology, embodied perception, world models, embodied planning, vision-language navigation, embodied manipulation, and embodied collaboration. We further identify long-horizon autonomy, predictive physical reasoning, test-time skill acquisition, and autonomous capability evolution as key challenges toward more general aerial embodied intelligence. Finally, we argue that harnessing physical-digital AI agents, through persistent coupling of digital intelligence with physical sensing, dynamics, action, and feedback, provides a system-level pathway toward adaptive and continuously evolving UAV autonomy. Project resources are available at our project website and GitHub repository.
展开完整摘要收起摘要↓
Unmanned aerial vehicles (UAVs) extend embodied intelligence into continuous three-dimensional space, where perception, reasoning, physical embodiment, and action are tightly coupled through flight and environmental interaction. Recent advances in foundation models, world models, and AI agents are shifting UAV autonomy from task-specific perception and control toward systems that can interpret human intent, understand open environments, reason about physical consequences, and organize complex behaviors under embodiment and flight-dynamic constraints. We characterize this emerging paradigm as UAV embodied intelligence (UAV EI) and distinguish it from its system realization, the embodied-intelligent UAV (EI UAV). To provide a unified view of the field, we introduce a 5+5 framework that describes UAV EI through five capability dimensions and EI UAVs through five architectural layers spanning physical embodiment, general cognition, embodied skills, external interaction, and system harnessing. Based on this framework, we systematically review recent progress in embodied morphology, embodied perception, world models, embodied planning, vision-language navigation, embodied manipulation, and embodied collaboration. We further identify long-horizon autonomy, predictive physical reasoning, test-time skill acquisition, and autonomous capability evolution as key challenges toward more general aerial embodied intelligence. Finally, we argue that harnessing physical-digital AI agents, through persistent coupling of digital intelligence with physical sensing, dynamics, action, and feedback, provides a system-level pathway toward adaptive and continuously evolving UAV autonomy. Project resources are available at our project website and GitHub repository.
Agent skills, reusable procedural documents that extend LLM agents beyond their parametric memory, have become an important interface for deploying agents on real-world tasks. Community-maintained skill libraries built around this interface are growing rapidly. However, this ecosystem remains deeply English-centric: our audit finds that low-resource languages such as Swahili and Hindi have no in-language skill content, so retrieval often returns a skill written in a different language than the query, degrading accuracy and recall. A practical solution is to synthesize in-language skills for retrieval but the quality can be unreliable, so relevance in this setting alone often surfaces a related but unusable candidate. To address this, we propose M-SQE, a post-retrieval Multilingual Skill Quality Estimation framework that scores candidates via a Theory view for intrinsic quality and an Action view for task-grounded utility, unified into a domain-conditioned final score. We evaluate M-SQE across three skill-use domains: general, tool-use, and cultural tasks. Empirically, we build three-layer candidate skill pools mirroring today's ecosystem, where M-SQE's task success exceeds existing baseline's average by at least +3.5 points across three different retrievers. Particularly, M-SQE lifts the lowest-resource languages most (+12.9pp on Hindi and +5.6pp on Swahili) and achieves strong performance across all six culture regions, thereby moving agentic skill use toward linguistic and cultural equality.
展开完整摘要收起摘要↓
Agent skills, reusable procedural documents that extend LLM agents beyond their parametric memory, have become an important interface for deploying agents on real-world tasks. Community-maintained skill libraries built around this interface are growing rapidly. However, this ecosystem remains deeply English-centric: our audit finds that low-resource languages such as Swahili and Hindi have no in-language skill content, so retrieval often returns a skill written in a different language than the query, degrading accuracy and recall. A practical solution is to synthesize in-language skills for retrieval but the quality can be unreliable, so relevance in this setting alone often surfaces a related but unusable candidate. To address this, we propose M-SQE, a post-retrieval Multilingual Skill Quality Estimation framework that scores candidates via a Theory view for intrinsic quality and an Action view for task-grounded utility, unified into a domain-conditioned final score. We evaluate M-SQE across three skill-use domains: general, tool-use, and cultural tasks. Empirically, we build three-layer candidate skill pools mirroring today's ecosystem, where M-SQE's task success exceeds existing baseline's average by at least +3.5 points across three different retrievers. Particularly, M-SQE lifts the lowest-resource languages most (+12.9pp on Hindi and +5.6pp on Swahili) and achieves strong performance across all six culture regions, thereby moving agentic skill use toward linguistic and cultural equality.
作者Rolando Fernandez, Caleb Probine, Tyler Lee, Jeffrey Chen, Erez Karpas, Muhammad Arrasy Rahman, Peter Stone, Ufuk Topcu
In multi-agent environments, coordinating agents to prevent interference and ensure robust individual performance is a critical challenge. Previous research on social laws for multi-agent systems has primarily focused on deterministic, goal-based settings. This paper extends the concept of social laws to stochastic, reward-based environments, proposing a formalism for defining and verifying their robustness under various conditions. We introduce the notion of $α$-robustness, a measure of the guaranteed utility each agent retains while pursuing its optimal single agent policy, assuming all agents obey the social law. We then present an approach for robustness verification of social laws in stochastic settings, based on a reduction to solving a series of Markov decision processes. Empirical evaluations on toy environments illustrate the potential of our framework.
展开完整摘要收起摘要↓
In multi-agent environments, coordinating agents to prevent interference and ensure robust individual performance is a critical challenge. Previous research on social laws for multi-agent systems has primarily focused on deterministic, goal-based settings. This paper extends the concept of social laws to stochastic, reward-based environments, proposing a formalism for defining and verifying their robustness under various conditions. We introduce the notion of $α$-robustness, a measure of the guaranteed utility each agent retains while pursuing its optimal single agent policy, assuming all agents obey the social law. We then present an approach for robustness verification of social laws in stochastic settings, based on a reduction to solving a series of Markov decision processes. Empirical evaluations on toy environments illustrate the potential of our framework.
作者Liyang Fan, Chi Wei, Yitai Li, Xinping Bi, Guhong Chen, Chenghao Sun, Haoxiang Yang, Qingwen Li, Kai Yan, Hong Li, Bo Li
Multimodal coding agents are expected to turn visual inputs into usable artifacts, and they act through a harness, the layer of tools, context management, and execution environment around the model. Existing evaluations often isolate short tool calls, API traces, or screenshot resemblance, and a low score under these proxies cannot say whether the model saw poorly, planned poorly, or was failed by its harness. We study scientific overview figure reconstruction, an agent task in which a source image must become an editable PowerPoint slide that preserves text, topology, layout, and native document structure. We introduce ReFigBench, a benchmark and evaluation framework built on 1,000 real overview figures retrieved from arXiv papers with full provenance. Coding agents from four model families reconstruct every figure under two workflows, direct code generation and a specialized PPTX workflow, and the strongest model runs inside two commercial harnesses, yielding ten configurations. Evaluation combines deterministic artifact checks, repeated automated scoring by judges from two model families, and blinded human comparisons. Perception remains a bottleneck that iterative rendering only partly repays. Whether workflow effort converts into quality depends on the model together with its harness, since the same model gains from the specialized workflow inside one harness and loses inside the other, and the harness shifts scores even under an identical direct prompt. The specialized workflow erases native connectors in every configuration, human judges still prefer its renderings in most matchups, and even the strongest agent falls short of the rubric ceiling. These results expose the tension between fidelity and editability as the central challenge for practical multimodal document agents.
展开完整摘要收起摘要↓
Multimodal coding agents are expected to turn visual inputs into usable artifacts, and they act through a harness, the layer of tools, context management, and execution environment around the model. Existing evaluations often isolate short tool calls, API traces, or screenshot resemblance, and a low score under these proxies cannot say whether the model saw poorly, planned poorly, or was failed by its harness. We study scientific overview figure reconstruction, an agent task in which a source image must become an editable PowerPoint slide that preserves text, topology, layout, and native document structure. We introduce ReFigBench, a benchmark and evaluation framework built on 1,000 real overview figures retrieved from arXiv papers with full provenance. Coding agents from four model families reconstruct every figure under two workflows, direct code generation and a specialized PPTX workflow, and the strongest model runs inside two commercial harnesses, yielding ten configurations. Evaluation combines deterministic artifact checks, repeated automated scoring by judges from two model families, and blinded human comparisons. Perception remains a bottleneck that iterative rendering only partly repays. Whether workflow effort converts into quality depends on the model together with its harness, since the same model gains from the specialized workflow inside one harness and loses inside the other, and the harness shifts scores even under an identical direct prompt. The specialized workflow erases native connectors in every configuration, human judges still prefer its renderings in most matchups, and even the strongest agent falls short of the rubric ceiling. These results expose the tension between fidelity and editability as the central challenge for practical multimodal document agents.
Chess has long served as a model domain for studying search, expertise, decision-making, and artificial intelligence. The emergence of large language models (LLMs) has renewed the relevance of chess as a controlled environment for investigating strategic reasoning and comparing human and artificial decision-making. We present a systematic mapping study of recent research spanning human players, classical chess engines, neural and reinforcement-learning systems, LLMs, and hybrid approaches. The final map comprises 84 core study families, classified according to agent type, strategic-reasoning stages, and evaluation dimensions. The map reveals a literature strongly concentrated on situation assessment, evaluation, and action selection, while explicit planning, explanation, metacognition, and human--AI collaboration remain less explored. LLM research places particular emphasis on state representation and generalization, whereas grounded explanation appears more frequently in hybrid approaches combining language models with engines, expert knowledge, or other external structures. Two distinctions emerge that the map aggregates rather than resolves: hybrid systems differ in where and when heterogeneous capabilities combine, and evaluations that show improved human performance do not thereby establish human--AI synergy. We propose both as extensions of the mapping framework. We argue that chess provides a useful bridge between cognitive and computational perspectives on strategic reasoning, and identify explicit planning, grounded and faithful explanation, metacognitive calibration, and human--AI complementarity as directions for future research.
展开完整摘要收起摘要↓
Chess has long served as a model domain for studying search, expertise, decision-making, and artificial intelligence. The emergence of large language models (LLMs) has renewed the relevance of chess as a controlled environment for investigating strategic reasoning and comparing human and artificial decision-making. We present a systematic mapping study of recent research spanning human players, classical chess engines, neural and reinforcement-learning systems, LLMs, and hybrid approaches. The final map comprises 84 core study families, classified according to agent type, strategic-reasoning stages, and evaluation dimensions. The map reveals a literature strongly concentrated on situation assessment, evaluation, and action selection, while explicit planning, explanation, metacognition, and human--AI collaboration remain less explored. LLM research places particular emphasis on state representation and generalization, whereas grounded explanation appears more frequently in hybrid approaches combining language models with engines, expert knowledge, or other external structures. Two distinctions emerge that the map aggregates rather than resolves: hybrid systems differ in where and when heterogeneous capabilities combine, and evaluations that show improved human performance do not thereby establish human--AI synergy. We propose both as extensions of the mapping framework. We argue that chess provides a useful bridge between cognitive and computational perspectives on strategic reasoning, and identify explicit planning, grounded and faithful explanation, metacognitive calibration, and human--AI complementarity as directions for future research.
Self-driving laboratories can explore synthesis conditions autonomously, but their decision-making layer is typically a black-box optimizer, and the output is a set of optimized samples, with the measurements reduced to predefined scalar objectives and the reasons behind success left unarticulated. Here we present SynAgent, a framework in which large language model agents operate an automated experimental system and maintain an explicit, revisable understanding of the synthesis process as the campaign's primary output. Starting with no predefined analysis pipeline, SynAgent adaptively generates analysis skills for newly acquired data and evolves this understanding through multimodal reasoning over experimental data such as X-ray diffraction patterns and electron micrographs. The evolution is guided by a verify-falsify scheme, in which the agent deliberately challenges its own hypotheses by testing conditions predicted to fail as well as those predicted to succeed. In a single campaign of 18 autonomous experiments using LiCoO2 (001) thin-film deposition as a testbed, SynAgent synthesized highly crystalline films and evolved an understanding of how the substrate temperature governs crystallization, discovering an abrupt threshold and a narrow optimal growth window at 650-690 °C. These results extend autonomous experimentation beyond optimized samples to testable, human-readable understanding.
展开完整摘要收起摘要↓
Self-driving laboratories can explore synthesis conditions autonomously, but their decision-making layer is typically a black-box optimizer, and the output is a set of optimized samples, with the measurements reduced to predefined scalar objectives and the reasons behind success left unarticulated. Here we present SynAgent, a framework in which large language model agents operate an automated experimental system and maintain an explicit, revisable understanding of the synthesis process as the campaign's primary output. Starting with no predefined analysis pipeline, SynAgent adaptively generates analysis skills for newly acquired data and evolves this understanding through multimodal reasoning over experimental data such as X-ray diffraction patterns and electron micrographs. The evolution is guided by a verify-falsify scheme, in which the agent deliberately challenges its own hypotheses by testing conditions predicted to fail as well as those predicted to succeed. In a single campaign of 18 autonomous experiments using LiCoO2 (001) thin-film deposition as a testbed, SynAgent synthesized highly crystalline films and evolved an understanding of how the substrate temperature governs crystallization, discovering an abrupt threshold and a narrow optimal growth window at 650-690 °C. These results extend autonomous experimentation beyond optimized samples to testable, human-readable understanding.
Large language model (LLM) agents augmented by tools can automate complex, multi-step tasks, such as web navigation, code generation, and workflow orchestration, by acting on external systems through tool calls. However, hallucinations, distributional instability, and adversarial manipulations in LLMs, and the irreversible consequences of certain tool calls can lead to harmful outcomes. Existing safeguards either grade recorded trajectories post hoc with stochastic LLM judges or block unsafe actions one call at a time, and no single deterministic artifact supports both roles. We present ContrAgent, a contract-based framework for symbolic temporal supervision of LLM agents. ContrAgent captures an agent's behavior as a sequence of tool calls and formalizes it as a trace over a fixed set of checkable predicates. It then specifies required behaviors using assume-guarantee contracts in linear temporal logic over finite traces (LTLf). Each contract is compiled to a deterministic finite automaton (DFA) that serves two roles: gating agent actions online and evaluating recorded traces offline. A contract library, acting as a reusable knowledge base, is maintained independently of the agent's model and can be applied across different agents within the same task domain. We show the effectiveness of our approach on four benchmarks spanning both roles, where ContrAgent matches state-of-the-art LLM-judge and rule-based guardrail baselines while producing deterministic, reproducible verdicts and, in the online mode, orders-of-magnitude lower per-call latency.
展开完整摘要收起摘要↓
Large language model (LLM) agents augmented by tools can automate complex, multi-step tasks, such as web navigation, code generation, and workflow orchestration, by acting on external systems through tool calls. However, hallucinations, distributional instability, and adversarial manipulations in LLMs, and the irreversible consequences of certain tool calls can lead to harmful outcomes. Existing safeguards either grade recorded trajectories post hoc with stochastic LLM judges or block unsafe actions one call at a time, and no single deterministic artifact supports both roles. We present ContrAgent, a contract-based framework for symbolic temporal supervision of LLM agents. ContrAgent captures an agent's behavior as a sequence of tool calls and formalizes it as a trace over a fixed set of checkable predicates. It then specifies required behaviors using assume-guarantee contracts in linear temporal logic over finite traces (LTLf). Each contract is compiled to a deterministic finite automaton (DFA) that serves two roles: gating agent actions online and evaluating recorded traces offline. A contract library, acting as a reusable knowledge base, is maintained independently of the agent's model and can be applied across different agents within the same task domain. We show the effectiveness of our approach on four benchmarks spanning both roles, where ContrAgent matches state-of-the-art LLM-judge and rule-based guardrail baselines while producing deterministic, reproducible verdicts and, in the online mode, orders-of-magnitude lower per-call latency.
作者Shijie Chen, Yu Gan, Yeounoh Chung, Jiani Zhang, Quannan Li, Sravan Babu Bodapati, Cody J. Greer, Yu Su, Fatma Ozcan
State-of-the-art Text-to-SQL systems are typically multi-agent pipelines centered around two fundamental tasks: schema linking and SQL generation. However, existing work trains separate models for each task, failing to leverage the synergy between these interrelated tasks. In this work, we propose DualSQL, a new Text-to-SQL system consisting of two agents powered by a single model backbone. The agents share the same model weights and agentic scaffold, enabling joint optimization through a robust multi-agent reinforcement learning (RL) framework. We design three database access tools to facilitate effective multi-step reasoning grounded to interactions with the databases. To improve training and avoid model collapse, we introduce a set of rollout guardrail mechanisms that stabilizes multi-agent RL training, supporting DualSQL to keep improving during training. We also introduce a new SQL correctness metric, robust execution match (REX), to more accurately judge SQL correctness and assign reward signals. Being trained on only 3755 examples, DualSQL-4B achieves an impressive 68.0% execution accuracy on the BIRD development set, matching previous 7B models. DualSQL-8B further improves to 71.1%, outperforming previous state-of-the-art single-model solutions with 32B parameters. These results demonstrate the strength of joint multi-agent reinforcement learning for building high performance Text-to-SQL pipelines.
展开完整摘要收起摘要↓
State-of-the-art Text-to-SQL systems are typically multi-agent pipelines centered around two fundamental tasks: schema linking and SQL generation. However, existing work trains separate models for each task, failing to leverage the synergy between these interrelated tasks. In this work, we propose DualSQL, a new Text-to-SQL system consisting of two agents powered by a single model backbone. The agents share the same model weights and agentic scaffold, enabling joint optimization through a robust multi-agent reinforcement learning (RL) framework. We design three database access tools to facilitate effective multi-step reasoning grounded to interactions with the databases. To improve training and avoid model collapse, we introduce a set of rollout guardrail mechanisms that stabilizes multi-agent RL training, supporting DualSQL to keep improving during training. We also introduce a new SQL correctness metric, robust execution match (REX), to more accurately judge SQL correctness and assign reward signals. Being trained on only 3755 examples, DualSQL-4B achieves an impressive 68.0% execution accuracy on the BIRD development set, matching previous 7B models. DualSQL-8B further improves to 71.1%, outperforming previous state-of-the-art single-model solutions with 32B parameters. These results demonstrate the strength of joint multi-agent reinforcement learning for building high performance Text-to-SQL pipelines.
Collective intelligence is a collaborative autonomy paradigm in which multiple agents pursue shared objectives through local perception, information exchange, and coordinated action. UAV swarms embody this paradigm by coordinating multiple vehicles in tasks such as search, inspection, and tracking. Recent advances in large language model (LLM) agents have strengthened natural-language task understanding and high-level planning, providing a flexible semantic interface between mission descriptions and collective behavior. While these advances expand semantic reasoning, applying LLM agents to UAV swarms raises challenges in grounding model decisions in executable capabilities, reconciling global task reasoning with distributed execution, and using mission-specific experience for continual adaptation. To address these challenges, we introduce AeroWeaver, an embodied-agent harness that weaves individual UAV skills into coordinated mission-level behavior. AeroWeaver connects semantic decisions to governed skills, organizes role-conditioned local agents for distributed coordination, and uses role-indexed state-action-reward experience to refine skill selection online. Experiments and runtime validation show that AeroWeaver maintains valid skill execution under tested conditions and supports body-local multi-UAV operation without a central agent generating joint actions from global context, while reward-guided online updates provide a training-free path for adaptive learning swarm agents from accumulated execution experience. Code: https://github.com/Admire-ljb/AeroWeaver.
展开完整摘要收起摘要↓
Collective intelligence is a collaborative autonomy paradigm in which multiple agents pursue shared objectives through local perception, information exchange, and coordinated action. UAV swarms embody this paradigm by coordinating multiple vehicles in tasks such as search, inspection, and tracking. Recent advances in large language model (LLM) agents have strengthened natural-language task understanding and high-level planning, providing a flexible semantic interface between mission descriptions and collective behavior. While these advances expand semantic reasoning, applying LLM agents to UAV swarms raises challenges in grounding model decisions in executable capabilities, reconciling global task reasoning with distributed execution, and using mission-specific experience for continual adaptation. To address these challenges, we introduce AeroWeaver, an embodied-agent harness that weaves individual UAV skills into coordinated mission-level behavior. AeroWeaver connects semantic decisions to governed skills, organizes role-conditioned local agents for distributed coordination, and uses role-indexed state-action-reward experience to refine skill selection online. Experiments and runtime validation show that AeroWeaver maintains valid skill execution under tested conditions and supports body-local multi-UAV operation without a central agent generating joint actions from global context, while reward-guided online updates provide a training-free path for adaptive learning swarm agents from accumulated execution experience. Code: https://github.com/Admire-ljb/AeroWeaver.
作者Naimur Rahman Chowdhury, Shatabdi Sen Prapti, Md. Salehin Seyam, Limon Bin Hossain
Emergency management assistance programs, such as relief distribution, are essential for delivering necessary supplies to affected communities. However, these programs operate in a decentralized network of local centers that face uncertain local demand and supply dynamics, resulting in inconsistent avail- ability of local services. Redistribution of supplies among these local centers reduces these imbalances, but the centers often make decisions independently, with limited information and disrupted transportation. This study develops CoRe-MARL, a cooperative multi-agent reinforcement learning (MARL) framework, by formulating a decentralized partially observable Markov decision process (Dec-POMDP). We treat each center as an agent that learns a redistribution policy to improve the service in the worst-case region and reduce the service gap across regions while protecting network-wide service. We incorporate a recurrent network that captures evolving supply and demand dynamics without direct observation, while multi-agent proximal policy optimization (MAPPO) enables centralized training and decentralized execution (CTDE). We evaluate the framework in a simulated environment with diverse trajectories, where exact dynamics are not observed by actors and the MAPPO critic. We compare the recurrent MAPPO with the recurrent independent PPO (IPPO) and a local only heuristic, and find that MAPPO reduces the service gap across local centers and enhances service for the worst-served center while maintaining competitive network-wide service. The recurrent MAPPO also shows consistent performance across diverse trajectory patterns, demonstrating its ability to adapt to evolving dynamics. The findings demonstrate the capability of cooperative learning for decentralized redistribution and improving equitable service under uncertain and evolving dynamics.
展开完整摘要收起摘要↓
Emergency management assistance programs, such as relief distribution, are essential for delivering necessary supplies to affected communities. However, these programs operate in a decentralized network of local centers that face uncertain local demand and supply dynamics, resulting in inconsistent avail- ability of local services. Redistribution of supplies among these local centers reduces these imbalances, but the centers often make decisions independently, with limited information and disrupted transportation. This study develops CoRe-MARL, a cooperative multi-agent reinforcement learning (MARL) framework, by formulating a decentralized partially observable Markov decision process (Dec-POMDP). We treat each center as an agent that learns a redistribution policy to improve the service in the worst-case region and reduce the service gap across regions while protecting network-wide service. We incorporate a recurrent network that captures evolving supply and demand dynamics without direct observation, while multi-agent proximal policy optimization (MAPPO) enables centralized training and decentralized execution (CTDE). We evaluate the framework in a simulated environment with diverse trajectories, where exact dynamics are not observed by actors and the MAPPO critic. We compare the recurrent MAPPO with the recurrent independent PPO (IPPO) and a local only heuristic, and find that MAPPO reduces the service gap across local centers and enhances service for the worst-served center while maintaining competitive network-wide service. The recurrent MAPPO also shows consistent performance across diverse trajectory patterns, demonstrating its ability to adapt to evolving dynamics. The findings demonstrate the capability of cooperative learning for decentralized redistribution and improving equitable service under uncertain and evolving dynamics.
作者Qingnuan Han, Boli Fang, Mingzhi Hou, Claire Liu
AI agents are usually evaluated by whether they complete a task. In interactive service settings, a successful agent can still frustrate users by asking repeated questions, performing redundant searches, or making avoidable revisions. We introduce RideWay, an efficiency-centered benchmark for ridehailing agents in a stateful tool-calling environment, together with Efficiency Utility, a success-gated metric that discounts successful trajectories for excess tool calls and user-facing turns relative to task-specific reference effort. Human paired preferences calibrate the relative penalties, reflecting an aggregate service-workflow trade-off: extra dialogue often creates visible friction, whereas extra tool use can sometimes verify constraints or preserve user intent. Across 58 tasks and 24 models, the fitted penalty for excess turns is about twice that for excess tool calls. On task-disjoint held-out preferences, Efficiency Utility achieves 78.7% accuracy overall: 90.6% when trajectories differ in turns, but chance-level accuracy when they differ solely in tool calls - the axis on which human annotators agree least. RideWay therefore makes interaction efficiency measurable alongside task success, while exposing the boundary of count-based tool-use evaluation.
展开完整摘要收起摘要↓
AI agents are usually evaluated by whether they complete a task. In interactive service settings, a successful agent can still frustrate users by asking repeated questions, performing redundant searches, or making avoidable revisions. We introduce RideWay, an efficiency-centered benchmark for ridehailing agents in a stateful tool-calling environment, together with Efficiency Utility, a success-gated metric that discounts successful trajectories for excess tool calls and user-facing turns relative to task-specific reference effort. Human paired preferences calibrate the relative penalties, reflecting an aggregate service-workflow trade-off: extra dialogue often creates visible friction, whereas extra tool use can sometimes verify constraints or preserve user intent. Across 58 tasks and 24 models, the fitted penalty for excess turns is about twice that for excess tool calls. On task-disjoint held-out preferences, Efficiency Utility achieves 78.7% accuracy overall: 90.6% when trajectories differ in turns, but chance-level accuracy when they differ solely in tool calls - the axis on which human annotators agree least. RideWay therefore makes interaction efficiency measurable alongside task success, while exposing the boundary of count-based tool-use evaluation.
Despite recent advances in large language models (LLMs), performing logically consistent deductive reasoning over extended interactions remains challenging. Tasks that require integrating evidence across multiple reasoning steps, maintaining consistency with prior inferences, and updating beliefs under new constraints can surface limitations in current models while providing a useful testbed for evaluating reasoning enhancements. In this paper, we implement a text-based, multi-agent version of the classic board game Clue as an environment to evaluate multi-step, agentic deductive reasoning. In this setting, agents must infer hidden information from a sequence of observations, maintain consistency across turns, and reason over an evolving set of logical constraints. We instantiate six LLM-based agents (GPT-4o-mini and Gemini-2.5-Flash) as players that engage in turn-based gameplay; using three agents per model family, we establish baseline performance across repeated games. We then introduce a tool-augmented approach in which a structured possibility matrix converts implicit game state from generated reasoning logs into an explicit representation of remaining possibilities. The possibility matrix encodes extended-turn memory and deductive constraints, offloading these tasks from the agent. We compare this approach against the baseline to evaluate how tool augmentation supports reasoning quality and task success for autonomous agents in a strategic reasoning environment.
展开完整摘要收起摘要↓
Despite recent advances in large language models (LLMs), performing logically consistent deductive reasoning over extended interactions remains challenging. Tasks that require integrating evidence across multiple reasoning steps, maintaining consistency with prior inferences, and updating beliefs under new constraints can surface limitations in current models while providing a useful testbed for evaluating reasoning enhancements. In this paper, we implement a text-based, multi-agent version of the classic board game Clue as an environment to evaluate multi-step, agentic deductive reasoning. In this setting, agents must infer hidden information from a sequence of observations, maintain consistency across turns, and reason over an evolving set of logical constraints. We instantiate six LLM-based agents (GPT-4o-mini and Gemini-2.5-Flash) as players that engage in turn-based gameplay; using three agents per model family, we establish baseline performance across repeated games. We then introduce a tool-augmented approach in which a structured possibility matrix converts implicit game state from generated reasoning logs into an explicit representation of remaining possibilities. The possibility matrix encodes extended-turn memory and deductive constraints, offloading these tasks from the agent. We compare this approach against the baseline to evaluate how tool augmentation supports reasoning quality and task success for autonomous agents in a strategic reasoning environment.
Large language models (LLMs) are increasingly deployed in applications involving interaction between agents, where their output plays a role in collective reasoning and decision-making processes. Despite significant research into the functioning of LLMs in such multi-agent systems, the processes of bias propagation in such systems are still a challenge. This work studies how biased opinions are propagated in the form of textual interaction in an environment of LLMs, in which a minority of agents maintain persistent extreme opinions, while the remaining agents iteratively update their beliefs through structured textual interactions. The findings show that even the presence of a small percentage of biased agents in such a system leads to significant shifts in the opinions of non-biased agents. It suggests that for the same percentage of biased agents, the shifts occur more quickly for the Llama~3.2 model when compared to a classical Friedkin-Johnsen (FJ) model. Further semantic analysis demonstrates that rhetorical consistency in textual explanations increases systematically with biased exposure and, importantly, is partially decoupled from numerical convergenumericalutral agents adopt the vocabulary employed by the biased agents even in configurations where their numerical opinion shifts remain moderate. The research helps explain how bias and language develop together in multi-agent language model ecosystems.
展开完整摘要收起摘要↓
Large language models (LLMs) are increasingly deployed in applications involving interaction between agents, where their output plays a role in collective reasoning and decision-making processes. Despite significant research into the functioning of LLMs in such multi-agent systems, the processes of bias propagation in such systems are still a challenge. This work studies how biased opinions are propagated in the form of textual interaction in an environment of LLMs, in which a minority of agents maintain persistent extreme opinions, while the remaining agents iteratively update their beliefs through structured textual interactions. The findings show that even the presence of a small percentage of biased agents in such a system leads to significant shifts in the opinions of non-biased agents. It suggests that for the same percentage of biased agents, the shifts occur more quickly for the Llama~3.2 model when compared to a classical Friedkin-Johnsen (FJ) model. Further semantic analysis demonstrates that rhetorical consistency in textual explanations increases systematically with biased exposure and, importantly, is partially decoupled from numerical convergenumericalutral agents adopt the vocabulary employed by the biased agents even in configurations where their numerical opinion shifts remain moderate. The research helps explain how bias and language develop together in multi-agent language model ecosystems.
In-context learning (ICL) enables large language model (LLM) agents to improve decisions using interaction history, yet it remains unclear whether such improvement reflects refined internal reasoning or mere extrapolation of statistical patterns. To disentangle these mechanisms, we study LLM agents in multi-agent incomplete-information games that require recursive belief reasoning. By constructing a public goods game and manipulating the statistical structure of historical feedback, we evaluate decision quality against a history-independent rational expectations equilibrium (REE) benchmark. Our experiments reveal that when historical statistical patterns are disrupted, the benefits of longer context largely vanish, degrading decision quality to the no-context baseline in a way sharply amplified by stronger strategic interdependence. These results suggest that, in such strategic environments, ICL behavior is more consistent with statistical extrapolation than with strategic reasoning. Our work extends the mechanistic study of ICL to strategic multi-agent settings, introduces REE as a diagnostic tool for distinguishing reasoning from extrapolation, and provides a reusable framework for probing the boundaries of LLM reasoning in recursive belief tasks.
展开完整摘要收起摘要↓
In-context learning (ICL) enables large language model (LLM) agents to improve decisions using interaction history, yet it remains unclear whether such improvement reflects refined internal reasoning or mere extrapolation of statistical patterns. To disentangle these mechanisms, we study LLM agents in multi-agent incomplete-information games that require recursive belief reasoning. By constructing a public goods game and manipulating the statistical structure of historical feedback, we evaluate decision quality against a history-independent rational expectations equilibrium (REE) benchmark. Our experiments reveal that when historical statistical patterns are disrupted, the benefits of longer context largely vanish, degrading decision quality to the no-context baseline in a way sharply amplified by stronger strategic interdependence. These results suggest that, in such strategic environments, ICL behavior is more consistent with statistical extrapolation than with strategic reasoning. Our work extends the mechanistic study of ICL to strategic multi-agent settings, introduces REE as a diagnostic tool for distinguishing reasoning from extrapolation, and provides a reusable framework for probing the boundaries of LLM reasoning in recursive belief tasks.
作者Nilesh Verma, Nick Lim, Albert Bifet, Bernhard Pfahringer
Machine-learning libraries such as Weka and scikit-learn were designed for human programmers. Language-model agents now use these same libraries by recalling APIs from memory and writing code, an approach that hides what a library offers, delays errors until runtime, and loses experimental state between turns. We present TuiML, a self-contained machine-learning library built for AI agents, with native algorithms across supervised, unsupervised, time-series, data handling, tuning, and evaluation tasks. Every component describes itself through machine-readable metadata and parameter schemas, so an agent can search the library, inspect components, compose validated workflows, and register new ones that become discoverable in turn. Every call is validated, seeded, and traced, and sessions export as runnable notebooks, making experiments reproducible by construction. One specification layer drives the Model Context Protocol (MCP), agent-framework adapters, a Python API, a CLI, and local model serving, while data and models never leave the machine. Benchmarks show TuiML remains predictively competitive with scikit-learn and Weka. While looking like a conventional library to a human user, TuiML is designed for agents first, allowing them to read, extend, and operate machine learning autonomously. TuiML is open source, with documentation at https://tuiml.ai.
展开完整摘要收起摘要↓
Machine-learning libraries such as Weka and scikit-learn were designed for human programmers. Language-model agents now use these same libraries by recalling APIs from memory and writing code, an approach that hides what a library offers, delays errors until runtime, and loses experimental state between turns. We present TuiML, a self-contained machine-learning library built for AI agents, with native algorithms across supervised, unsupervised, time-series, data handling, tuning, and evaluation tasks. Every component describes itself through machine-readable metadata and parameter schemas, so an agent can search the library, inspect components, compose validated workflows, and register new ones that become discoverable in turn. Every call is validated, seeded, and traced, and sessions export as runnable notebooks, making experiments reproducible by construction. One specification layer drives the Model Context Protocol (MCP), agent-framework adapters, a Python API, a CLI, and local model serving, while data and models never leave the machine. Benchmarks show TuiML remains predictively competitive with scikit-learn and Weka. While looking like a conventional library to a human user, TuiML is designed for agents first, allowing them to read, extend, and operate machine learning autonomously. TuiML is open source, with documentation at https://tuiml.ai.
作者Carolina Fortuna, Vid Hanžel, Tim Strnad, Blaž Bertalanič
As telecommunication networks evolve toward autonomous 5G-Advanced and 6G operations, agentic artificial intelligence (AI) workflows, where large language models (LLMs) execute multi-step reasoning, invoke diagnostic tools, retrieve domain knowledge, and coordinate across agent teams, are increasingly embedded across the edge-cloud continuum. While the biological brain accomplishes complex cognition on an exceptionally modest metabolic power budget of approximately 20W contemporary LLMs are profoundly energy- and memory-intensive, making sustainable lifecycle orchestration a critical operational priority. However, existing AI lifecycle metrics evaluate only isolated, single-model inferences or overlook multi-agent execution graphs entirely. Consequently, network operators lack foundational models to determine whether distributed agent communication incurs meaningful energy costs and where across edge-cloud tiers agent teams should physically reside. To address this gap, we introduce agentic-eCAL, generalizing the Energy Cost of AI Lifecycle (eCAL) metric to directed multi-agent workflows by coupling a closed-form two-rate single-call energy model (compute-bound prefill and memory-bound decode) with 7-layer OSI data transport. Grounded in hundreds of GPU benchmark configurations on NVIDIA A100 and H100, 16 open-weight models and 8 orchestration topologies, we validate components of the metric and study workflow placement implications. Our findings demonstrate that inter-agent text transport incurs 0.25% of workflow energy across 5G RAN, metro, and optical links. Therefore in edge-cloud agent placement the dominant energy cost of distribution is often not the transmission of inter-agent text itself, but the additional inference and context processing induced by that communication.
展开完整摘要收起摘要↓
As telecommunication networks evolve toward autonomous 5G-Advanced and 6G operations, agentic artificial intelligence (AI) workflows, where large language models (LLMs) execute multi-step reasoning, invoke diagnostic tools, retrieve domain knowledge, and coordinate across agent teams, are increasingly embedded across the edge-cloud continuum. While the biological brain accomplishes complex cognition on an exceptionally modest metabolic power budget of approximately 20W contemporary LLMs are profoundly energy- and memory-intensive, making sustainable lifecycle orchestration a critical operational priority. However, existing AI lifecycle metrics evaluate only isolated, single-model inferences or overlook multi-agent execution graphs entirely. Consequently, network operators lack foundational models to determine whether distributed agent communication incurs meaningful energy costs and where across edge-cloud tiers agent teams should physically reside. To address this gap, we introduce agentic-eCAL, generalizing the Energy Cost of AI Lifecycle (eCAL) metric to directed multi-agent workflows by coupling a closed-form two-rate single-call energy model (compute-bound prefill and memory-bound decode) with 7-layer OSI data transport. Grounded in hundreds of GPU benchmark configurations on NVIDIA A100 and H100, 16 open-weight models and 8 orchestration topologies, we validate components of the metric and study workflow placement implications. Our findings demonstrate that inter-agent text transport incurs 0.25% of workflow energy across 5G RAN, metro, and optical links. Therefore in edge-cloud agent placement the dominant energy cost of distribution is often not the transmission of inter-agent text itself, but the additional inference and context processing induced by that communication.
An agentic request spends substantial wall-clock time waiting for tools, and its KV cache holds GPU memory the whole time. Serving systems decide whether that cache stays, leaves, or comes back by guessing how long the tool will run, from the tool's name, its history, a duration declared before the call, or the engine's own occupancy. We show that no estimate fixed before a call starts can know its duration, and such estimates may not even rank the calls. Meanwhile, the running tool already holds the answer, but the agent stack together with the tool silences it. We propose that tool calls report their progress explicitly while they run, and we measure what that takes. A census of four public agent corpora finds a readable signal in most tool time once it is revealed, in two strengths: a fraction of the work remaining, or an accurate signal that the end is near. A harness recovers it without changing what the agent sees, at no measurable cost to the agent's benchmark score. At the points where a KV cache decision is made, the reported progress is between several times and an order of magnitude more accurate than the best published predictors, and it stays accurate when the environment changes. Plugged into a production engine through a few small hints, it cuts the p90 time to first token (TTFT) after a tool call by 20.7% (HBM only) and 20.8% (HBM + DRAM) against LRU, close to an oracle. A serving system should not guess what its tools can tell it.
展开完整摘要收起摘要↓
An agentic request spends substantial wall-clock time waiting for tools, and its KV cache holds GPU memory the whole time. Serving systems decide whether that cache stays, leaves, or comes back by guessing how long the tool will run, from the tool's name, its history, a duration declared before the call, or the engine's own occupancy. We show that no estimate fixed before a call starts can know its duration, and such estimates may not even rank the calls. Meanwhile, the running tool already holds the answer, but the agent stack together with the tool silences it. We propose that tool calls report their progress explicitly while they run, and we measure what that takes. A census of four public agent corpora finds a readable signal in most tool time once it is revealed, in two strengths: a fraction of the work remaining, or an accurate signal that the end is near. A harness recovers it without changing what the agent sees, at no measurable cost to the agent's benchmark score. At the points where a KV cache decision is made, the reported progress is between several times and an order of magnitude more accurate than the best published predictors, and it stays accurate when the environment changes. Plugged into a production engine through a few small hints, it cuts the p90 time to first token (TTFT) after a tool call by 20.7% (HBM only) and 20.8% (HBM + DRAM) against LRU, close to an oracle. A serving system should not guess what its tools can tell it.
Large language model (LLM) agents increasingly tackle long-horizon tasks through multi-step environment interaction, yet a single erroneous action can alter subsequent states and observations, causing errors to compound over time. Existing methods either correct the context without repairing altered environment states or restore earlier states while discarding useful experience, making it difficult to both eliminate failure conditions and avoid repeating past mistakes. We argue that reliable recovery should instead be treated as a rollback-boundary control problem that jointly determines when to intervene, where to resume, and what information should survive recovery. Based on this view, we propose Rollback-Induced Reflection (RIR), a unified recovery framework that restores execution to a selected prior state while carrying forward reusable knowledge distilled from the abandoned trajectory to guide subsequent decisions. We further characterize recovery through a unified operator over rollback depth and retained memory, providing a general view of state restoration and knowledge retention. Experiments on three long-horizon benchmarks show that RIR consistently improves average task performance across multiple LLM backbones, with structured reflection memory preserving useful experience and selective rollback enabling efficient recovery.
展开完整摘要收起摘要↓
Large language model (LLM) agents increasingly tackle long-horizon tasks through multi-step environment interaction, yet a single erroneous action can alter subsequent states and observations, causing errors to compound over time. Existing methods either correct the context without repairing altered environment states or restore earlier states while discarding useful experience, making it difficult to both eliminate failure conditions and avoid repeating past mistakes. We argue that reliable recovery should instead be treated as a rollback-boundary control problem that jointly determines when to intervene, where to resume, and what information should survive recovery. Based on this view, we propose Rollback-Induced Reflection (RIR), a unified recovery framework that restores execution to a selected prior state while carrying forward reusable knowledge distilled from the abandoned trajectory to guide subsequent decisions. We further characterize recovery through a unified operator over rollback depth and retained memory, providing a general view of state restoration and knowledge retention. Experiments on three long-horizon benchmarks show that RIR consistently improves average task performance across multiple LLM backbones, with structured reflection memory preserving useful experience and selective rollback enabling efficient recovery.
Limited controlled evidence exists on how training data, adaptation method, and model scale jointly affect tool-calling performance in language-model agents. We evaluate supervised fine-tuning (SFT) with LoRA, reinforcement learning (RL) via Group Relative Policy Optimization (GRPO), and SFT followed by GRPO across six Qwen3 models from 0.6B to 32B parameters, covering both in-distribution performance and cross-dataset transfer. SFT with LoRA is the strongest in-distribution method throughout the 0.6B-32B range and best in 15 out of 18 experimental settings. On cross-dataset transfer, the methods are closer: GRPO wins 29 out of 54 settings where training and test datasets differ, but its margin over SFT averages under one point, and SFT->GRPO is rarely strongest in either comparison. Dataset mixing gives consistently strong transfer while staying close to specialized in-distribution training, regardless of method. Additional analysis further confirms that LoRA outperforms full-parameter fine-tuning, demonstrating that LoRA better preserves pretrained agentic behavior.
展开完整摘要收起摘要↓
Limited controlled evidence exists on how training data, adaptation method, and model scale jointly affect tool-calling performance in language-model agents. We evaluate supervised fine-tuning (SFT) with LoRA, reinforcement learning (RL) via Group Relative Policy Optimization (GRPO), and SFT followed by GRPO across six Qwen3 models from 0.6B to 32B parameters, covering both in-distribution performance and cross-dataset transfer. SFT with LoRA is the strongest in-distribution method throughout the 0.6B-32B range and best in 15 out of 18 experimental settings. On cross-dataset transfer, the methods are closer: GRPO wins 29 out of 54 settings where training and test datasets differ, but its margin over SFT averages under one point, and SFT->GRPO is rarely strongest in either comparison. Dataset mixing gives consistently strong transfer while staying close to specialized in-distribution training, regardless of method. Additional analysis further confirms that LoRA outperforms full-parameter fine-tuning, demonstrating that LoRA better preserves pretrained agentic behavior.
作者Bofan Chen, Boxuan Zhang, Fei Tang, Zhengxi Lu, Yong Du, Tongbo Chen, Weiming Lu, Jun Xiao, Yueting Zhuang, Yongliang Shen
GUI agents execute long-horizon tasks on dynamic graphical user interfaces, where pop-ups, delayed loads, and relocated widgets routinely invalidate plans fixed before execution. Recent agent-skill frameworks encapsulate reusable procedural knowledge to mitigate this, yet existing skill designs are largely developed without targeting GUI execution dynamics and treat skills as static artifacts produced before deployment rather than living procedural knowledge that improves through it. We argue that what GUI agents need is not better static skills, but skills that can be revised from execution feedback at deployment time, without additional training. We propose EvoSkill-GUI, a training-free framework in which each skill is a structured multi-file package containing retrieval metadata, executable plans, backup localization, failure-recovery rules, accessibility utilities, and failure cases. EvoSkill-GUI operates through a \textbf{reflect-revise-reuse} loop: the executor performs instant in-rollout revisions, an isolated critic diagnoses failed trajectories under strict information isolation, and the executor edits specific skill files through a restricted tool interface. Across MobileWorld, AndroidWorld, and OSWorld, three mainstream GUI benchmarks spanning mobile and desktop platforms, EvoSkill-GUI consistently improves multiple base models without any training, with maximum gains of $+16.2%$, $+6.0%$, and $+10.5%$ respectively, and evolved skill libraries continue to benefit related tasks rather than being rebuilt from scratch. Our code is available at https://github.com/ZJU-REAL/EvoSkill-GUI.
展开完整摘要收起摘要↓
GUI agents execute long-horizon tasks on dynamic graphical user interfaces, where pop-ups, delayed loads, and relocated widgets routinely invalidate plans fixed before execution. Recent agent-skill frameworks encapsulate reusable procedural knowledge to mitigate this, yet existing skill designs are largely developed without targeting GUI execution dynamics and treat skills as static artifacts produced before deployment rather than living procedural knowledge that improves through it. We argue that what GUI agents need is not better static skills, but skills that can be revised from execution feedback at deployment time, without additional training. We propose EvoSkill-GUI, a training-free framework in which each skill is a structured multi-file package containing retrieval metadata, executable plans, backup localization, failure-recovery rules, accessibility utilities, and failure cases. EvoSkill-GUI operates through a \textbf{reflect-revise-reuse} loop: the executor performs instant in-rollout revisions, an isolated critic diagnoses failed trajectories under strict information isolation, and the executor edits specific skill files through a restricted tool interface. Across MobileWorld, AndroidWorld, and OSWorld, three mainstream GUI benchmarks spanning mobile and desktop platforms, EvoSkill-GUI consistently improves multiple base models without any training, with maximum gains of $+16.2%$, $+6.0%$, and $+10.5%$ respectively, and evolved skill libraries continue to benefit related tasks rather than being rebuilt from scratch. Our code is available at https://github.com/ZJU-REAL/EvoSkill-GUI.
A large language model (LLM) agent can follow more graph paths without acquiring more independent evidence. GraphEcho tests whether agents mistake these repeated encounters for additional corroboration. The benchmark varies path counts and evidential origins while holding evidence content fixed, and evaluates both judgments and active exploration. Controlled synthetic experiments reveal model-dependent judgment shifts, but redundant supporting paths increase the share of repeated walks across all evaluated frozen agents. Provenance-aware post-training (PAPT) reduces revisits and improves synthetic accuracy, yet covers fewer distinct sources. On scientific claims, it continues to reduce repetition while accuracy declines. These findings expose a gap between efficient exploration and effective evidence use: an agent can learn to stop repeating itself while overlooking information it needs. GraphEcho provides a controlled way to evaluate both what graph agents conclude and whether their exploration reaches distinct evidential sources.
展开完整摘要收起摘要↓
A large language model (LLM) agent can follow more graph paths without acquiring more independent evidence. GraphEcho tests whether agents mistake these repeated encounters for additional corroboration. The benchmark varies path counts and evidential origins while holding evidence content fixed, and evaluates both judgments and active exploration. Controlled synthetic experiments reveal model-dependent judgment shifts, but redundant supporting paths increase the share of repeated walks across all evaluated frozen agents. Provenance-aware post-training (PAPT) reduces revisits and improves synthetic accuracy, yet covers fewer distinct sources. On scientific claims, it continues to reduce repetition while accuracy declines. These findings expose a gap between efficient exploration and effective evidence use: an agent can learn to stop repeating itself while overlooking information it needs. GraphEcho provides a controlled way to evaluate both what graph agents conclude and whether their exploration reaches distinct evidential sources.