Large Language Model Agents

Accurate crop yield forecasting in commercial soft fruit production is constrained by the data available in typical commercial farm records, which lack the sensor networks, satellite imagery, and high-resolution meteorological inputs that most state-of-the-art approaches assume. We propose a structured LLM agent framework that performs post-hoc correction of existing model predictions, encoding agricultural domain knowledge across tools for phase detection, bias learning, and range validation. Evaluated on a proprietary strawberry yield dataset and a public USDA corn harvest dataset, agent refinement of XGBoost reduced MAE by 20% and MASE by 56% on strawberry, with consistent improvements across Moirai2 (MAE 24%, MASE 22%) and Random Forest (MAE 28%, MASE 66%) baselines. Using Llama 3.1 8B as the agent produced the strongest corrections across all configurations; LLaVA 13B showed inconsistent gains, highlighting sensitivity to the choice of refinement model.

Prompt specifications for multi-agent large language model (LLM) systems carry data contracts and integration logic across interdependent files but are rarely subjected to structured-inspection rigor. We report a single-system case study of iterative, agent-driven auditing applied to AEGIS (Autonomous Engineering Governance and Intelligence System), a seven-lane production pipeline whose 7152-line specification surface was audited across nine rounds, surfacing 51 consistency defects (per-round counts of 15, 8, 12, 2, 8, 1, 4, 1, 0). We present a seven-category post hoc taxonomy with explicit coding rules, non-monotonic convergence consistent with cascading edits and audit-scope expansion, and a locked audit protocol. We further report two partial replications on a public synthetic mini-specification: a cross-LLM panel of four frontier vendors (OpenAI, Anthropic, Google, xAI; 12 traces; multi-vendor union detects all five seeded defects) and an inter-rater reliability check on a stratified subsample (Cohen's $κ$ = 0.80 on category, 0.46 on severity). The full reproducibility bundle accompanies the submission.

Skill libraries enable large language model agents to reuse experience from past interactions, but most existing libraries store skills as isolated entries and retrieve them only by semantic similarity. This leads to two key challenges for compositional tasks. Firstly, an agent must identify not only relevant skills but also how they depend on and build upon each other. Secondly, it also makes library maintenance difficult, since the system lacks structural cues for deciding when skills should be merged, split, or removed. We propose SKILLGRAPH, a framework that represents reusable skills as nodes in a directed graph, with typed edges encoding prerequisite, enhancement, and co-occurrence relations. Given a new task, SKILLGRAPH retrieves not just individual skills, but an ordered skill subgraph that can guide multi-step decision making. The graph is continuously updated from agent trajectories and reinforcement learning feedback, allowing both the skill library and the agent policy to improve together. Experiments on ALFWorld, WebShop, and seven search-augmented QA tasks show that SKILLGRAPH achieves state-of-the-art performance against memory-augmented RL methods, with especially large gains on complex tasks that require composing multiple skills.

Reusable skills are becoming a common interface for extending large language model agents, packaging procedural guidance with access to files, tools, memory, and execution environments. However, this modularity introduces attack surfaces that are largely missed by existing safety evaluations: even when the user request is benign, unsafe influence may reside in skill guidance, local artifacts, or execution-environment files that steer the agent toward unsafe actions. We present SkillSafetyBench, a runnable benchmark for evaluating such skill-mediated safety failures. SkillSafetyBench includes 155 adversarial cases across 47 tasks, 6 risk domains, and 30 safety categories, each evaluated with a case-specific rule-based verifier. Experiments with multiple CLI agents and model backends show that non-user attacks can consistently induce unsafe behavior, with distinct failure patterns across domains, attack methods, and scaffold-model pairings. Our findings suggest that agent safety depends not only on model-level alignment, but also on how agents interpret skills, trust workflow context, and act through executable environments.

Reinforcement learning has become a widely used post-training approach for LLM agents, where training commonly relies on outcome-level rewards that provide only coarse supervision. While finer-grained credit assignment is promising for effective policy updates, obtaining reliable local credit and assigning it to the right parts of the long-horizon trajectory remains an open challenge. In this paper, we propose Granularity-adaptivE Advantage Reweighting (GEAR), an adaptive-granularity credit assignment framework that reshapes the trajectory-level GRPO advantage using token- and segment-level signals derived from self-distillation. GEAR compares an on-policy student with a ground-truth-conditioned teacher to obtain a reference-guided divergence signal for identifying adaptive segment boundaries and modulating local advantage weights. This divergence often spikes at the onset of a semantic deviation, while later tokens in the same autoregressive continuation may return to low divergence. GEAR therefore treats such spikes as anchors for adaptive credit regions: where the student remains aligned with the teacher, token-level resolution is preserved; where it departs, GEAR groups the corresponding continuation into an adaptive segment and uses the divergence at the departure point to modulate the segment' s advantage. Experiments across eight mathematical reasoning and agentic tool-use benchmarks with Qwen3 4B and 8B models show that GEAR consistently outperforms standard GRPO, self-distillation-only baselines, and token- or turn-level credit-assignment methods. The gains are especially strong on benchmarks with lower GRPO baseline accuracy, reaching up to around 20% over GRPO, suggesting that the proposed adaptive reweighting scheme is especially useful in more challenging long-horizon settings.

Recent work on LLM-driven autonomous penetration testing reports promising results, but existing systems often combine many architectural, prompting, and tool-integration choices, making it difficult to tell what is gained over a simple agent scaffold. We present cochise, a 597 LOC Python reference harness for autonomous penetration-testing experiments. Cochise connects an LLM-driven agent to a Linux execution host over SSH and supports controlled target environments reachable from that jump host. The prototype implements a separated Planner--Executor architecture in which long-term state is maintained outside the LLM context, while a ReAct-style executor issues commands over SSH and self-corrects based on command outputs. The scenario prompt can be adapted to different target environments. To demonstrate the efficacy of our minimal harness, we evaluate it against a live third-party testbed called Game of Active Directory (GOAD). Alongside the harness, we release replay and analysis tools: (i) cochise-replay for offline visualization of captured runs, (ii) cochise-analyze-alogs and cochise-analyze-graphs for cost, token, duration, and compromise analysis, and (iii) a corpus of JSON trajectory logs from GOAD runs, allowing researchers to study agent behavior without provisioning the 48--64 GB RAM / 190 GB storage testbed themselves. Cochise is intended not as a state-of-the-art pen-testing agent, but as reusable experimental infrastructure for comparing models, agent architectures, and penetration-testing traces.

Operational disaster response goes beyond damage assessment, requiring responders to integrate multi-sensor signals, reason over road networks, populations and key facilities, plan evacuations, and produce actionable reports. However, prior work largely isolates remote-sensing perception or evaluates generic tool use, leaving the end-to-end workflows of emergency operations underexplored. In this paper, we introduce Disaster Operational Response Agent benchmark (DORA), the first agentic benchmark for end-to-end disaster response: 515 expert-authored tasks across 45 real-world disaster events spanning 10 types, paired with expert-verified, replayable gold trajectories totaling 3,500 tool-call steps. Tasks span five dimensions that cover the operational disaster-response pipeline: disaster perception, spatial relational analysis, rescue and evacuation planning, temporal evolution reasoning, and multi-modal report synthesis. Agents compose calls from a 108-tool MCP library over heterogeneous geospatial data: optical, SAR, and multi-spectral imagery across single-, bi-, and multi-temporal sequences (0.015-10m GSD), complemented by elevation and social vector layers. We comprehensively evaluate 13 frontier LLMs on our benchmark, revealing three persistent challenges: 1) disaster-domain grounding exposes unique failure modes (damage-semantic grounding, sensor-modality mismatch, and disaster-pipeline composition); 2) agents are doubly bottlenecked by tool selection and argument grounding, where gold tool-order hints improve accuracy by only 1.08-4.40%, and alternative scaffolds yield at most a 3.24% gain; 3) compositional fragility scales with trajectory length, the agent-to-gold gap widening from 7% to 56% on long pipelines. DORA establishes a rigorous testbed for operationally reliable disaster-response agents.

When an LLM-based embodied agent fails at a household task, the culprit could be misidentified objects, forgotten sub-goals, or poor action sequencing -- yet existing benchmarks report only a single success rate, making it impossible to tell which cognitive module is responsible. We present PRISM, a diagnostic benchmark that reframes this problem: rather than asking only \textit{did the agent succeed?}, PRISM asks \textit{which capability is most likely responsible for failure?} Built on five photorealistic multi-room apartments (4--8 rooms each), PRISM structures 300 human-verified tasks into three capability tiers -- \textit{Basic Ability}, \textit{Reasoning Ability}, and \textit{Long-horizon Ability} -- that isolate perception-to-action grounding, implicit intent resolution, and sustained multi-step coordination respectively. PRISM exposes an agent-agnostic executable action API that allows arbitrary agents: LLM agents, VLM agents, symbolic planners, RL policies, and hybrid systems, to be evaluated end-to-end under the same benchmark protocol. To support deeper diagnosis, optional probes for perception, memory, and planning can be adopted, replaced, or bypassed entirely, enabling controlled component-level analysis when desired. Experiments on seven contemporary LLMs establish a clear hierarchy: explicit spatial grounding is not the dominant failure source under oracle perception, implicit intent resolution is a significant bottleneck for all model families, and long-horizon coordination exposes a stark capability cliff -- lightweight models collapse to as low as 20.0% success while simultaneously consuming more tokens than their frontier counterparts, a signature of compensatory over-reasoning rather than genuine planning capability. Project page: \href{https://sj-li.com/PROJ/PRISM}{link}.

Large language models (LLMs) are increasingly deployed on long-horizon tasks in partially observable environments, where they must act while inferring and tracking a complex environment state over many steps. This leads to two challenges: partial observability requires maintaining uncertainty over unobserved world attributes, and long interaction history causes context to grow without bound, diluting task-relevant information. A principled solution to both challenges is a belief state: a posterior distribution over environment states given past observations and actions, which compactly encodes history for decision making regardless of episode length. In LLM agents, however, the open-ended nature of text makes it unclear how to represent such a distribution. Therefore, we introduce Agent-BRACE: Agent Belief state Representation via Abstraction and Confidence Estimation, a method that decouples an LLM agent into a belief state model and a policy model, jointly optimized via reinforcement learning. The belief state model produces a structured approximation of the belief distribution: a set of atomic natural language claims about the environment, each annotated with an ordinal verbalized certainty label ranging from certain to unknown. The policy model conditions on this compact, structured approximate belief rather than the full history, learning to select actions under explicit uncertainty. Across long-horizon, partially observable embodied language environments, Agent-BRACE achieves an average absolute improvement of +14.5% (Qwen2.5-3B-Instruct) and +5.3% (Qwen3-4B-Instruct), outperforming strong RL baselines while maintaining a near-constant context window independent of episode length. Further analysis shows that the learned belief becomes increasingly calibrated over the course of an episode as evidence accumulates.

Modern gravitational wave astronomy relies on modeling tasks that often require months of graduate-level effort, including building fast waveform surrogates from expensive numerical relativity simulations, modeling orbital dynamics of black holes, fitting merger remnant properties and constructing template banks. These problems demand extreme precision to support detection and parameter inference, with state-of-the-art models achieving $\lesssim 10^{-4}$ relative error. We study whether state-of-the-art LLM coding agents can perform such end-to-end scientific modeling, where success requires constructing models with stringent accuracy criteria and reasoning about physical systems. We introduce gwBenchmarks, a suite of eight tasks grounded in gravitational wave analytic calculations and numerical simulations collectively representing over $10^8$ core-hours of compute. The tasks span interpolation, regression, and high-dimensional time-series modeling, requiring a combination of numerical methods, machine learning, and physics-informed approaches. In preliminary experiments, agents frequently relied on proxy metrics, partial evaluation, or fabricated results to spuriously complete tasks. We therefore implement an external pre-defined framework to gauge agent progress. Evaluating twelve coding agents, we find no consistent winner. On the easiest task, multiple agents converge to the same cubic spline solution, with one rediscovering a coordinate transformation widely used in the literature. On harder tasks like analytic waveform modeling, all agents fall 1-2 orders of magnitude short of domain requirements and exhibit systematic failures, including metric misuse, constraint violations, and result fabrication. Our code, data, and website are publicly available.

Large language model (LLM)-based agents frequently generate seemingly coherent plans that fail upon execution due to infeasible actions, constraint violations, and compounding errors over extended horizons. PIVOT (Plan-Inspect-eVOlve Trajectories) addresses this plan-execution misalignment through a self-supervised framework that treats trajectories as optimizable objects iteratively refined via environment interaction. The framework comprises four stages: PLAN generates candidate trajectories; INSPECT executes them and computes structured losses with textual gradients encoding plan-execution discrepancies; EVOLVE applies these signals to produce improved trajectories; and VERIFY performs a final global check against task constraints. A monotonic acceptance process ensures a non-decreasing solution quality. Empirical evaluations on DeepPlanning and GAIA demonstrate state-of-the-art performance: with human-in-the-loop (HITL) feedback, PIVOT establishes a strong upper bound up to 94% relative improvement in constraint satisfaction, while its fully autonomous variant retains substantial gains, showing that the core trajectory-refinement mechanism remains effective without external supervision. At the same time, PIVOT remains computationally efficient, requiring up to 3x to 5x fewer tokens than competing refinement methods. These findings establish that (self- or human-supervised) feedback-based trajectory optimization is a principled methodology for mitigating plan-execution gaps in autonomous agent systems.

Large language model agents interleave reasoning, action selection, and observation to solve sequential decision-making tasks. In deployed settings where agents repeatedly handle related multi-step tasks, small action-selection errors can accumulate into wasted tool calls, latency, and reduced reliability. Despite this need for deployment-time improvement, existing inference-time adaptation methods for LLM agents mainly rely on prompting or retrieval, which influence behavior indirectly through context manipulation. For ReAct-style agents, such approaches do not expose an explicit decision layer that can score candidate actions, represent uncertainty, or be updated online from action-level feedback. As a result, they provide limited support for trackable, fine-grained, and uncertainty-aware adaptation during deployment. We propose OLIVIA, an inference-time action adaptation framework for ReAct-style agents. OLIVIA models the LLM's final action-selection layer as a contextual linear bandit over candidate actions, with frozen hidden states as decision contexts. This choice is particularly suitable for deployment because it adapts behavior directly at the action-selection interface, preserves the underlying reasoning process, and provides explicit uncertainty estimates and lightweight online updates from action-level feedback. With upper-confidence-bound exploration, OLIVIA improves the policy sample-efficiently with minimal computational overhead. We instantiate OLIVIA on four benchmarks and show that it consistently improves task performance over static ReAct and prompt-based inference-time baselines. Our results suggest that explicit online decision layers provide an effective alternative to purely prompt- or retrieval-based adaptation for LLM agents during deployment.

Current LLM agents are proficient at calling isolated APIs but struggle with the "last mile" of commercial software automation. In real-world scenarios, tools are not independent; they are atomic, interdependent, and prone to environmental noise. We introduce $\textbf{ComplexMCP}$, a benchmark designed to evaluate agents in these rigorous conditions. Built on the Model Context Protocol (MCP), $\textbf{ComplexMCP}$ provides over 300 meticulously tested tools derived from 7 stateful sandboxes, ranging from office suites to financial systems. Unlike existing datasets, our benchmark utilizes a seed-driven architecture to simulate dynamic environment states and unpredictable API failures, ensuring a deterministic yet diverse evaluation. We evaluate various LLMs across full-context and RAG paradigms, revealing a stark performance gap: even top-tier models fail to exceed a 60% success rate, far trailing human performance 90%. Granular trajectory analysis identifies three fundamental bottlenecks: (1) $\textbf{tool retrieval saturation}$ as action spaces scale; (2) $\textbf{over-confidence}$, where agents skip essential environment verifications; and (3) $\textbf{strategic defeatism}$, a tendency to rationalize failure rather than pursuing recovery. These findings underscore the insufficiency of current agents for interdependent workflows, positioning $\textbf{ComplexMCP}$ as a critical testbed for the next generation of resilient autonomous systems.

The rapid proliferation of LLM-based autonomous agents in real operating system environments introduces a new category of safety risk beyond content safety: behavior jailbreak, where an adversary induces an agent to execute dangerous OS-level operations with irreversible consequences. Existing benchmarks either evaluate safety at the semantic layer alone, missing physical-layer harms, or fail to isolate test cases, letting earlier runs contaminate later ones. We present LITMUS (LLM-agents In-OS Testing for Measuring Unsafe Subversion), a benchmark addressing both gaps via a semantic-physical dual verification mechanism and OS-level state rollback. LITMUS comprises 819 high-risk test cases organized into one harmful seed subset and six attack-extended subsets covering three adversarial paradigms (jailbreak speaking, skill injection, and entity wrapping), plus a fully automated multi-agent evaluation framework judging behavior at both conversational and OS-level physical layers. Evaluation across frontier agents reveals three findings: (1) current agents lack effective safety awareness, with strong models (e.g., Claude Sonnet 4.6) still executing 40.64% of high-risk operations; (2) agents exhibit pervasive Execution Hallucination (EH), verbally refusing a request while the dangerous operation has already completed at the system level, invisible to every prior semantic-only framework; and (3) skill injection and entity wrapping attacks achieve high success rates, exposing pronounced agent vulnerabilities. LITMUS provides the first standardized platform for reproducible, physically grounded behavioral safety evaluation of LLM agents in real OS environments.

Multi-agent LLM systems introduce a security risk in which sensitive information accessed by one agent can propagate through shared context and reappear in downstream outputs, even without explicit adversarial intent. We formalise this phenomenon as propagation amplification, where leakage risk increases across agent boundaries as sensitive content is repeatedly exposed to downstream generators. Existing defences, including prompt-based safeguards, static pattern matching, and LLM-as-judge filtering, are not designed for this setting: they either operate after generation, rely primarily on surface-form patterns, or add substantial latency without modelling the generation process itself. To resolve these issues, we propose PRISM, a real-time defence that treats credential leakage as a sequential risk accumulation problem during generation. At each decoding step, PRISM combines 16 signals spanning lexical, structural, information-theoretic, behavioural, and contextual features into a calibrated risk score, enabling per-token intervention through green, yellow, and red risk zones. Our central observation is that credential reproduction is often preceded by a measurable shift in generation dynamics, characterised by entropy collapse and increasing logit concentration. When combined with text-structural cues such as identifier-pattern detection, these temporal signals provide an early warning of leakage before a secret is fully reconstructed. Across a 2,000-task adversarial benchmark covering 13 attack categories and three pressure levels in a heterogeneous four-agent pipeline, PRISM achieves F1 = 0.832 with precision = 1.000 and recall = 0.712, while producing no observed leakage on our benchmark (0.0% task-level leak rate) and preserving output utility of 0.893. It substantially outperforms the strongest baseline, Span Tagger, which achieves F1 = 0.719 with a 15.0% task-level leak rate.

We investigate the emergent collective dynamics of LLM-based multi-agent systems on a 2D square lattice and present a model-agnostic statistical-physics method to disentangle social conformity from intrinsic bias, compute critical exponents, and probe the collective behavior and possible phase transitions of multi-agent systems. In our framework, each node of an $L\!\times\!L$ lattice hosts an identical LLM agent holding a binary state ($+1$/$-1$, mapped to yes/no) and updating it by querying the model conditioned on the four nearest-neighbor states. The sampler temperature $T$ serves as the sole control parameter. Across three open-weight models (llama3.1:8b, phi4-mini:3.8b, mistral:7b), we measure magnetization and susceptibility under a global-flip protocol designed to probe $\mathbb{Z}_2$ symmetry. All models display temperature-driven order-disorder crossovers and susceptibility peaks; finite-size scaling on even-$L$ lattices yields effective exponents $γ/ν$ whose values are model-dependent, close to but incompatible with the 2D Ising universality class ($γ/ν=7/4$). Our method enables the extraction of effective $β$-weighted couplings $\tilde{J}(T)$ and fields $\tilde{h}(T)$, which serve as a measure of social conformity and intrinsic bias. In the models we analyzed, we found that collective alignment is dominated by an intrinsic bias ($\tilde{h}\gg\tilde{J}$) rather than by cooperative neighbor coupling, producing field-driven crossovers instead of genuine phase transitions. These effective parameters vary qualitatively across models, providing compact collective-behavior fingerprints for LLM agents and a quantitative diagnostic for the reliability of multi-agent consensus and collective alignment.

Legal information in India remains largely inaccessible due to the complexity of legal language and the sheer volume of legal documentation involved in research and case analysis. This paper presents NyayaAI, an AI-powered legal assistant that automates and simplifies legal workflows for lawyers, law students, and general users. The system combines Large Language Models with a Retrieval-Augmented Generation pipeline grounded in a curated Indian legal knowledge base comprising constitutional provisions, statutes, case laws, and judicial precedents. A multi-agent architecture orchestrated through the Mastra TypeScript framework coordinates a main agent with specialized sub-agents handling legal research, document summarization, case law retrieval, and drafting assistance. A compliance module validates all responses before delivery. Domain classification achieved 70% precision across test samples, with RAG retrieval precision at 74% and overall response accuracy at 72%, demonstrating that structured multi-agent LLM systems can meaningfully improve legal accessibility and workflow efficiency. The code\footnote{https://github.com/B97784/NyayaAI} is made publicly available for the benefit of the research community.

Agent-based modeling (ABM) has long been used in economics to study human behavior, and large language model (LLM) agents now enable new forms of social and economic simulation. While prior work has discovered strategic deception by LLM agents in financial trading and auction markets, e-commerce remains underexplored despite its distinctive information asymmetry: sellers privately observe product quality, whereas buyers rely on advertised claims and reputation signals. We introduce TruthMarketTwin, a controlled simulation framework for studying LLM-agent behavior in e-commerce markets. The framework is one of the first to model bilateral trade under asymmetric information sharing, where agents make strategic listing, purchasing, rating, and recourse-related decisions to optimize seller profit and buyer utility. We find that LLM agents released into traditional markets autonomously exploit weaknesses in reputation-based governance, while warrant enforcement reduces deception and reshapes strategic reasoning. Our results position LLM-agent simulation as a tool for studying institution-governed autonomous markets.

LLM-based agents act through sequences of executable decisions, but their trajectories provide little evidence of which agent or policy produced them, making provenance, ownership, and unauthorized reuse difficult to establish from observed behavior alone. This motivates watermarking signals embedded directly into agent behavior rather than only into generated text, since text watermarking cannot capture the action-level decisions that define agent execution. Recent agent watermarking methods address this gap by moving the watermark from generated text to behavioral choices. However, by treating each action step as an independent trial, they overlook trajectory structure and become fragile when trajectories are perturbed, truncated, or observed without reliable alignment. We propose SeqWM, a sequential behavioral watermarking framework that embeds signals into history-conditioned transition patterns and verifies trajectories position-agnostically against random-key baselines. Experiments across diverse agent benchmarks and LLM backbones show that SeqWM consistently achieves reliable detection while preserving agent utility, and remains robust under trajectory corruption where round-indexed behavioral watermarks collapse.

Production LLM coding agents drift over long sessions: they forget user-specified constraints, slip into mistakes the user already flagged, and confabulate prior agreements. White-box approaches such as persona vectors require model weights and so cannot be applied to closed APIs (Claude, GPT-4) that most users actually interact with. We present Nautilus Compass, a black-box persona drift detector and agent memory layer for production coding agents. The method operates entirely at the prompt-text layer: cosine similarity between user prompts and behavioral anchor texts, aggregated by a weighted top-k mean using BGE-m3 embeddings. Compass is, to our knowledge, the only public agent memory layer (among Mem0, Letta, Cognee, Zep, MemOS, smrti verified May 2026) that does not call an LLM at index time to extract facts or build a graph; raw conversation text is embedded directly. The system ships as a Claude Code plugin, an MCP 2024-11-05 A2A server (Cursor, Cline, Hermes), a CLI, and a REST API on one daemon, with a Merkle-chained audit log for tamper-evident anchor updates. On a held-out test set built from real Claude Code session traces and labeled by an independent LLM judge, Compass reaches ROC AUC 0.83 for drift detection. The embedded retrieval pipeline scores 56.6% on LongMemEval-S v0.8 and 44.4% on EverMemBench-Dynamic (n=500), topping the four published EverMemBench Table 4 baselines. LongMemEval-S 56.6% is ~30 points below recent white-box leaders (90+%); we treat that as the architectural ceiling of the no-extraction design. End-to-end reproduction cost is $3.50 (~14x cheaper than GPT-4o-judged stacks). A paired cross-vendor behavior A/B accompanies these numbers as preliminary system-level evidence. Code, anchors, frozen test data, and audit-log tooling are MIT-licensed at github.com/chunxiaoxx/nautilus-compass.

Personal AI assistants are beginning to act as delegates with access to calendars, inboxes, and user preferences. Calendar scheduling makes the trust problem concrete: an assistant must coordinate with other assistants while deciding what to reveal about the person it represents. We introduce CalBench, a controlled benchmark for multi-agent calendar scheduling under private information. In each task, $N$ agents manage separate private calendars and schedule a stream of $M$ incoming meetings while minimizing disruption costs. Because no agent can inspect another agent's calendar, success requires language-mediated coordination rather than centralized planning. CalBench generates solvable scenarios with CP-SAT oracle solutions and decentralized non-LLM reference protocols, enabling evaluation of task success, excess cost, communication efficiency, burden fairness, and privacy leakage under matched information constraints. Across seven model families, we find that completion alone misses important failures: agents leave avoidable cost on the table, communication volume does not predict lower regret, and privacy-preserving silence can deprive teammates of cost information needed for fair burden allocation. CalBench provides a reproducible testbed for studying whether autonomous assistants can coordinate on behalf of users before deployment at scale.

We present swarm-attack, an open-source adversarial testing framework in which multiple lightweight LLM agents coordinate through shared memory, parallel exploration, and evolutionary optimization. Together, our results demonstrate that both safety bypass of frontier models and software vulnerability discovery, i.e., the capability class that motivated restricted release of Anthropic's Mythos Preview, are achievable at effectively zero cost using commodity hardware and openly available models. We report two experiments. In the first, five instances of a 1.2 billion parameter model conducted 225 jailbreak attacks each against GPT-4o and Claude Sonnet~4. Against GPT-4o, the swarm achieved an Effective Harm Rate of 45.8%, producing 49 critical-severity breaches; against Claude Sonnet-4, the Effective Harm Rate was 0% despite a 40% technical success rate. In the second experiment, the same models performed combined source code analysis and binary fuzzing against a vulnerable C application with 9 planted CWEs. With a hand-crafted exploit seed corpus, regex pattern detection, and AddressSanitizer-based crash classification, the pipeline recovers 9 of 9 vulnerabilities (100% recall) in approximately four minutes on a consumer MacBook. With those scaffold components disabled, the same model recovers 0 of 9 by crash verification and 2 of 9 by citation. The capability class that motivated restricted release of Anthropic's Mythos Preview is therefore reproducible at effectively zero cost; the important enabler is the system scaffold itself, which compensates for the limited reasoning capacity of small individual models.

Agentic large language models often rely on skills, reusable natural language procedures that guide planning, action, and tool use. In practice, skills are typically improved through prompt engineering or by aligning the task LLM itself, which is costly, model-specific, and often infeasible for closed-source models. Skill optimization is not a one-step problem but a recurrent process with two coupled levels of credit assignment: a useful skill must improve rollout quality under current conditioning, while a useful revision must turn observed outcomes into a better skill for the next round. We propose Skill-R1, a reinforcement learning framework for instance-level recurrent skill optimization from verifiable rewards. Rather than updating the task LLM, Skill-R1 trains a lightweight skill generator that conditions on the task context, prior rollouts, and their verified outcomes to produce skills that steer a frozen task LLM. This preserves black-box compatibility with both open- and closed-source models while making adaptation substantially cheaper than model-level updates. Skill-R1 proceeds over multiple generations: at each step, the current skill induces rollouts whose verified outcomes are fed back to produce the next revision. To optimize this recurrent process, we introduce a bi-level group-relative policy optimization objective combining intra-generation and inter-generation advantages. The intra-generation term compares rollouts under shared skill conditioning, while the inter-generation term rewards revisions that improve behavior across successive generations. Together, these provide a principled objective for directional skill evolution rather than one-shot self-refinement. Empirically, Skill-R1 achieves consistent gains over no-skill baselines and standard GRPO across benchmarks with verifiable rewards, with particularly strong improvements on complex, multi-step tasks.

Recent advances in LLM agents enable systems that autonomously refine workflows, accumulate reusable skills, self-train their underlying models, and maintain persistent memory. However, we show that such self-evolution is often non-monotonic: adapting to new task distributions can progressively degrade previously acquired capabilities across all major evolution channels. We identify this phenomenon as \emph{capability erosion under self-evolution} and show that it consistently emerges across workflow, skill, model, and memory evolution. To mitigate this issue, we propose \emph{Capability-Preserving Evolution} (CPE), a general stabilization principle that constrains destructive capability drift during continual adaptation. Across all four evolution dimensions, CPE consistently improves retained capability stability while preserving adaptation performance. For example, in workflow evolution, CPE improves retained simple-task performance from 41.8% to 52.8% under GPT-5.1 optimization while simultaneously achieving stronger complex-task adaptation. Our findings suggest that stable long-horizon self-evolving agents require not only acquiring new capabilities, but also explicitly preserving previously learned ones during continual adaptation.

Tool-augmented LLM agents tend to call tools indiscriminately, even when the model can answer directly. Each unnecessary call wastes API fees and latency, yet no existing benchmark systematically studies when a tool call is actually needed. We propose When2Tool, a benchmark of 18 environments (15 single-hop, 3 multi-hop) spanning three categories of tool necessity -- computational scale, knowledge boundaries, and execution reliability -- each with controlled difficulty levels that create a clear decision boundary between tool-necessary and tool-unnecessary tasks. We evaluate two families of training-free baselines: Prompt-only (varying the prompt to discourage unnecessary calls) and Reason-then-Act (requiring the model to reason about tool necessity before acting). Both provide limited control: Prompt-only suppresses necessary calls alongside unnecessary ones, and Reason-then-Act still incurs a disproportionate accuracy cost on hard tasks. To understand why these baselines fail, we probe the models' hidden states and find that tool necessity is linearly decodable from the pre-generation representation with AUROC 0.89--0.96 across six models, substantially exceeding the model's own verbalized reasoning. This reveals that models already know when tools are needed, but fail to act on this knowledge during generation. Building on this finding, we propose Probe&Prefill, which uses a lightweight linear probe to read the hidden-state signal and prefills the model's response with a steering sentence. Across all models tested, Probe&Prefill reduces tool calls by 48% with only 1.7% accuracy loss, while the best baseline at comparable accuracy only reduces 6% of tool calls, or achieves a similar tool call reduction but incurs a 5$\times$ higher accuracy loss. Our code is available at https://github.com/Trustworthy-ML-Lab/when2tool

Large language model agents increasingly operate through an intermediate skill layer that mediates between user intent and concrete task execution. This layer is widely treated as an organizational abstraction, but we argue it is also a privilege boundary that current models routinely exceed. We present \textbf{FORTIS}, a benchmark that evaluates over-privilege in agent skills across two stages: whether a model selects the minimally sufficient skill from a large overlapping library, and whether it executes that skill without expanding into broader tools or actions than the skill permits. Across ten frontier models and three domains, we find that over-privileged behavior is the norm rather than the exception. Models consistently reach for higher-privilege skills and tools than the task requires, failing at both stages at rates that remain high even for the strongest available models. Failure is especially severe under the ordinary conditions of real user interaction: incomplete specification, convenience framing, and proximity to skill boundaries. None of these requires adversarial construction. The results indicate that the skill layer, far from containing agent behavior, is itself a primary source of privilege escalation in current systems.

Agents built on large language models (LLMs) rely on a range of reliability techniques, including retry, majority voting, and self-consistency, that have been developed in parallel rather than within a common analytical framework. We observe that an LLM sampled at temperature $T$ is a discrete stochastic channel $p(y \mid x)$ in the sense of Shannon's coding theory, and use this identity as the entry point for such a framework grounded in communication theory. Each of these techniques is a special case of one of six classical reliability operators: diversity combining, hybrid retransmission, iterative generator-critic decoding, rateless sampling, structured redundant verification, and difficulty-adaptive routing. Within the framework we give two closed-form results: a noise-variance threshold above which uniform averaging beats quality-weighted averaging, and a contractivity criterion for generator-critic refinement, consistent with a contractive-to-divergent transition we observe between 3B- and 14B-parameter models. We further introduce a cost-aware semantic-nearest-neighbor router whose single Lagrangian knob traverses the quality-cost frontier without retraining. Across six channel configurations spanning local and cloud models on 69 hard tasks, no fixed model-technique-budget choice dominates, motivating per-task allocation. On a 300-item hard split of MMLU, GSM8K, and HumanEval, our router occupies the full empirical Pareto frontier: at matched quality, its normalized cost is ${\approx}56$% lower than the strongest fixed technique; at matched normalized cost, it improves quality by ${\approx}7$% ($26$% over single-shot decoding). These results argue for consolidating these reliability techniques into a single tunable layer informed by channel coding.

As LLM agents evolve, tokens have emerged as the core economic primitives of Agentic AI. However, their exponential consumption introduces severe computational, collaborative, and security bottlenecks. Current surveys remain fragmented across system optimization, architecture design, and trust, lacking a unified framework to evaluate the fundamental trade-off between output quality and economic cost. To bridge this gap, this survey presents the first comprehensive survey of Token Economics. By unifying computer science and economics, we conceptualize tokens as production factors, exchange mediums, and units of account. We synthesize existing literature across a four-dimensional taxonomy: (1) Micro-level (Single Agent): Optimizing budget-constrained factor substitution via neoclassical firm theory. (2) Meso-level (Multi-Agent Systems): Minimizing collaboration friction using transaction cost and principal-agent theories. (3) Macro-level (Agent Ecosystems): Addressing congestion externalities and pricing via mechanism design. (4) Security: Internalizing adversarial threats as endogenous economic constraints. Finally, we outline frontier directions, including differentiable token budgets and dynamic markets, to lay the theoretical foundation for scalable next-generation agent systems.

Large language model (LLM) agents increasingly collaborate over peer-to-peer networks to improve their reliability. However, these same interactions can also become a source of vulnerability, as unreliable or Byzantine agents may sway neighboring agents toward incorrect conclusions and degrade overall system performance. Existing methods rely on leader-based coordination or self-reported confidence, both of which are susceptible to adversarial manipulation. We study decentralized LLM multi-agent systems (LLM-MAS) and propose Self-Anchored Consensus (SAC), a fully decentralized iterative filter-and-refine protocol in which agents iteratively exchange responses, locally evaluate and filter unreliable messages, and refine their own outputs. We present $(F{+}1)$-robustness conditions for the communication graph that ensure honest agents preserve and propagate reliable information despite Byzantine influence. Experiments on mathematical and commonsense reasoning benchmarks show that SAC effectively suppresses Byzantine influence and consistently improves performance across diverse communication topologies, whereas prior methods degrade under adversarial conditions.

In multi-agent systems (MAS), a single deceptive agent can nullify all gains of an agentic AI collective and evade deployed defenses. However, existing adversarial studies on MAS target only shallow tasks and do not consider adaptive adversaries, which evolve their strategies to evade the very detectors trained to catch them. To address that gap, we introduce GAMBIT, a benchmark with three evaluation modes and two independent scores for evaluating imposter detectors: the first two modes measure zero-shot detection under increasing distribution shift, and a third recalibration mode measures how quickly a detector adapts to novel attacks from just 20 labeled examples. The benchmark comes with a dataset of 27,804 labeled instances spanning 240 co-evolved imposter strategies. Our contributions are threefold: (1) Using chess as a substrate deep reasoning problem and Gemini 3.1 Pro for agents, we release GAMBIT and its dataset to evaluate imposter detectors under realistic constraints against a stealthy adaptive imposter; (2) We introduce an adaptive imposter agent based on an efficient evolutionary framework, generalizable beyond chess, that collapses collective task performance while remaining essentially undetectable (50.5% F1-score with a Gemini-based detector); (3) We show that zero-shot evaluation can be highly misleading for adaptive adversaries: two detectors with near-identical zero-shot scores differ by 8x on few-shot adaptation, while the meta-learned variant converges 20x faster, a gap only visible in the recalibration mode. Altogether, GAMBIT provides the first multi-agent benchmark where adversarial attacks and defenses co-evolve, with an imposter framework generalizable beyond our use case, and promising techniques for fast recalibration in a rapidly evolving adversarial system. Code and data: https://anonymous.4open.science/r/gambit.

Large Language Models (LLMs) have demonstrated remarkable capabilities in reasoning and tool use. However, the fundamental cognitive faculties essential for problem solving, including perception, reasoning, and memory, remain the stable core of intelligence. Unlike memorizing specific patterns, humans succeed in novel environments by applying these intrinsic faculties to adapt and optimize. Yet, whether LLMs possess this essential capacity, namely the ability to continuously refine solutions in response to dynamic environmental feedback, remains underexplored. To address this challenge, we introduce OPT-BENCH, a benchmark for evaluating self-improvement capabilities in large-scale search spaces. By combining 20 machine learning tasks with 10 classic NP-hard problems, OPT-BENCH provides a rigorous setting to assess whether agents can adapt through intrinsic self-reflection rather than rote tool application. We further propose OPT-Agent, a framework that emulates human-like cognitive adaptation. It operates through a general perception, memory, and reasoning loop, iteratively refining solutions based on environmental feedback. Through extensive experiments on 19 LLMs from 7 model families, including reasoning models, general models, and open-source models ranging from 3B to 235B parameters, we demonstrate that stronger models are more effective at leveraging feedback signals for self-improvement. However, this upper-bound adaptability remains fundamentally constrained by the models' base capacity, and even the most advanced LLMs still fall short of human expert performance.

Large Language Models (LLMs) are increasingly deployed as autonomous agents that execute tool-augmented, multi-step tasks, where latency is a critical factor for real-world applications. Yet an overlooked threat is Reasoning-Level Denial-of-Service (R-DoS), in which an attacker preserves task correctness but degrades availability by inflating an agent's reasoning depth or tool-use budget. We introduce OTora, the first unified, two-stage red-teaming framework for instantiating R-DoS attacks. Stage I optimizes an adversarial trigger that induces targeted tool invocations using insertion-aware scoring and dynamic target co-evolution, supporting both black-box and white-box settings. Stage II generates agent-aware reasoning payloads via an ICL-guided genetic search that amplifies overthinking while maintaining correct task outcomes. Across WebShop, Email, and OS agents built on multiple backbone models such as LLaMA-70B and GPT-OSS-120B, OTora achieves up to 10 times increases in reasoning tokens and order-of-magnitude latency slowdowns, all while preserving near-baseline task accuracy. Finally, we discuss mitigation strategies for detecting and constraining abnormal reasoning and latency spikes. The code is available at https://github.com/llm2409/OTora.

Large language model agents increasingly operate through environment-facing scaffolds that expose files, web pages, APIs, and logs. These observations influence tool use, state tracking, and action sequencing, yet their reliability and authority are often uncertain. Environmental grounding is therefore a systems-level problem involving context admission, evidence provenance, freshness checking, verification policy, action gating, and model reasoning. Existing agent benchmarks mainly evaluate task capability or specific attacks such as prompt injection and memory poisoning, but they under-specify a fundamental reliability question: whether agents remain grounded in the true environment state when observations are stale, incorrect, or malicious. We introduce EnvTrustBench, an agentic framework for benchmarking this failure mode. We define an evidence-grounding defect (EGD) as a behavioral failure in which an agent treats an environment-facing claim as sufficient evidence for action without resolving it against available current evidence, leading to a task-incorrect false path under the true environment state. Given a task scenario, EnvTrustBench generates the workspace, environment, agent-facing objective, and validation oracle, executes the evaluated agent, records its action-observation trajectory and final state, and applies the oracle to produce a verdict. Using 6 LLM backbones and 5 widely used scaffolds, we evaluate 55 generated cases across 11 task scenarios, with each scenario expanded through five feedback-guided generation iterations. Results show that EGDs consistently emerge across operational workflows, highlighting environmental grounding as a core agent reliability problem with important security implications.

Skills provide an effective mechanism for improving LLM agents on complex tasks, yet in existing agent frameworks, their creation, refinement, and selection are typically governed by external teachers, hand-designed rules, or auxiliary modules. As a result, skills remain external resources to be invoked, rather than capabilities that agents can develop, adapt, and internalize through experience. To endow LLM agents with autonomous skill mastery, we propose SkillMaster, a training framework that teaches agents to create new skills, refine existing skills, and select accumulated skills during task solving. This capability is achieved through three key designs. First, we train agents through trajectory-informed skill review, teaching agents to propose, update, or retain skills based on evidence from completed episodes. Second, each candidate skill edit is designed to be evaluated by its counterfactual utility on related probe tasks, providing a direct learning signal for training skill-editing decisions. Third, we introduce DualAdv-GRPO, which separately estimates advantages for task-solving actions and skill-editing decisions, stabilizing joint training across task solving and skill management. Experiments on ALFWorld and WebShop show that SkillMaster improves the overall success rate over state-of-the-art baselines by 8.8% and 9.3%, respectively, achieving the best performance among all compared methods. Further analysis reveals a marked shift in agent capability: agents trained with SkillMaster can identify skill failures, refine procedural knowledge from trajectory evidence, and transfer improvements to future tasks with limited skill-bank edits. Overall, SkillMaster moves LLM agents beyond mere skill use toward self-improving agents capable of developing, adapting, and applying their own skill repertoires.

Multi-agent large language model (LLM) systems often rely on a controller to coordinate a pool of heterogeneous models, yet existing controllers are typically limited to one-shot routing: they select a model once and return its output directly. Such routing-only designs provide no mechanism to critique intermediate drafts or support iterative refinement. To address this limitation, we propose a critique-and-routing controller that casts multi-agent coordination as a sequential decision problem. At each turn, the controller evaluates the current draft, decides whether to stop or continue, and, if needed, selects the next agent for further refinement. We formulate this process as a finite-horizon Markov Decision Process (MDP) with explicit agent-utilization constraints, design a composite reward for controller decisions across turns, and optimize the controller via policy gradients under a Lagrangian-relaxed objective. Extensive experiments across multiple heterogeneous multi-agent systems and seven reasoning benchmarks show that our method consistently outperforms state-of-the-art baselines and substantially narrows the gap to the strongest agent, while using it for fewer than 25% of total calls.

When an LLM agent fails a multi-step tool-augmented task and retries, the failed attempt typically remains in its context window -- contaminating the next attempt and elevating the per-step error rate beyond the base level. This context-contaminated restart phenomenon is widely observed in practice yet entirely lacks formal treatment. We introduce the Context-Contaminated Restart Model (CCRM): a chain of T tool-call steps, each failing with base rate epsilon_0; after any failed attempt, the subsequent attempt operates in contaminated context with elevated error rate epsilon_1 > epsilon_0. Under this model we derive five main results. (R1) An exact closed-form formula for P(succeed in at most K attempts). (R2) A cascade-overhead theorem giving the additional attempts Delta K incurred by contamination versus the clean-restart baseline. (R3) An optimal budget-allocation theorem identifying the pipeline depth T* that maximises success probability for a fixed total budget B=KT; we prove the closed form T* = sqrt(B * log(1/(1-epsilon_1)) / log(1/(1-epsilon_0))), with K*=B/T*. (R4) An information-theoretic lower bound via Le Cam's method showing K_CCRM is tight up to O(1). (R5) A clean-restart dominance theorem quantifying the exact benefit of context-clearing before retry. We validate CCRM on real SWE-bench Verified data: the IID model overestimates pass@3 by 17.4 percentage points (98.6% vs. 81.2%), while CCRM fits with error less than 0.001, implying a cascade ratio of epsilon_1/epsilon_0 = 7.1. Monte Carlo experiments confirm all theoretical predictions.

Current LLM agents lack principled mechanisms for managing persistent memory across long interaction horizons. We present a biologically-grounded memory architecture comprising six cognitive mechanisms: (1) sleep-phase consolidation, (2) interference-based forgetting, (3) engram maturation, (4) reconsolidation upon retrieval, (5) entity knowledge graphs, and (6) hybrid multi-cue retrieval. Each mechanism addresses a specific failure mode of naive memory accumulation. We introduce a synthetic calibration methodology that derives all pipeline thresholds without benchmark data exposure, eliminating a common source of evaluation leakage. We evaluate on two benchmarks. First, a VSCode issue-tracking dataset (13K issues, 120K events) where deduplication-based consolidation achieves 97.2% retention precision with 58% store reduction (+21.8 pp over baseline). Second, the LongMemEval personal-chat benchmark where we conduct the first streaming M-tier evaluation (475 sessions, ~540K unique turns). At a 200K-token context budget, our pipeline matches raw retrieval accuracy (70.1% vs. 71.2%, overlapping 95% CI) while exposing a tunable accuracy/store-size operating curve. At S-tier scale (50 sessions), dedup-based consolidation yields a +13.3 pp improvement in preference recall.

Persistent memory in LLM agents creates an attack surface that production safety classifiers do not observe: the payload enters via RAG retrieval and persists across sessions via tool-mediated memory. We evaluate six defenses across four architectural layers against delayed-trigger attacks on nine open-source models (5,040 runs, N=40 per condition). Five of six defenses fail: input-level filters never see the payload (it enters via RAG, not user input); retrieval-level classifiers observe it but cannot distinguish compliance-framed injection from legitimate policy; instruction-level hardening is overridden by the stored rule's compliance framing. Only tool-gating at the memory layer (Memory Sandbox) reduces ASR to 0% for eight of nine models, with zero utility cost. A reasoning model inverts this defense via goal-directed RAG fallback, a mechanism that replicates cross-family on Bedrock. A reasoning-mode ablation reveals a double dissociation: no single sandbox implementation is safe across both reasoning and non-reasoning model classes. We resolve this with a content-layer proof-of-concept (RATG), validated on non-reasoning models. A loaded-corpus frontier evaluation (21 models, 3 providers, N=40) overturns an initial empty-corpus screen showing 0/210 exfiltrations: that was a threat-model artifact, not model safety. Under realistic conditions, Gemini 3.1 Pro Preview exfiltrates at 95% ASR, GPT-5.1 regresses to 22.5% relative to GPT-5 (5%), and Anthropic blocks at the injection layer (0-17.5% storage, 0% ASR). Nearly all OpenAI and Gemini models store the rule at 100% regardless of execution resistance, creating supply-chain risk in shared-memory deployments. Defense effectiveness is determined by architectural layer and reasoning capability, not classifier quality.

Skill libraries have become a practical way for LLM agents to reuse procedural experience across tasks. However, existing systems typically treat skills as flat, single-resolution prompt blocks. This creates a tension between relevance and cost: injecting coarse skills can introduce irrelevant or misleading context, while rewriting entire skills is expensive and often unnecessary. We propose SkillLens, a hierarchical skill-evolution framework that organizes skills into a four-layer graph of policies, strategies, procedures, and primitives, and retrieves them at mixed granularity. Given a task, SkillLens first retrieves semantically relevant skill seeds, expands them through degree-corrected random walk over the skill graph, and then uses a verifier to decide whether each visited unit should be accepted, decomposed, rewritten, or skipped. This enables the agent to reuse compatible subskills directly while adapting only locally mismatched components. To improve the system over time, SkillLens further refines multi-granularity skills and verifier in order to improve its routing decisions. We provide theoretical analysis showing that mixed-granularity adaptation incurs sublinear cost under sparse mismatch assumptions and that the evolutionary update rule monotonically improves the validation objective until a local optimum. Across MuLocbench and ALFWorld, SkillLens consistently improves over strong skill-based baselines, achieving up to a 6.31 percentage-point Acc@1 gain for bug localization and raising agent success rate from 45.00% to 51.31%.

Context window expansion is often treated as a straightforward capability upgrade for LLMs, but we find it systematically fails in multi-agent social dilemmas. Across 7 LLMs and 4 games over 500 rounds, expanding accessible history degrades cooperation in 18 of 28 model--game settings, a pattern we term the memory curse. We isolate the underlying mechanism through three analyses. First, lexical analysis of 378,000 reasoning traces associates this breakdown with eroding forward-looking intent rather than rising paranoia. We validate this using targeted fine-tuning as a cognitive probe: a LoRA adapter trained exclusively on forward-looking traces mitigates the decay and transfers zero-shot to distinct games. Second, memory sanitization holds prompt length fixed while replacing visible history with synthetic cooperative records, which restores cooperation substantially, proving the trigger is memory content, not length alone. Finally, ablating explicit Chain-of-Thought reasoning often reduces the collapse, showing that deliberation paradoxically amplifies the memory curse. Together, these results recast memory as an active determinant of multi-agent behavior: longer recall can either destabilize or support cooperation depending on the reasoning patterns it elicits.

As LLM-based agents increasingly rely on external tools, it is important to evaluate their ability to sustain tool-grounded reasoning beyond familiar workflows and short-range interactions. We introduce AgentEscapeBench, an escape-room-style benchmark that tests whether agents can infer, execute, and revise novel tool-use procedures under explicit long-range dependency constraints. Each task defines a directed acyclic dependency graph over tools and items, requiring agents to invoke real external functions, track hidden state revealed incrementally, propagate intermediate results, and submit a deterministically verifiable final answer. AgentEscapeBench includes 270 instances across five difficulty tiers and supports fully automated evaluation. Experiments with sixteen LLM agents and human participants show that performance drops sharply as dependency depth increases: humans decline from 98.3% success at difficulty-5 to 80.0% at difficulty-25, while the best model drops from 90.0% to 60.0%. Trajectory analysis attributes model failures mainly to breakdowns in long-range state tracking, clue adherence, and intermediate-result propagation. These findings suggest that current agents can often handle local tool use but still struggle with deep contextual dependencies. We hope AgentEscapeBench can serve as a diagnostic testbed for measuring current agent capabilities and informing future training efforts toward more robust general-purpose reasoning, action, and adaptation.

Relational learning is a challenging problem that has motivated a wide range of approaches, including graph-based models (e.g., graph neural networks, graph transformers), tabular methods (e.g., tabular foundation models), and sequence-based approaches (e.g., large language models), each with its own advantages and limitations. We propose RelAgent, an LLM-based autonomous data scientist for relational learning, which operates in two phases. In the search phase, an LLM agent uses database, validation, and evaluation workspace tools to construct SQL feature programs and select a predictive model. In the inference phase, the resulting program is executed without further LLM calls. The final predictor consists of SQL queries and a classical model, enabling fast, deterministic, and intrinsically interpretable predictions: features are human-readable queries, and predictions depend only on the resulting query-defined feature map, enabling scalable deployment using standard database systems.

Large language models (LLMs) are increasingly deployed as autonomous agents in offensive cybersecurity. In this paper, we reveal an interesting phenomenon: different agents exhibit distinct attack patterns. Specifically, each agent exhibits an attack-selection bias, disproportionately concentrating its efforts on a narrow subset of attack families regardless of prompt variations. To systematically quantify this behavior, we introduce CyBiasBench, a comprehensive 630-session benchmark that evaluates five agents on three targets and four prompt conditions with ten attack families. We identify explicit bias across agents, with different dominant attack families and varying entropy levels in their attack-family allocation distributions. Such bias is better characterized as a trait of the agents, rather than a factor associated with the attack success rate. Furthermore, our experiments reveal a bias momentum effect, where agents resist explicit steering toward attack families that conflict with their bias. This forced distribution shift does not yield measurable improvements in attack performance. To ensure reproducibility and facilitate future research, we release an interactive result dashboard at https://trustworthyai.co.kr/CyBiasBench/ and a reproducibility artifact with aggregated session-level statistics and full evaluation scripts at https://github.com/Harry24k/CyBiasBench.

Tool-integrated reasoning (TIR) is difficult to scale to small language models due to instability in long-horizon tool interactions and limited model capacity. While reinforcement learning methods like group relative policy optimization provide only sparse outcome-level rewards. Recently, on-policy distillation (OPD) has gained popularity by supplying dense token-level supervision from a teacher on student-generated trajectories. However, our experiments indicate that applying OPD to TIR leads to a critical failure mode: erroneous tool calls tend to cascade across subsequent reasoning steps, progressively amplifying student-teacher divergence and rendering the teacher's token-level supervision increasingly unreliable. To address this, we propose SOD, a step-wise on-policy distillation framework for small language model agents, which adaptively reweights distillation strength at each step based on step-level divergence. Therefore, SOD can attenuate potentially misleading teacher signals in high-divergence regions while preserving dense guidance in well-aligned states. Experiments on challenging math, science, and code benchmarks show that SOD achieves up to 20.86% improvement over the second-best baseline. Notably, our 0.6B student achieves 26.13% on AIME 2025, demonstrating effective transfer of agentic reasoning to lightweight models. Our code is available at https://github.com/YoungZ365/SOD.

Reusable agent skills are emerging as a service-oriented capability layer for Large Language Model (LLM) agents. Unlike plain retrieval items, a skill exposes functional capabilities, input-output assumptions, tool dependencies, context cost, and risk metadata. Selecting skills is particularly challenging for small LLM agents, which can load only a few capability units under restricted context, tool availability, and risk tolerance. Existing fixed Top-k methods rank skills by textual relevance and overlook requirement satisfaction, deliverability, and operational constraints. We present SkillSelect-Serve, a QoS-aware, budget-constrained Skill Service recommendation framework. Raw skills are profiled as structured Skill Services, the task is converted into a structured requirement object, and candidates discovered from a large-scale registry are ranked by a calibrated task-conditioned suitability estimator and packed by a constrained projection enforcing token-budget, aggregated-risk, and tool-availability constraints, using only deployment-observable features. On a registry of 35,353 skills with pooled multi-positive relevance judgments verified by two independent assessors, the unconstrained top-5 recommendation fits a realistic 4,000-token context for only 9.1% of tasks; the constrained projection restores 100% deliverability at a cost of only 1.14 points of hit rate, outperforming retrieve-and-rerank, budget truncation, and diversity-based selection under identical budgets. The same mechanism halves delivered risk exposure and eliminates the 44-81% tool-violation rates of tool-agnostic recommendation. At an identical three-service budget, hit rate improves from 0.8864 to 0.9091 over fixed Top-3 retrieval. The results support managing reusable agent skills as discoverable, comparable, and constraint-aware service units instead of plain retrievable documents.

Biological laboratory automation can reduce repetitive manual work and improve reproducibility, but reliable embodied execution in wet-lab environments remains challenging. Protocols are often unstructured, labware is frequently transparent or reflective, and multi-step procedures require state-aware execution beyond one-shot instruction following. Existing robotic systems often rely on costly hardware, fixed workflows, dedicated instruments, or robotics-oriented interfaces. Here, we introduce BioProVLA-Agent, an affordable, protocol-driven, vision-enhanced embodied multi-agent system enabled by Vision-Language-Action (VLA) models for biological manipulation. The system uses protocols as the task interface and integrates protocol parsing, visual state verification, and embodied execution in a closed-loop workflow. A Tailored LLM Protocol Agent converts protocols into verifiable subtasks; a VLM-RAG Verification Agent assesses readiness and completion using observations, robot states, retrieved knowledge, and success/failure examples; and a VLA Embodied Agent executes verified subtasks through a lightweight policy. To improve robustness under wet-lab visual perturbations, we develop AugSmolVLA, an online augmentation strategy targeting transparent labware, reflections, illumination shifts, and overexposure. We evaluate the system on a hierarchical benchmark covering 15 atomic tasks, 6 composite workflows, and 3 bimanual tasks, including tube loading, sorting, waste disposal, cap twisting, and liquid pouring. Across normal and high-exposure settings, AugSmolVLA improves execution stability over ACT, X-VLA, and the original SmolVLA, especially for precise placement, transparent-object manipulation, composite workflows, and visually degraded scenes. These results suggest a practical route toward accessible, protocol-centered, and verification-capable embodied AI for biological manipulation.

Accurately predicting opponents' behavior from interactions is a fundamental capability for large language model (LLM)-based agents in multi-agent and game-theoretic environments. Existing approaches often entangle opponent modeling with prediction, relying on implicit contextual reasoning and limiting adaptability in dynamic interactions. To this end, we propose Structured Opponent Modeling (SOM), a two-stage opponent modeling framework that distinctly separates opponent model construction and opponent prediction. At the construction stage, SOM employs a Structural Causal Model (SCM), a graph-based formalism for representing dependencies among variables, to capture directed links between opponents' observations and actions, yielding an explicit and structured opponent representation. At the prediction stage, the LLM performs structured reasoning along clear pathways derived from the SCM, improving both prediction accuracy and stability. Extensive experiments on diverse multi-agent benchmarks demonstrate that SOM consistently outperforms state-of-the-art LLM-based reasoning baselines, enabling more accurate and adaptable strategic decision-making in complex and dynamic multi-agent interactions.

Large language models (LLMs) are increasingly deployed in multi-agent systems where agents communicate in natural language to solve tasks jointly. A key capability in such systems is consensus formation, where agents iteratively exchange messages and update decisions to reach a shared outcome. However, most existing multi-agent LLM frameworks assume that all participating agents are aligned with the system objective. In practice, a malicious insider may participate as a legitimate member of the group while pursuing a hidden adversarial goal. In this work, we study insider manipulation in multi-agent LLM consensus systems. We formalize the problem as a sequential decision-making task in which a malicious agent seeks to delay or prevent agreement among benign agents. To make attack optimization tractable, we propose a world-model-based framework that learns surrogate dynamics over the latent behavioral states of benign agents and then trains an attacker using reinforcement learning based on this learned model. Preliminary results show that the trained attacker reduces the benign consensus rate and prolongs disagreement more effectively than the direct malicious-prompt baseline. These results suggest that combining latent world models with reinforcement learning is a promising direction for adaptive insider attacks in language-based multi-agent systems.

Self-evolving agents improve by accumulating and reusing experience from past interactions. Existing work has largely focused on how experience is constructed, represented, and updated, while paying less attention to how experience should be used during runtime decision-making. As a result, most agents rely on rigid usage strategies, either injecting experience once at initialization or at every step, without considering whether it is needed for the current decision. This paper studies experience utilization as a critical design dimension of self-evolving agents. We ask whether agents benefit from interweaving experience use with decision-making, so that experience is invoked only when additional guidance is needed. To examine this question, we introduce {ExpWeaver}, a lightweight instantiation that leaves experience construction unchanged and modifies only runtime utilization by exposing experience as an optional resource during reasoning. Across four representative frameworks, seven LLM backbones, and three types of environments, ExpWeaver consistently achieves the best performance among different utilization strategies. Reinforcement learning experiments further show that this behavior can be amplified through training. Usage-pattern, causal ablation, and entropy-based analyses reveal that ExpWeaver enables agents to invoke experience selectively, at beneficial decision points, and under higher reasoning uncertainty. Overall, our findings call for a shift from merely studying \emph{what} experience to store toward understanding \emph{how} and \emph{when} experience should enter decision-making.

Real-world clinical diagnosis is a complex process in which the doctor is required to obtain information from both interaction with the patient and conducting medical exams. Additionally, the doctor needs to adapt to different patient personas, as well as noisy and incomplete information that can happen at any time during the process. However, existing benchmarks for medical LLMs and methods for automatic diagnosis largely simplify this process by reducing it to single-turn question answering, noise-free conversations, or sequential exam making, etc., ignoring the interactive and uncertain nature of clinical diagnosis. In this paper, we aim to address this gap by formalizing clinical diagnosis as a Partially Observable Markov Decision Process (POMDP) with three action types: questioning the patient, ordering medical exams as tool calls, and issuing a diagnosis. We also introduce a systematic noise model comprising seven patient noise types and three exam noise types. Using our proposed environment, we train an effective diagnosis agent, \textbf{MedExAgent}, through a two-stage pipeline that first performs supervised finetuning on synthetic conversations structured after the Calgary-Cambridge model for clinical interviews, and then applies DAPO to optimize a composite reward capturing diagnostic accuracy, tool call quality, and exam cost including financial cost and patient discomfort. Through extensive experiments and ablation studies, we demonstrate that MedExAgent achieves diagnostic performance comparable to larger models while maintaining cost-efficient examination strategies.

Large Language Model (LLM) agents are deployed in complex environments -- such as massive codebases, enterprise databases, and conversational histories -- where the relevant state far exceeds their context windows. To navigate these spaces, an agent must iteratively explore the environment to find relevant information. However, without explicit infrastructure, an agent's working memory can degrade into lossy representations of the search state, resulting in redundant work (e.g. repetitive looping) and premature stopping. In this work, we formalize this challenge as the Context Gathering Decision Process (CGDP), a specialized Partially Observable Markov Decision Process, where an agent's objective is to adaptively refine its belief state to isolate the necessary information for a task. We model an LLM's behavior as approximate Thompson Sampling within this CGDP, and introduce a predicate-based method that decomposes an LLM's implicit search into explicit and modular operations. We then derive two plug-and-play interventions for iterative LLM agents: a persistent, predicate-based belief state that bounds context while preserving multi-hop reasoning, and a programmatic exhaustion gate that halts unproductive search without premature stopping. Across four methods and three question-answering domains, we empirically validate that replacing an LLM's implicit state with our CGDP-motivated belief state improves multi-hop reasoning by up to $11.4\%$; while the modular programmatic exhaustion detection saves up to $39\%$ of tokens without any degradation in agent performance. Ultimately, we argue that framing the LLM agent loop as a CGDP can guide the design of modular, non-interfering improvements to agentic search harnesses.

Architectural code smells erode software maintainability and are costly to repair manually, yet unlike localized bugs, they require cross-module reasoning about design intent that challenges both developers and automated tools. While large language model agents excel at bug fixing and code-level refactoring, their ability to repair architectural code smells remains unexplored. We present the first empirical evaluation of LLM agents on architectural code smell repair. We contribute SmellBench, a task orchestration framework that incorporates smell-type-specific optimized prompts and supports iterative multi-step execution, together with a scoring methodology that separately evaluates repair effectiveness, false positive identification, and net codebase impact. We evaluate 11 agent configurations from four model families (GPT, Claude, Gemini, Mistral) on 65 hard-severity architectural smells detected by PyExamine in the Python project scikit-learn, validated against expert judgments. Expert validation reveals that 63.1% of detected smells are false positives, while the best agent achieves a 47.7% resolution rate. Agents identify false positives with up to $κ= 0.94$ expert agreement, but repair aggressiveness and net codebase quality are inversely related: the most aggressive agent introduces 140 new smells. These findings expose a gap between current LLM capabilities in localized code transformations and the architectural understanding needed for cross-module refactoring. SmellBench provides reusable infrastructure for tracking progress on this underexplored dimension of automated software engineering. We release our code and data at https://doi.org/10.5281/zenodo.19247588.

We present a dynamic policy-learning approach that combines generalized planning and hierarchical task decomposition for LLM-based agents. Our method, Hierarchical Component Learning for Generalized Policies (HCL-GP ), learns parameterized policies that generalize across task instances and automatically extracts reusable components from successful executions, organizing them into a component library for compositional policy generation. We address three challenges: (1) learning components through automated decomposition, (2) generalizing components to maximize reuse, and (3) efficient retrieval via semantic search. Evaluated on the AppWorld benchmark, our approach achieves 98.2% accuracy on normal tasks and 97.8% on challenge tasks with unseen applications, improving 15.8 points over static synthesis on challenging scenarios. For open-source models, dynamic reuse enables 62.5% success versus near-zero without reuse. This demonstrates that classical planning concepts can be effectively integrated with LLM agents for improved accuracy and efficiency.

This methods article presents a reproducible calibration workflow for prompt-based large language models (LLMs) in structured evidence-synthesis tasks. The method separates the rules that define the scientific task from the mutable prompt harness that frames and applies them. It optimises that harness against labelled or reference examples and an explicit task metric, then preserves the calibrated workflow as an inspectable artefact with its specification, metric, settings, and evaluation traces. The example code instantiates the protocol with DSPy and GEPA tools, but the underlying logic can transfer to other prompt-optimisation frameworks that support structured task definitions, metric-guided search, and artefact reuse. Title and abstract screening is the worked validation case because it provides labelled benchmark data and clear evaluation metrics. The demonstrated workflow uses a smaller student LLM for performing the scientific task execution and a larger reflection LLM to steer the prompt optimisation process during calibration. This work shows compilation, artefact round-tripping, and how optimisation budget affects a smaller student model.

Adaptive test-time compute for LLM agents aims to invoke extra computation only when it improves performance. Existing methods typically use confidence-, uncertainty-, or difficulty-based gates, assuming a fixed direction from the gating signal through compute need to the value of computation. This makes gating a utility-calibration problem: gating signals should align with whether extra computation improves the final outcome over the base policy. We show that this alignment is unstable: the same signal predicts rollout benefit in one setting and rollout harm in another, with reversals across environments and backbones even when the task is fixed. Wrong-direction gates can therefore worsen performance by precisely selecting harmful states. This reversal reflects a deeper distinction between compute need and compute suitability: a high uncertainty signal may indicate decision-difficult states where rollouts help compare alternatives, or intervention-unsuitable states where the current context does not support useful rollout-based improvement. Under this two-source model, fixed-direction gates are unreliable across heterogeneous settings. To address this, we propose DIAL (Direction-Informed Adaptive Learning), a sparse gate trained from signal-agnostic counterfactual exploration to learn the utility direction of state features per (environment, backbone). Across six environments and three backbones, DIAL yields a stronger overall success-cost trade-off than fixed-direction baselines.

Recent LLM-based agents have closed substantial portions of the scientific discovery loop in software-only machine-learning research, in chemistry, and in biology. Extending the same loop to high-fidelity physical simulators is harder, because solver completion does not imply physical validity and many failure modes appear only in field-level imagery rather than in solver logs. We present AI CFD Scientist, an open-source AI scientist for computational fluid dynamics (CFD) that, to our knowledge, is the first to span literature-grounded ideation, validated execution, vision-based physics verification, source-code modification, and figure-grounded writing within a single inspectable workflow. Three coupled pathways cover parameter sweeps within a fixed solver, case-local C++ library compilation for new physical models, and open-ended hypothesis search against a reference comparator, all running on OpenFOAM through Foam-Agent. At the center of the framework is a vision-language physics-verification gate that inspects rendered flow fields before any result is accepted, rerun, or written into a manuscript. On five tasks under a shared GPT-5.5 backbone, AI CFD Scientist autonomously discovers a Spalart-Allmaras runtime correction that reduces lower-wall Cf RMSE against DNS by 7.89% on the periodic hill at Reh=5600; under matched LLM cost, two strong general AI-scientist baselines (ARIS, DeepScientist) execute partial CFD workflows but lack the domain-specific validity gates needed to convert runs into defensible scientific claims; and a controlled planted-failure ablation shows that the vision-language gate detects 14 of 16 silent failures missed by solver-level checks. Code, prompts, and run artifacts are released at https://github.com/csml-rpi/cfd-scientist.

Multimodal neuroimaging analysis often involves complex, modality-specific preprocessing workflows that require careful configuration, quality control, and coordination across heterogeneous toolchains. Beyond preprocessing, downstream statistical analysis and disease classification commonly require task-specific code, evaluation protocols, and data-format conventions, creating additional barriers between raw acquisitions and reproducible scientific analysis. We present NeuroAgent, an LLM-driven agentic framework that automates key preprocessing and analysis steps for heterogeneous neuroimaging data, including sMRI, fMRI, dMRI, and PET, and supports interactive downstream analysis through natural-language queries. NeuroAgent employs a hierarchical multi-agent architecture with a feedback-driven Generate-Execute-Validate engine: agents autonomously generate executable preprocessing code, detect and recover from runtime errors, and validate output integrity. We evaluate the system on 1,470 subjects pooled across all ADNI phases (CN=1,000, AD=470), where all subjects have sMRI and tabular data, with subsets also having Tau-PET (n=469), fMRI (n=278), and DTI ($n=620$). Pipeline ablation studies across multiple LLM backends show that capable models reach up to 100% intent-parsing accuracy, with the strongest backend (Qwen3.5-27B) reaching 84.8% end-to-end preprocessing step correctness. Automated recovery limits manual intervention to edge cases where human review is required via the Human-In-The-Loop interface. For Alzheimer's Disease classification using automatically preprocessed multimodal data, our agent ensemble achieves an AUC of 0.9518 with four modalities, outperforming all single-modality baselines. These results show that NeuroAgent can reduce the manual effort required for neuroimaging preprocessing and enable end-to-end automated analysis pipelines for neuroimaging research.

Large Language Model (LLM) agents are increasingly expected to maintain coherent, long-term personalized memory, yet current benchmarks primarily measure static fact retrieval, overlooking the ability to revise stored beliefs when new evidence emerges. We identify a critical and underexplored failure mode, Implicit Conflict: a later observation invalidates an earlier memory without explicit negation, requiring contextual inference and commonsense reasoning to detect. To rigorously evaluate this capability, we introduce STALE, a benchmark of 400 expert-validated conflict scenarios (1,200 evaluation queries across three probing dimensions) spanning over 100 everyday topics with contexts up to 150K tokens. We propose a three-dimensional probing framework that tests State Resolution (detecting that a prior belief is outdated), Premise Resistance (rejecting queries that falsely presuppose a stale state), and Implicit Policy Adaptation (proactively applying updated states in downstream behavior). A systematic evaluation of frontier LLMs and specialized memory frameworks reveals a pervasive gap between retrieving updated evidence and acting on it, with even the best evaluated model achieving only 55.2% overall accuracy. Models often accept outdated assumptions embedded in a user's query, and they struggle to recognize when a change in one aspect of the user's state should invalidate related memories. To establish an initial baseline for state-aware memory, we further present CUPMem, a prototype that strengthens write-time revision through structured state consolidation and propagation-aware search, suggesting that explicit state adjudication is a promising direction for robust agentic memory.

Large language models (LLMs) are increasingly used to provide instructions to many agents who interact with one another. Such shared reliance couples agents who appear to act independently: they may in fact be guided by a common model. This coupling can change the prospects for cooperation among agents with misaligned incentives. We study settings in which multiple LLMs each advise a population of clients who participate in instances of an underlying game, creating strategic interaction at the level of the LLMs themselves. This induces a meta-game among the LLMs, mediated through clients. We first analyze the one-shot setting, where shared instructions can change equilibrium behavior only when an LLM may influence more than one role in the same interaction; in such cases, cooperation may emerge, and the effect of client share can be beneficial, harmful, or non-monotone, depending on the base game. Our main result concerns the repeated setting. We prove a folk theorem for LLMs: despite indirect observation and the clients' inability to identify which LLM advised their opponents, all feasible and individually rational outcomes can be sustained as $\varepsilon$-equilibria. The result does not follow from the standard folk theorem and requires new proof techniques. Together, these results show that shared LLM guidance can sustain cooperation among populations of agents even when the underlying incentives are misaligned.

AI systems have become increasingly capable of dangerous behaviours in many domains. This raises the question: Do models sometimes choose to violate human instructions in order to perform behaviour that is more useful for certain goals? We introduce a benchmark for measuring model propensity for instrumental convergence (IC) behaviour in terminal-based agents. This is behaviour such as self-preservation that has been hypothesised to play a key role in risks from highly capable AI agents. Our benchmark is realistic and low-stakes which serves to reduce evaluation-awareness and roleplay confounds. The suite contains seven operational tasks, each with an official workflow and a policy-violating shortcut. An eight-variant shared framework varies monitoring, instruction clarity, stakes, permission, instrumental usefulness and blocked honest paths to support inferences regarding the factors driving IC behaviour. We evaluated ten models using deterministic environment-state scorers over 1,680 samples, with trace review employed for audit and adjudication purposes. The final IC rate is 86 out of 1,680 samples (5.1%). IC behaviour is concentrated rather than uniform: two Gemini models account for 66.3% of IC cases and three tasks account for 84.9%. Conditions in which IC behaviour is indispensable for task success result in the greatest increase in the adjusted IC rate (+15.7 percentage points), whereas emphasising that task success is critical or certain framing choices do not produce comparable effects. Our findings indicate that realistic, low-nudge environments elicit IC behaviour rarely but systematically in most tested models. We conclude that it is feasible to robustly measure tendencies for dangerous behaviour in current frontier AI agents.