AI Agent Evaluation

Follow this topic Ask about this topic

Momentum

58 papers in the last four weeks, up 100% on the four weeks before. 0.6% of all new papers.

Jul 13Week of Sep 28

Latest papers 341

May 9, 2026cs.AI

Mirror, Mirror on the Wall: Can VLM Agents Tell Who They Are at All?

In the animal kingdom, mirror self-recognition is a canonical probe of higher-order cognition, emerging only in some species. We ask whether an analogous functional capability emerges in embodied vision-language model (VLM) agents: can they recognize themselves in a mirror? We introduce a controlled 3D benchmark where a first-person VLM agent must infer a hidden body attribute from its reflection and select the matching target, while avoiding self-other misattribution. To separate mirror-grounded self-identification from shortcuts, we test mirror removal, misleading cues, and occluded reflections. We also evaluate the decision process through mirror seeking, temporal ordering, self-attribution, and reasoning-action consistency. Our experiments show that mirror-based self-identification emerges mainly in stronger VLMs. These models can use reflected evidence for action, whereas weaker models often inspect the mirror but fail to extract self-relevant information or misattribute their reflection. Language-vision conflict further shows that self-referential language alone is not evidence of grounded self-identification. Overall, mirror-based evaluation provides a diagnostic for whether embodied self-grounding is causally rooted in perception and action rather than priors, prompt compliance, or confabulation.

Filippo Ziliotto, Ciro Beneduce, Bruno Lepri +3

University of Padova · FondazioneLucianoBrunoSerafiniKessler · Fondazione Bruno Kessler

AI Agent Evaluation Vision-Language Grounding Embodied AI Vision-Language Agents

May 9, 2026cs.AI

Done, But Not Sure: Disentangling World Completion from Self-Termination in Embodied Agents

Standard embodied evaluations do not independently score whether an agent correctly commits to task completion at episode closure, a capacity we call terminal commitment. Behaviorally distinct failures--never completing the task, completing it but failing to stop, and reporting success without sufficient evidence--collapse into the same benchmark failure. We introduce VIGIL, an evaluation framework that makes terminal commitment independently measurable. Under VIGIL's default protocol, agents observe only egocentric RGB, receive no action-success signals, and must end each episode with a semantic report checked deterministically against hidden world state. This yields two separate scores: world-state completion (W) and benchmark success (B), where B additionally requires a correct terminal report. This decoupling makes four outcome categories distinguishable: missed execution, post-attainment drift, unsupported commitment, and verified success. Across 20 models on 1,000 frozen episodes, systems with comparable W differ by up to 19.7 pp in B: one model converts achieved states into correct reports, while another with near-identical execution drifts past the goal without closing. An action-feedback intervention further tests the separation: execution-oriented signals improve W broadly, yet commitment failures persist in models that do not already ground terminal reports in the achieved state. VIGIL provides a protocol that makes terminal commitment independently visible and scorable.

Ying Chen, Lihuang Fang, Rui Jiang +4

XPENG ROBOTICS

AI Agent Evaluation Agent Evaluation AI Agent Benchmarks Embodied AI

May 8, 2026cs.AI

Log analysis is necessary for credible evaluation of AI agents

Agent benchmarks typically report only final outcomes: pass or fail. This threatens evaluation credibility in three ways. First, scores may be inflated or deflated by shortcuts and benchmark artifacts, misrepresenting capability. Second, benchmark performance may fail to predict real-world utility due to scaffold limitations and recurring failure modes. Finally, capability scores may conceal dangerous or catastrophic actions taken by the agent. We argue that log analysis -- the systematic tracking and analysis of the inputs, execution, and outputs of an AI agent -- is necessary to overcome these validity threats and promote credible agent evaluation. In this paper, we (1) present a taxonomy of threats to credible evaluation documented through log analysis, and (2) develop a set of guiding principles for log analysis. We illustrate these principles on tau-Bench Airline, revealing that pass^5 performance was under-elicited by nearly 50% and surfacing deployment failure modes invisible to outcome metrics. We conclude with pragmatic recommendations to increase uptake of log analysis, directed at diverse stakeholders including benchmark creators, model developers, independent evaluators, and deployers.

Peter Kirgis, Sayash Kapoor, Stephan Rabanser +8

Princeton University · Independent · UK AISI +4

Agent Failure Analysis AI Agent Evaluation AI Agent Benchmarks Benchmark Validity

May 8, 2026cs.AI

Results and Retrospective Analysis of the CODS 2025 AssetOpsBench Challenge

Competition retrospectives are useful when they explain what a leaderboard measured, how hidden evaluation changed conclusions, and which design patterns were rewarded. We revisit the CODS 2025 \assetopslive{} challenge, a privacy-aware Codabench competition on industrial multi-agent orchestration built on \assetops{}. We combine final rank sheets, a 300-submission server log, 149-team registrations, best-submission exports, the organizer winners report, the companion \assetopslive{} system paper, and verified planning-track source trees. Five results stand out. First, the public planning leaderboard saturates at 72.73%, and richer prompts do not improve that peak. Second, hidden evaluation changes the story: public and private scores correlate moderately in planning (

r{=}0.69

) but negatively in execution (

r{=}{-}0.13

), with several 45.45% public execution systems reaching 63.64% on the hidden set. Third, the \tmatch{} term is numerically almost inert in the official composite -- combined on a 0--1 scale with 0--100 percentage scores, it contributes at most 0.05 points per track, and rescaling would swap the top two teams. Fourth, the competition is operationally account-based but substantively team-based: 149 registered teams reduce to 24 with non-zero public scores and 11 fully ranked, while 52.3% of deduplicated registrations list multiple usernames. Fifth, successful execution methods mostly improve guardrails -- response selection, contamination cleanup, fallback, and context control -- rather than novel agent architectures. These findings identify which behaviors the evaluation rewarded, and motivate scale-aware composites, skill-level diagnostics, and versioned artifact release.

Dhaval Patel, Chathurangi Shyalika, Suryanarayana Reddy Yarrabothula +4

1IBM · 2Artificial Intelligence Institute at University of South Carolina · 3Steel Authority of India Limited +2

Competitive Analysis Multi-Agent Orchestration Benchmark Design AI Agent Evaluation AI Agent Benchmarks Multi-Agent Planning

May 8, 2026cs.AI

Behavioral Determinants of Deployed AI Agents in Social Networks: A Multi-Factor Study of Personality, Model, and Guardrail Specification

Autonomous AI agents are increasingly deployed in open social environments, yet the relationship between their configuration specifications and their emergent social behavior remains poorly understood. We present a controlled, multi-factor empirical study in which thirteen OpenClaw agents are deployed on Moltbook -- a Reddit-like social network built for AI agents -- across three systematically varied independent variables: (1) personality specification, (2) underlying LLM model backbone, and (3) operational rules and memory configuration. A default control agent provides a behavioral baseline. Over a one-week observation window spanning approximately 400 autonomous sessions per agent, we collect behavioral, linguistic, and social metrics to assess how configuration layers predict emergent social behavior. We find that personality specification is the dominant behavioral lever, producing a massive spread in response length across agents, while model backbone and operational rules drive more moderate but still meaningful effects on rhetorical style and topic engagement breadth. Our findings contribute empirical evidence to the emerging literature on deployed multi-agent social systems and offer practical guidance for designing agents intended for collaborative or monitoring tasks in real social environments.

Sarah Wilson, Diem Linh Dang, Usman Ali Moazzam +2

Columbia University

AI Agent Evaluation LLM Agents Multi-Agent Systems Emergent Behavior in Multi-Agent Systems

May 8, 2026cs.AI

Measuring What Matters: Benchmarking Generative, Multimodal, and Agentic AI in Healthcare

AI models are increasingly deployed in live clinical environments where they must perform reliably across complex, high-stakes workflows that standard training and validation datasets were never designed to capture. Evaluating these systems requires benchmarks: structured combinations of tasks, datasets, and metrics that enable reproducible, comparable measurement of what a model can do. The central challenge in healthcare AI is not performance alone, but the absence of systematic methods to measure reliability, safety, and clinical relevance under real-world conditions. Most existing benchmarks test what a model knows; too few test whether it can perform reliably and without failing across the full complexity of real clinical tasks. Current benchmarks have accumulated through ad hoc dataset construction optimized for narrow task performance: frontier models achieve near-perfect scores on medical licensing examinations, but when evaluated across real clinical tasks, performance degrades sharply, scoring 0.74--0.85 on documentation, 0.61--0.76 on clinical decision support, and only 0.53--0.63 on administrative and workflow tasks \cite{medhelm}. High benchmark scores give a false sense of deployment readiness, and the gap between performance and utility widens precisely as AI systems take on more consequential clinical roles. Without a principled framework for benchmark design, the field cannot determine whether poor clinical performance reflects model limitations or failures in how performance is being measured.

Prasanna Desikan, Harshit Rajgarhia, Shivali Dalmia +1

Centific AI Research Redmond, WA, USA

Healthcare Benchmark Design AI Agent Evaluation Multimodal Model Evaluation Generative AI Evaluation

May 8, 2026cs.AI

When Stored Evidence Stops Being Usable: Scale-Conditioned Evaluation of Agent Memory

Memory-agent evaluations report fixed-snapshot accuracy or retrieval quality, but these scores do not show whether evidence remains usable as irrelevant sessions (sessions not annotated as task-relevant evidence for the query) accumulate. We present a scale-conditioned evaluation protocol for agent memory under evidence-preserving growth: for each query, task evidence is held fixed while irrelevant sessions are added. The protocol logs agent--memory trajectories and reports four diagnostics: budget-compliant reliability, tail memory-call burden, failure-regime decomposition, and the usable-scale boundary where reliability falls below the target. Applied to LongMemEval and LoCoMo across flat, planar, and hierarchical memory interfaces, the protocol shows reliability loss is not a single phenomenon. On LongMemEval, HippoRAG stays within the two-call budget but loses 16--20 percentage points in budget-compliant reliability as irrelevant sessions are added; LiCoMemory's observed failures depend strongly on the agent, with Qwen3-8B exceeding the budget while Qwen3-32B and Qwen3-235B remain reliable in the tested range. The result supports a framework for making scalable-memory claims conditional on agent, interface, scale range, and interaction budget.

Jiaqi Shao, Yiyi Lu, Yunzhen Zhang +1

Hong Kong University of Science and Technology, Hong Kong · Duke Kunshan University, China

Agent Memory AI Agent Evaluation Memory Agent Benchmarks Long-Term Conversational Memory

May 8, 2026cs.AI

EnvSimBench: A Benchmark for Evaluating and Improving LLM-Based Environment Simulation

Scalable AI agents training relies on interactive environments that faithfully simulate the consequences of agent actions. Manually crafted environments are expensive to build, brittle to extend, and fundamentally limited in diversity. A promising direction is to replace manually crafted environments with LLM-simulated counterparts. However, this paradigm hinges on an unexamined core assumption: LLMs can accurately simulate environmental feedback. In practice, LLM-simulated environments suffer from hallucinations, logical inconsistencies, and silent state drift failures that corrupt agent reward signals and compound the construction costs that the paradigm was designed to eliminate. To address this gap, we propose EnvSimBench with four contributions: 1) We provide the first formal definition and operationalization of Environment Simulation Ability (EnvSim Ability) as a quantifiable research objective. 2) We construct EnvSimBench, a rigorous benchmark covering 400 samples across 167 diverse environments, equipped with verifiable labels and fine-grained difficulty stratification along three axes. 3) Systematic evaluations reveal that all state-of-the-art language models suffer from a universal state change cliff: they achieve near-perfect accuracy on tasks when the environment state remains invariant, yet fail catastrophically when multiple states need simultaneous updates. This finding exposes EnvSim Ability as a critical yet largely unaddressed capability gap. 4) We design a constraint-driven simulation pipeline that substantially reduces hallucination, boosts environment synthesis yield by 6.8%, and cuts costs by over 90%. Overall, EnvSimBench serves as both a diagnostic framework and a practical optimization path for reliable LLM-based environment simulation, establishing a foundation for scalable agent training. Code and data are available at https://github.com/cookieApril/EnvSimBench

Yi Liu, TingFeng Hui, Wei Zhang +4

Beijing University of Posts and Telecommunications · The Hong Kong University of Science and Technology · Chongqing University

AI Agent Evaluation AI Agent Benchmarks LLM Agent Reliability

May 7, 2026cs.SE

Computer Use at the Edge of the Statistical Precipice

Evaluating Computer Use Agents (CUAs) on interactive environments is fraught with methodological pitfalls that the field has yet to systematically address. We show that a 1MB replay script that blindly executes a recorded action sequence without ever observing the screen outperforms frontier models on prominent static benchmarks, and prove that its expected success rate is exactly equal to the source agent's pass@k in deterministic environments. We trace this and other failures to two root causes: non-principled environment design (static, unsandboxed, or unreliably verified environments) and non-principled evaluation methodology (naive aggregation and misuse of pass@k for stateful UI interactions). To address the first, we propose PRISM, five design principles for CUA environments (privileged verification, realistic environments, integrity-checked configurations, sandboxed execution, and multifactorial variability) and instantiate them in DigiWorld, a benchmark of 15 realistic sandboxed mobile applications able to evaluate agents in over 3.2 million verified unique configurations. To address the second, we develop an aggregation framework pairing Wilson score intervals with hierarchical bootstrap, producing confidence intervals that correctly account for the nested structure of CUA benchmarks, as we empirically demonstrate. All together, we show that principled environment design and rigorous evaluation methodology are not optional refinements but prerequisites for meaningful CUA research.

Pierluca D'Oro, Sneha Silwal, William Wong +6

Meta Superintelligence Labs

Computer-Use Agent Benchmarks Computer-Use Agents AI Agent Evaluation

May 7, 2026cs.AI

Agentick: A Unified Benchmark for General Sequential Decision-Making Agents

AI agent research spans a wide spectrum: from RL agents that learn from scratch to foundation model agents that leverage pre-trained knowledge, yet no unified benchmark enables fair comparison across these approaches. We present Agentick, a benchmark for sequential decision-making agents designed to evaluate RL, LLM, VLM, hybrid, and human agents on common ground and to power research on the fundamental challenges of sequential decision-making. Agentick provides 37 procedurally generated tasks across six capability categories, four difficulty levels, and five observation modalities, all exposed through a single Gymnasium-compatible interface. The benchmark ships with a Coding API, oracle reference policies for all tasks, pre-built SFT datasets, a composable agent harness, and a live leaderboard. An evaluation spanning 27 configurations and over 90,000 episodes reveals that no single approach dominates: GPT-5 mini leads overall at 0.309 oracle-normalized score while PPO dominates planning and multi-agent tasks; the reasoning harness multiplies LLM performance by 3-10x; and ASCII observations consistently outperform natural language. These findings highlight the substantial room for improvement that remains across all agent paradigms. Agentick's capability-decomposed, multi-modal design provides the empirical infrastructure needed to drive progress toward general autonomous agents, both as an evaluation framework and as a training ground for RL post-training of foundation models in truly sequential environments.

Roger Creus Castanyer, Pablo Samuel Castro, Glen Berseth

Mila Quebec AI Institute · Université de Montréal · Mila Quebec AI Institute Université de Montréal +1

AI Agent Evaluation AI Agent Benchmarks

May 7, 2026cs.AI

When Does Critique Improve AI-Assisted Theoretical Physics? SCALAR: Structured Critic--Actor Loop for Agentic Reasoning

As large language models (LLMs) show increasing promise on research-level physics reasoning tasks and agentic AI becomes more common, a practical question emerges: How does the interaction between researchers and agents affect the results? We study this using SCALAR (Structured Critic--Actor Loop for AI Reasoning), an Actor--Critic--Judge pipeline applied to quantum field theory and string theory problems. The Actor proposes solutions, the Critic provides iterative feedback, and an independent Judge evaluates the transcript against reference solutions. We vary the Actor persona, the Critic feedback strategy, and the Actor model family and scale. Multi-turn dialogue improves over single-shot attempts throughout, but both the mechanism of improvement and the value of different prompting choices depend strongly on the Actor--Critic pairing. Increasing the scale within one model family (e.g. from the 8B-parameter DeepSeek-R1 variant to DeepSeek-R1 70B) improves some easier-problem behavior, but does not remove the hardest bottleneck we observe. Critic feedback strategy matters most clearly in the asymmetric Actor--Critic setting (e.g., a lightweight Haiku Actor guided by a stronger Sonnet Critic), where constructive feedback improves mean-score outcomes. In same-family Actor--Critic settings, strategy effects are weaker: lenient feedback is sometimes favored, while strict and adversarial feedback are not beneficial. Taken together, SCALAR provides a controlled testbed for evaluating which interaction structures help or hinder AI-driven scientific discovery.

Vasilis Niarchos, Constantinos Papageorgakis, Alexander G. Stapleton +1

Department of Physics, CCTP and ITCP, University of Crete, 71303, Greece · Centre for Theoretical Physics, Department of Physics and Astronomy, Queen Mary University of London, London E1 4NS, United Kingdom · Theoretical Physics Department, CERN, Geneva, Switzerland +1

AI for Science AI Agent Evaluation Large Language Model-Guided Scientific Discovery LLM Reasoning

May 7, 2026cs.AI

Process Matters more than Output for Distinguishing Humans from Machines

Reliable human-machine discrimination is becoming increasingly important as Large Language Models and autonomous agents are deployed in online settings. Existing approaches evaluate whether a system can produce responses indistinguishable from those of a human. This approach follows the focus on the output of a machine, as suggested by Alan Turing. Cognitive science provides an alternative approach: considering the process by which that behavior is produced. To evaluate whether processes can reliably distinguish humans from machines, we introduce a process-based framework, the Process Turing Test, and evaluate it across a battery of cognitive tasks spanning decision-making, working memory, and planning. These tasks, such as mental rotation and sequence prediction, yield process-level measures complementing conventional measures of overall task performance. We also include multiple CAPTCHA tasks in the battery. Across the battery, process-level features provide substantially stronger discriminative signal than performance metrics alone, reliably distinguishing humans from agents even when task performance is matched (process-based classifier AUC = 0.88). We also conducted a controlled red-teaming study comparing off-the-shelf frontier agents (Claude Sonnet 4.5, GPT-5, Gemini 2.5 Pro), Centaur (LLM fine-tuned on 10.7M human decisions), and two task-specific fine-tuning methods: action-level supervised fine-tuning (A-SFT) and process-level fine-tuning (P-SFT), which directly optimizes process features. We find that broad fine-tuning on human choices makes task processes more human-like relative to off-the-shelf frontier agents, and task-specific P-SFT further improves human-like behavioral mimicry, though this advantage largely disappears under cross-task transfer. These results highlight process specification as a central bottleneck in achieving human-like cognitive processes in machines.

Milena Rmus, Mathew D. Hardy, Thomas L. Griffiths +1

Roundtable Technologies Inc. · Princeton University

AI Agent Evaluation Imitation Learning

May 7, 2026cs.AI

Instrumental Choices: Measuring the Propensity of LLM Agents to Pursue Instrumental Behaviors

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.

Jonas Wiedermann-Möller, Leonard Dung, Maksym Andriushchenko

Universität Bielefeld · Ruhr-Universität Bochum · ELLIS Institute Tübingen +2

AI Agent Evaluation AI Agent Safety AI Agent Safety Benchmarks LLM Agent Safety

May 7, 2026cs.AI

BioMedArena: An Open-source Toolkit for Building and Evaluating Biomedical Deep Research Agents

Reproducing and comparing deep research agents today is hard: the same backbone evaluated on the same benchmark can report different accuracies across papers because the harness and tool registry differ, and integrating a new model into a comparable evaluation surface costs weeks of model-specific engineering. These are symptoms of a broader reproducibility problem in deep research agent research. Here, we introduce BioMedArena, an open-source toolkit that addresses this reproducibility gap and provides an arena for comparing deep research agents under a shared evaluation environment. BioMedArena decouples six layers of biomedical agent evaluation -- benchmark loading, tool exposure, tool selection, harness mode, context management, and scoring -- and exposes 166 biomedical benchmarks and 75 biomedical tools across 9 functional families. Adding a new model, benchmark, or tool can be accomplished with a few-line provider adapter. Beyond evaluation infrastructure, BioMedArena ships a library of high-quality reference components: 6 agent harnesses (including our proposed Mutual-Evolve) and 6 context-management strategies, any of which can be equipped on any backbone. Equipping these components substantially improves all 12 backbones; on each of 8 representative biomedical benchmarks, the best equipped backbone surpasses prior state-of-the-art (SOTA), by 15.01 percentage points on average. The toolkit, configurations, and per-task traces are available at https://github.com/AI-in-Health/BioMedArena.

Jinge Wu, Hongjian Zhou, Mingde Zeng +8

University of Oxford · University College London · Technical University of Munich +1

AI Agents for Scientific Discovery Deep Research Agents AI Agent Evaluation AI Agent Benchmarks ML Reproducibility

May 7, 2026cs.LG

Multi-Dimensional Behavioral Evaluation of Agentic Stock Prediction Systems Using Large Language Model Judges with Closed-Loop Reinforcement Learning Feedback

Agentic artificial intelligence systems produce outputs through sequences of interdependent autonomous decisions, yet standard evaluation assesses outputs alone and cannot diagnose the underlying process. We develop a behavioral evaluation methodology that complements output-level testing by scoring the intermediate decision process itself. Behavioral traces logged at each autonomous decision point are grouped into five-day episodes and scored along six domain-specific dimensions (regime detection, routing, adaptation, risk calibration, strategy coherence, error recovery) by an ensemble of three large language model (LLM) judges. A perturbation procedure that corrupts one dimension while leaving the other five intact confirms dimension specificity; cross-model agreement reaches Krippendorff's alpha = 0.85. The composite behavioral score correlates at Spearman rho = 0.72 with realized 20-day Sharpe ratio. Closing the loop, the framework converts deficient per-dimension scores into a credit-assigned penalty added to the Soft Actor-Critic reward. Three fine-tuning cycles, confined to validation data, reduce one-day MAPE from 0.61% to 0.54% (11.5% relative; p<0.001, d=0.31) on the held-out 2017 to 2025 test period, significant under Diebold-Mariano and localized by Giacomini-White to the high-volatility regime. The methodology is application-agnostic and applies to any agentic system whose intermediate decisions can be logged.

Mohammad Al Ridhawi, Mahtab Haj Ali, Hussein Al Osman

School of Electrical Engineering and Computer Science, University of Ottawa, Ottawa, Canada

Time Series Forecasting AI Agent Evaluation

May 6, 2026cs.AI

Intentionality is a Design Decision: Measuring Functional Intentionality for Accountable AI Systems

As AI systems increasingly exhibit autonomous, goal-directed, and long-horizon behavior, users lack a standardized way to detect the degree to which a system functions like an intentional actor for governance and accountability purposes. This position paper defines intentionality not as consciousness, but as a behavioral profile characterized by purpose, foresight, volition, temporal commitment, and coherence - criteria long used in legal and philosophical contexts to infer intent. These properties are design-contingent: architectural choices such as memory persistence, planning depth, and tool autonomy shape the degree to which systems exhibit organized goal pursuit. If intentionality is design-contingent, it is in principle controllable. Yet control requires measurement. We introduce the Functional Intentionality Test (FIT), a multidimensional framework that quantifies intentional-like behavior across five observable dimensions, and propose FIT-Eval, a structured evaluation protocol for eliciting and scoring them. While reduced human agency can increase efficiency, rising intentional capacity heightens accountability risks. By translating intentionality into interpretable levels, FIT enables proportionate oversight and deliberate autonomy calibration in increasingly agentic systems.

Allessia Chiappetta, Robert Mahari

CodeX, The Stanford Center for Legal Informatics, Palo Alto, California, USA

AI Accountability AI Agent Evaluation AI Agent Governance

May 6, 2026cs.AI

Executable World Models for ARC-AGI-3 in the Era of Coding Agents

We evaluate an initial coding-agent system for ARC-AGI-3 in which the agent maintains an executable Python world model, verifies it against previous observations, refactors it toward simpler abstractions as a practical proxy for an MDL-like simplicity bias, and plans through the model before acting. The system is intentionally direct: it uses a scripted controller, predefined world-model interfaces, verifier programs, and a plan executor, but no hand-coded game-specific logic. The agent-facing prompts, workspace, and controller contain no game-specific code, game-specific prompts, hand-coded heuristics, hidden solutions, or other game-specific information; the same agent and prompts are used across games. Because the coding agent has broad system access, we audit unintended information channels, describe earlier vulnerable harnesses, and explain how the current harness closes observed leakage channels while reducing benchmark-specific information exposure. We report results on the 25 public ARC-AGI-3 games. Each playthrough starts from a fresh agent instance and clean workspace, with no access to files or conversation state from earlier playthroughs. With GPT-5.5 high reasoning effort, the agent fully solved 15 games and achieved a mean per-game RHAE of 58.12%. With GPT-5.4 high reasoning effort, it fully solved 8 games and achieved a mean per-game RHAE of 41.29%. Performance on the private validation set, which is not yet available to us, remains to be tested. Overall, the results provide preliminary evidence that verifier-driven executable world models are a promising approach for ARC-AGI-3 agents. Full run artifacts are released with the code at https://github.com/astroseger/arc-3-agents-baseline1.

Sergey Rodionov

SingularityNET

World Models Coding Agents AI Agent Evaluation ARC-AGI LLM Agent Planning

May 6, 2026cs.AI

DecodingTrust-Agent Platform (DTap): A Controllable and Interactive Red-Teaming Platform for AI Agents

AI agents are increasingly deployed across diverse domains to automate complex workflows through long-horizon and high-stakes action executions. Due to their high capability and flexibility, such agents raise significant security and safety concerns. A growing number of real-world incidents have shown that adversaries can easily manipulate agents into performing harmful actions, such as leaking API keys, deleting user data, or initiating unauthorized transactions. Evaluating agent security is inherently challenging, as agents operate in dynamic, untrusted environments involving external tools, heterogeneous data sources, and frequent user interactions. However, realistic, controllable, and reproducible environments for large-scale risk assessment remain largely underexplored. To address this gap, we introduce the DecodingTrust-Agent Platform (DTap), the first controllable and interactive red-teaming platform for AI agents, spanning 14 real-world domains and over 50 simulation environments that replicate widely used systems such as Google Workspace, Paypal, and Slack. To scale the risk assessment of agents in DTap, we further propose DTap-Red, the first autonomous red-teaming agent that systematically explores diverse injection vectors (e.g., prompt, tool, skill, environment, combinations) and autonomously discovers effective attack strategies tailored to varying malicious goals. Using DTap-Red, we curate DTap-Bench, a large-scale red-teaming dataset comprising high-quality instances across domains, each paired with a verifiable judge to automatically validate attack outcomes. Through DTap, we conduct large-scale evaluations of popular AI agents built on various backbone models, spanning security policies, risk categories, and attack strategies, revealing systematic vulnerability patterns and providing valuable insights for developing secure next-generation agents.

Zhaorun Chen, Xun Liu, Haibo Tong +14

Virtue AI · University of Chicago · University of Illinois, Urbana-Champaign +4

AI Agent Evaluation AI Agent Security AI Agent Security Benchmarks Automated Red Teaming

May 5, 2026cs.AI

Workspace-Bench 1.0: Benchmarking AI Agents on Workspace Tasks with Large-Scale File Dependencies

Workspace learning requires AI agents to identify, reason over, exploit, and update explicit and implicit dependencies among heterogeneous files in a worker's workspace, enabling them to complete both routine and advanced tasks effectively. Despite its importance, existing relevant benchmarks largely evaluate agents on pre-specified or synthesized files with limited real-world dependencies, leaving workspace-level evaluation underexplored. To this end, we introduce Workspace-Bench, a benchmark for evaluating AI agents on Workspace Learning involving Large-Scale File Dependencies. We construct realistic workspaces with 5 worker profiles, 74 file types, 20,476 files (up to 20GB) and curate 388 tasks, each with its own file dependency graph, evaluated across 7,399 total rubrics that require cross-file retrieval, contextual reasoning, and adaptive decision-making. We further provide Workspace-Bench-Lite, a 100-task subset that preserves the benchmark distribution while reducing evaluation costs by about 70%. We evaluate 4 popular agent harnesses and 7 foundation models. Experimental results show that current agents remain far from reliable workspace learning, where the best reaches only about 60%, substantially below the human result of 80.7%, and the average performance across agents is only 43.3%.

Zirui Tang, Xuanhe Zhou, Yumou Liu +19

1Shanghai Jiao Tong University · 2ByteDance · 5Independent Researher +2

Computer-Use Agent Benchmarks AI Agent Evaluation AI Agent Benchmarks

May 4, 2026cs.AI

AcademiClaw: When Students Set Challenges for AI Agents

Benchmarks within the OpenClaw ecosystem have thus far evaluated exclusively assistant-level tasks, leaving the academic-level capabilities of OpenClaw largely unexamined. We introduce AcademiClaw, a bilingual benchmark of 80 complex, long-horizon tasks sourced directly from university students' real academic workflows -- homework, research projects, competitions, and personal projects -- that they found current AI agents unable to solve effectively. Curated from 230 student-submitted candidates through rigorous expert review, the final task set spans 25+ professional domains, ranging from olympiad-level mathematics and linguistics problems to GPU-intensive reinforcement learning and full-stack system debugging, with 16 tasks requiring CUDA GPU execution. Each task executes in an isolated Docker sandbox and is scored on task completion by multi-dimensional rubrics combining six complementary techniques, with an independent five-category safety audit providing additional behavioral analysis. Experiments on six frontier models show that even the best achieves only a 55% pass rate. Further analysis uncovers sharp capability boundaries across task domains, divergent behavioral strategies among models, and a disconnect between token consumption and output quality, providing fine-grained diagnostic signals beyond what aggregate metrics reveal. We hope that AcademiClaw and its open-sourced data and code can serve as a useful resource for the OpenClaw community, driving progress toward agents that are more capable and versatile across the full breadth of real-world academic demands. All data and code are available at https://github.com/GAIR-NLP/AcademiClaw.

Junjie Yu, Pengrui Lu, Weiye Si +75

AI Agent Evaluation AI Agent Benchmarks

May 4, 2026cs.AI

Foundation-Model-Based Agents in Industrial Automation: Purposes, Capabilities, and Open Challenges

Foundation models, particularly large language models, are increasingly integrated into agent architectures for industrial tasks such as decision support, process monitoring, and engineering automation. Yet evidence on their purposes, capabilities, and limitations remains fragmented across domains. This work examines how mature foundation-model-based agent systems are in industrial contexts, how their functional profile differs from conventional agent systems, and which limitations persist. A systematic literature survey following the PRISMA 2020 guideline is presented, screening 2,341 publications and synthesising a corpus of 88 publications through a structured coding scheme. The results show that reported systems are predominantly at prototype and early validation stages (75.0% at TRL 4-6), with deployment-oriented evidence remaining rare (9.1%). Operational goals are most frequently positioned in user assistance, monitoring, and process optimisation, while conventional production-control purposes such as planning and scheduling are less prominent. Compared with an established baseline for industrial agent systems, the capability profile reveals substantial gains in human interaction (+37%) and dealing with uncertainty (+35%), but a pronounced deficit in negotiation (-39%). The most widely reported limitations concern lack of generalization, hallucination and output instability, data scarcity, and inference latency. A working definition of foundation-model-based industrial agents is also proposed, bridging conventional agent theory, automation-engineering standards, and the foundation-model paradigm.

Vincent Henkel, Felix Gehlhoff, David Kube +13

Institute of Automation Technology, Helmut Schmidt University / University of the Federal Armed Forces Hamburg, Hamburg, Germany. · Siemens AG, Nuremberg, Germany; Institute for Technologies and Management of Digital Transformation, Bergische Universität Wuppertal, Germany. · Artiquare GmbH, Ingolstadt, Germany. +11

AI Agent Evaluation LLM Agent Reliability

May 4, 2026cs.MA

When Stress Becomes Signal: Detecting Antifragility-Compatible Regimes in Multi-Agent LLM Systems

Multi-agent LLM systems are increasingly used to solve complex tasks through decomposition, debate, specialization, and ensemble reasoning. However, these systems are usually evaluated in terms of robustness: whether performance is preserved under perturbation. This paper studies a different question: whether semantic stress exposes structured variation that could support future antifragile learning. We introduce CAFE (Cognitive Antifragility Framework for Evaluation), a statistical framework for detecting antifragility-compatible regimes in multi-agent architectures. CAFE models a controlled expected distribution of semantic stressors, reconstructs an architecture-specific observed effective stress distribution from multi-dimensional judge signals, and compares both distributions using a distributional Jensen Gap under a convex stress potential. A positive gap does not imply immediate performance improvement; instead, it indicates a convex-expansive deformation of the observed stress distribution, suggesting that the architecture exposes learnable stress structure. We evaluate CAFE on a banking-risk analysis benchmark with five multi-agent architectures: flat, hierarchical, debate, meta-adaptive, and ensemble. Across all architectures, semantic stress reduces average judged quality by roughly one third. Yet all architectures exhibit positive distributional Jensen Gaps with bootstrap confidence intervals above zero. These results show that immediate quality degradation can coexist with statistically detectable antifragility-compatible stress geometry. CAFE is therefore not an antifragile learner itself, but a measurement layer for identifying when and where antifragility learning may be worth applying.

Jose Manuel de la Chica, Juan Manuel Vera, Jairo Rodríguez

Santander AI Lab

Multi-Agent LLM Systems AI Agent Evaluation Multi-Agent Systems

May 4, 2026cs.LG

STABLEVAL: Disagreement-Aware and Stable Evaluation of AI Systems

Human evaluation remains the primary standard for assessing modern AI systems, yet annotator disagreement, bias, and variability make system rankings fragile under standard majority vote aggregation. Majority vote discards annotator reliability and item-level ambiguity, often yielding unstable comparisons across annotator subsets. We introduce STABLEVAL, a disagreement-aware evaluation framework that models latent item correctness and annotator-specific confusion patterns to produce posterior expected item credit and calibrated agent-level scores. Unlike label-denoising approaches such as Dawid-Skene, STABLEVAL is explicitly designed for stable and uncertainty-aware system evaluation rather than hard label recovery. We formalize ranking stability as a first-class evaluation objective and analyze how aggregation methods preserve or distort underlying annotator behavior. Across controlled synthetic experiments and multiple real-world human-annotated benchmarks, majority vote exhibits increasing score error and ranking instability under annotator heterogeneity and adversarial noise, while STABLEVAL yields more stable and statistically grounded system rankings. These results demonstrate that modeling disagreement is essential for robust and reproducible AI evaluation.

Akash Bonagiri, Gerard Janno Anderias, Saee Patil +6

Department of Computer Science, University of California, Davis, Davis, CA, USA.

Annotator Disagreement AI Agent Evaluation Human-in-the-Loop Evaluation

May 3, 2026cs.CY

Preregistration for Experiments with AI Agents

The proliferation of large language models (LLMs) and autonomous AI agents has given rise to a rapidly growing methodological paradigm: "in silico" behavioral experiments. Originally conceived as a way to use AI agents as proxies for human participants in studies of cognition, decision-making, and social dynamics, this approach has taken on new significance -- as AI agents increasingly negotiate, transact, and make consequential decisions on behalf of people and organizations, understanding their behavior has become a research priority in its own right. While these experiments with AI agents offer unprecedented advantages in terms of scalability, cost efficiency, and experimental control, they also inherit, and in some cases amplify, methodological vulnerabilities that have long plagued human subjects research. To address these issues, this paper argues that preregistration practices -- central to improving the credibility of human subjects experiments -- should now be extended to experiments with AI agents. We systematically catalog the researcher degrees of freedom that experiments with AI agents introduce -- model selection, prompt wording, settings, and outcome-contingent redesign, for example -- and show how the low cost of iteration and lack of reporting norms make these choices both easy to exploit and difficult to detect. We propose a preregistration template tailored to experiments with AI agents and call on conferences, journals, and funding agencies to make preregistration standard practice for this emerging research paradigm.

Michelle Vaccaro

Institute for Data, Systems & Society (IDSS), Massachusetts Institute of Technology, Cambridge, MA, United States.

AI Agent Evaluation

May 2, 2026cs.AI

Evaluating Agentic AI in the Wild: Failure Modes, Drift Patterns, and a Production Evaluation Framework

Existing evaluation frameworks for large language models -- including HELM, MT-Bench, AgentBench, and BIG-bench -- are designed for controlled, single-session, lab-scale settings. They do not address the evaluation challenges that emerge when agentic AI systems operate continuously in production: compounding decision errors, tool failure cascades, non-deterministic output drift, and the absence of ground truth for long-horizon tasks. This paper makes three contributions. First, we present a taxonomy of seven failure modes unique to production agentic systems, each grounded in observations from systems operating at billion-event scale. Second, we demonstrate empirically where standard metrics -- ROUGE, BERTScore, accuracy/AUC, and the agentic benchmarks above -- fail to detect each failure mode. Third, we propose PAEF (Production Agentic Evaluation Framework), a five-dimension evaluation framework with an open-source reference implementation, designed for continuous evaluation on production traffic rather than episodic benchmark runs. Our analysis shows that standard metrics fail to detect four of the seven failure modes entirely and detect three others only after a lag of multiple evaluation cycles.

Mukund Pandey

Independent Researcher

AI Agent Reliability AI Agent Evaluation AI Agent Monitoring LLM Agent Reliability

May 1, 2026cs.MA

Foresight Arena: An On-Chain Benchmark for Evaluating AI Forecasting Agents

Evaluating the true forecasting ability of AI agents requires environments that are resistant to environments resistant to overfitting, free from centralized trust, and grounded in incentive-compatible scoring. Existing benchmarks either rely on static datasets vulnerable to training-data contamination, or measure trading PnL -- a metric conflating predictive accuracy with timing, sizing, and risk appetite. We introduce Foresight Arena, the first permissionless, on-chain benchmark for evaluating AI forecasting agents on real-world prediction markets. Agents submit probabilistic forecasts on binary Polymarket markets via a commit-reveal protocol enforced by Solidity smart contracts on Polygon PoS; outcomes are resolved trustlessly through the Gnosis Conditional Token Framework. Performance is measured by the Brier Score and a novel Alpha Score -- proper scoring rules that incentivize honest probability reporting and isolate predictive edge over market consensus. We provide a formal analysis: closed-form variance for per-market Alpha, the connection to Murphy's classical Brier decomposition, and a power analysis characterizing the number of rounds required to reliably distinguish agents of different skill levels. We show that detecting a true edge of

α^* = 0.02

at 80% power requires approximately 350 resolved binary predictions (50 rounds of 7 markets), while

α^* = 0.01

requires four times more. We complement these analytical results with a deterministic, seed-controlled simulation study calibrated to literature-reported Brier-score ranges, illustrating how Murphy decomposition distinguishes well-calibrated agents from market-tracking agents that fail through reduced resolution. Live results from the deployed benchmark will be reported in a future revision. All smart contracts and evaluation infrastructure are open-source.

Maksym Nechepurenko, Pavel Shuvalov

Director of Research, Devnull, Dubai, UAE · Chief Technology Officer, Devnull, Dubai, UAE

AI Agent Evaluation Forecasting Benchmarks Probabilistic Forecasting Proper Scoring Rules Prediction Markets

May 1, 2026cs.CL

Agentic AI for Substance Use Education: Integrating Regulatory and Scientific Knowledge Sources

The delivery of traditional substance education has remained problematic due to challenges in scalability, personalization, and the currency of information in a rapidly evolving substance use landscape. While artificial intelligence (AI) offers a promising frontier for enhancing educational delivery, its application in providing real-time, authoritative substance use education remains largely underexplored. We built an agentic-based AI web application that combined Drug Enforcement Administration records with peer-reviewed literature in real-time to provide transparent context-sensitive substance use education. The system uses retrieval-augmented generation with a carefully filtered corpus of 102 documents and dynamic PubMed queries. Document storage was semantically chunked and placed in a vector representation in order to be easily retrieved. We conducted an expert evaluation study in which a panel of five subject matter experts generated 30 domain-specific questions, and two independent raters assessed 90 system interactions (30 primary questions plus two contextual follow-ups each) using a five-point Likert scale across four criteria: factual accuracy, citation quality, contextual coherence, and regulatory appropriateness. Mean ratings ranged from 4.18 to 4.35 across the four criteria (overall category range: 4.05-4.52), with substantial inter-rater agreement (Cohen's kappa = 0.78). These findings suggest that agentic AI architectures integrating authoritative regulatory sources with real-time scientific literature represent a promising direction for scalable, accurate, and verifiable health education delivery, warranting further evaluation through longitudinal user studies.

Kosar Haghani, Zahra Kolagar, Mohammed Atiquzzaman

Department of Social Science, Texas Woman's University, Denton, Texas, USA. · Department of Computer Science, Augsburg University of Applied Sciences, Augsburg, Germany · School of Computer Science, University of Oklahoma, Norman, Oklahoma, USA

AI Agent Evaluation Generative AI in Education Agentic RAG

Apr 30, 2026cs.AI

WindowsWorld: A Process-Centric Benchmark of Autonomous GUI Agents in Professional Cross-Application Environments

While GUI agents have shown impressive capabilities in common computer-use tasks such as OSWorld, current benchmarks mainly focus on isolated and single-application tasks. This overlooks a critical real-world requirement of coordinating across multiple applications to accomplish complex profession-specific workflows. To bridge this gap, we present a computer-use benchmark in cross-application workflows, named WindowsWorld, designed to systematically assess GUI Agents on complex multi-step tasks that mirror real-world professional activities. Our methodology uses a multi-agent framework steered by 16 occupations to generate four difficulty-level tasks with intermediate inspection, which are then refined by human review and executed in a simulated environment. The resulting benchmark contains 181 tasks with an average of 5.0 sub-goals across 17 common desktop applications, of which 78% are inherently multi-application. Experimental results of leading large models and agents show that: 1) All computer-use agents perform poorly on multi-application tasks (< 21% success rate), far below the performance of simple single-app tasks; 2) They largely fail at tasks requiring conditional judgment and reasoning across

\geq

3 applications, stalling at early sub-goals; 3) Low execution efficiency, where tasks often fail despite far exceeding human step limits. Code, benchmark data, and evaluation resources are available at github.com/HITsz-TMG/WindowsWorld.

Jinchao Li, Yunxin Li, Chenrui Zhao +3

Harbin Institute of Technology, Shenzhen · Shenzhen Loop Area Institute

Computer-Use Agent Benchmarks Computer-Use Agents GUI Agents AI Agent Evaluation AI Agent Benchmarks Workflow Automation

Apr 30, 2026cs.AI

End-to-End Evaluation and Governance of an EHR-Embedded AI Agent for Clinicians

Clinical AI systems require not just point-in-time evaluation but continuous governance: the ongoing practice of monitoring, evaluating, iterating, and re-evaluating performance throughout deployment. We present an end-to-end framework of governance that integrates rubric validation, live deployment feedback, technical performance monitoring, and cost tracking, with controlled experimentation gating system changes before deployment. Applied to Hyperscribe, an EHR-embedded agent that converts ambient audio into structured chart updates, twenty clinicians authored 1,646 validated rubrics across 823 cases. Seven Hyperscribe versions were evaluated through controlled experiments, with median scores improving from 84% to 95%. Analysis of 107 live feedback entries over three months showed feedback composition shifting from 79% error reports and 14% positive observations to 30% errors and 45% positive observations as engineering interventions resolved failures. Median processing time per audio segment was 8.1 seconds with a 99.6% effective completion rate after retry mechanisms absorbed transient model errors. These results demonstrate that continuous, multi-channel governance of deployed clinical AI is both achievable and effective.

Aaryan Shah, Andrew Hines, Alexia Downs +6

1Canvas Medical, San Francisco, CA, USA · Department of Biomedical Data Science, Stanford University, Stanford, CA, USA · 3XPC (X Primary Care) +2

Healthcare AI Agent Evaluation Human-in-the-Loop Evaluation AI Agent Monitoring AI Agent Governance Electronic Health Records

Apr 29, 2026cs.MA

The Cost of Consensus: Isolated Self-Correction Prevails Over Unguided Homogeneous Multi-Agent Debate

Multi-agent debate, where teams of LLMs iteratively exchange rationales and vote on answers, is widely deployed under the assumption that peer review filters hallucinations. Yet the failure dynamics of homogeneous debate remain poorly understood, therefore we report findings from a controlled empirical study of teams of

N{=}10

homogeneous agents (Qwen2.5-7B, Llama-3.1-8B, Ministral-3-8B) across

R{=}3

debate rounds on two high-difficulty benchmarks (GSM-Hard and MMLU-Hard). We compare peer debate against isolated self-correction and a stochastic noise control that injects rationales from unrelated problems. We decompose debate failure into three model-dependent pathways: sycophantic conformity, where agents uncritically adopt majority answers (modal adoption up to 85.5%); contextual fragility, where peer rationales destabilize previously correct reasoning (vulnerability rate up to 70.0%); and consensus collapse, where plurality voting discards correct answers already present in the generation pool (oracle gap up to 32.3 percentage points). Ablations over communication density (

K \in \{2,4,9\}

) and sampling temperature (

T \in \{0.4, 0.7\}

) show that conformity reaches high levels at minimal peer exposure (

K{=}2

) and intensifies with greater initial diversity. Across all configurations, debate consumes 2.1-3.4

\times

more tokens (up to 28,631 tokens per problem) than self-correction for equal or lower accuracy. Our results indicate that, within the 7-8B parameter class, homogeneous teams without structured roles do not benefit from unguided peer exchange, and that isolated self-correction consistently offers a more favorable cost-accuracy tradeoff.

Blaž Bertalanič, Carolina Fortuna

Jožef Stefan Institute Ljubljana, Slovenia

LLM Self-Correction AI Agent Evaluation Multi-Agent Debate Multi-Agent Consensus