AI Agent Evaluation

We present DeepSpeak-Agentic, a dataset of videos comprising over 37 hours of semi-structured conversations between a human and an embodied AI agent. We use this dataset to evaluate the automatic forensic identification (audio, video, or text) of AI agents, study the nature of human-agent interactions, and provide a benchmark for future advances in the large-language models and AI-generated voices and faces that power embodied AI agents. We also contribute a scalable data-capture system that creates agents, automatically pairs them with human crowd workers, records audiovisual conversations across specified scenarios, and identifies and separates the human and agent in the combined stream.

While household robots are often evaluated based on task completion, everyday domestic environments involve value-conflicting situations where robots are expected to choose actions that prioritize diverse values such as human autonomy, efficiency, or social appropriateness. Yet, there are no benchmarks for evaluating robots' value preferences in such scenarios. We introduce RobotValues, a benchmark to evaluate household robot planners in 8K value-conflict scenarios. Each instance consists of a realistic, synthetically generated household image with multiple plausible robot actions that prioritize different human values. We construct ROBOTVALUES through LLM-assisted scenario generation, stakeholder-grounded value extraction, image generation and automatic quality control. We evaluate 10 VLMs used in robotics and find that models show default value preferences, including safety and accommodation, while underselecting privacy-prioritizing actions. When models are prompted to prioritize values that conflict with their preferences, models often fail to override the default actions, choosing incorrect actions 80% of the time on average across models. These findings highlight the need to go beyond task completion or safety evaluations and assess robots' decision-making capability when human values conflict.

Language agents spend substantial inference time solving individual tasks, yet the experience acquired in one episode is often underutilized in future episodes. Continual learning expects an agent to accumulate reusable experience across a stream of tasks, improve over time, and avoid interference from irrelevant experiences. Unfortunately, existing benchmarks struggle to evaluate continual learning in language agents rigorously. Most efforts focus on retrieval and reasoning over long-context conversations or documents, while recent lifelong-adaptation benchmarks often rely on naive task streams with limited analysis of cross-task relationships, making it difficult to understand what an agent learns and reuses over time. This paper presents an evaluation framework AgentCL for continual learning in agents, centered on controlled task streams and metrics for transfer gains. AgentCL constructs compositional streams where earlier sub-solutions, evidence, or workflows are intentionally reusable in later tasks, and contrasts them with naive streams where such reusability is not guaranteed. We use the benchmark to evaluate non-parametric memory designs for continual learning. To diagnose how memory design choices affect continual learning, we develop MemProbe, a probing method that stores interactions, insights, and skills, while filtering unreliable experiences during consolidation. Empirical analysis across coding, deep research, and language understanding/reasoning tasks shows that naive streams offer limited ability to distinguish memory designs, whereas controlled streams more clearly distinguish their plasticity. Meanwhile, naive and held-out settings often yield limited gains and can expose memory-induced degradation. These results highlight the need for stronger memory designs that balance plasticity and stable reuse.

Multimodal agents are increasingly expected to operate interfaces on behalf of users, raising a central deployment question: can they truly substitute for humans in workflows that services deliberately protect against automation? CAPTCHA verification makes this question concrete. It is not merely a visual puzzle, but a human-verification boundary placed before account creation, content access, form submission, and other protected actions. We introduce \textbf{Humanity's Last Line of Verification (HLL)}, a controlled benchmark that uses interactive CAPTCHA verification to evaluate whether agents can cross this boundary through grounded, human-like interaction rather than recognition alone. HLL covers diverse CAPTCHA interactions and exposes agents to controlled realism stressors, including cluttered webpages, harder task variants, and trace-conditioned validation of the solving process. We evaluate eight frontier multimodal agents in a closed-loop GUI environment. The results show that current agents remain brittle at this human-substitution boundary: performance varies sharply across verification types, degrades under realistic interface conditions, and drops further when correct answers must be supported by valid action traces. By exposing gaps in localization, action calibration, state tracking, and process consistency, HLL provides a concrete testbed for measuring how close multimodal agents are to acting as human substitutes in protected real-world workflows. Our code is available at https://github.com/XinhaoS0101/HLL

Autonomous LLM agents increasingly operate in stateful environments where they access tools, files, memory, and external services. While such capabilities enable complex real-world workflows, they also introduce security risks that are difficult to capture with existing evaluations. Current agent security benchmarks often rely on manually curated tasks, provide limited coverage of emerging threats, and focus primarily on final outcomes rather than the execution processes that lead to unsafe behavior. We introduce SeClaw, a framework that combines specification-driven security task synthesis with execution-based security evaluation for Autonomous agents. Spec-driven security task synthesis enables scalable and controllable construction of security tasks from structured risk specifications, while SeClaw docker provides a standardized testbed for evaluating agent behavior under diverse safety-risk scenarios. The benchmark covers risks arising from resources, user tasks, environments, and intrinsic agent behaviors, and supports trajectory-aware assessment of unsafe actions beyond final responses. By bridging systematic task synthesis and reproducible security evaluation, SeClaw provides a practical foundation for measuring, diagnosing, and comparing security failures in autonomous LLM agents. The code is available at https://github.com/seclaw-eval/seclaw-eval.

Enterprise AI systems that translate natural language into SQL queries and orchestrate multi-step agentic reasoning pipelines require evaluation approaches fundamentally different from academic benchmarks. Spider and BIRD established execution-accuracy protocols; G-Eval and RAGAS advanced LLM-based assessment; and recent work such as Spider 2.0, BEAVER, and BIRD-Interact has begun to address enterprise and agentic dimensions. No single framework unifies text-to-SQL assessment with agentic behavior evaluation into a production-grade pipeline calibrated against human expert judgment. We present BADGER, developed at Merkle, a unified evaluation framework integrating text-to-SQL assessment with agentic behavior evaluation. BADGER offers three contributions. First, LLM-assisted SQL component extraction extending Spider methodology to handle CTE-heavy, dialect-specific SQL. Second, a hybrid execution accuracy metric (Hybrid-EX) resolving column-aliasing and numeric-tolerance brittleness by using an LLM to infer structural alignments before deterministic cell-level scoring. Validated on 150 human-annotated industry queries, Hybrid-EX achieves Cohen's kappa=0.717 [95% CI: 0.600-0.822] (Substantial agreement) and 87.3% balanced accuracy, outperforming all six competing frameworks (Delta-kappa: 0.322-0.502, all p<=0.001). Third, an enterprise agentic evaluation suite assembling RAGAS, G-Eval, and agent benchmark metrics into a unified pipeline; Excess Tool Usage is the sole novel element. BADGER runs entirely within the client's governed data environment, supports configurable LLM judge backends, and enables rapid prototyping of client-specific judges and metrics, serving as a continuous evaluation backbone rather than a one-time quality gate.

Autonomous agents are increasingly expected to support end-to-end medical-AI research workflows, moving beyond isolated prediction tasks or short-form clinical question answering. However, existing medical agent benchmarks primarily evaluate final outputs, providing limited visibility into agent behavior within the research process. To address this gap, we present AutoMedBench, a workflow-aware benchmark for autonomous medical-AI research across diverse medical imaging and multimodal inference tasks, organizing agent execution into a unified five-stage workflow (S1-S5): Plan, Setup, Validate, Inference, and Submit. It comprises long-horizon tasks with each run averaging 33 agent turns, spanning five research tracks: segmentation, image enhancement, visual question answering (VQA), report generation, and lesion detection. Each task is evaluated under two difficulty tiers, Lite and Standard, which use the same data and metrics but differ in the amount of task-brief scaffolding, and each run is scored using both final task performance and S1-S5 stage scores, enabling stage-level analysis from the initial task brief to the final submitted artifact. Across thousands of recorded runs, stage-level scoring reveals that Validate is the weakest workflow stage on average, whereas Setup is the strongest, suggesting that current agents are better at making pipelines executable than at verifying their reliability. Post-run error analysis further shows that verification and submission failures dominate tagged errors, accounting for 37.7% and 38.1% of fired codes respectively, whereas task-understanding errors are rare at 0.9%, and runs with one fired error code have a 48% lower overall score than runs with no error code on average.

Agent benchmarks increasingly record rich interaction trajectories, yet evaluation often reduces each rollout to a pass rate or reward score. We introduce TraceGraph, a graph-based framework that turns released multi-model agent trajectories into shared decision landscapes. For each task, TraceGraph builds a graph over observable action-observation states from pooled rollouts before model identity is introduced. It then overlays outcome-informed productive cores and trap regions, and summarizes each rollout with three events: Access, Trap exposure, and Repair. Across trajectories spanning five benchmark splits, TraceGraph profiles reveal navigation differences hidden by aggregate scores and show that splits differ in whether they reward avoiding traps or recovering from them. The same TraceGraph landscape also motivates a trap-aware recovery pipeline for SWE-bench: aruntime detector fires on states matching historical trap regions, then lightweight continuation policies are evaluated from the same prefix. On fired states, the best pooled single-factor policy raises official resolved rate from 40.4% to 43.5% on the per-provider fired subset and from 41.0% to 44.8% on common-fired instances, with provider-specific active components. Overall, TraceGraph provides a process vocabulary for asking what agent benchmarks test, where models diverge on a shared landscape, and how failure regions can guide downstream improvement.

AI agents are rapidly evolving into autonomous systems capable of sustained interaction, tool use, and long-term collaboration. Yet their real-world adoption remains limited, suggesting that the key barrier lies not only in technical capability but also in a lack of design knowledge for successful human-agent interaction. This position paper argues that AI agents should not be solely evaluated or deployed based on autonomous task capability alone; because agents interact with, adapt to, influence, and sometimes fail humans, human-agent interaction must be treated as a core design and evaluation target for agentic AI. We present 14 design principles that articulate the ideal human-agent relationship across four interaction stages: initially, during interaction, over time, and when things go wrong. We use these principles to evaluate nine agent systems to illustrate that these design principles can provide actionable guidance for AI design teams to systematically design and evaluate agents that are usable, trustworthy, and effective in real-world interactive settings.

We introduce Gram, an automated alignment auditing framework to assess the propensity of AI agents to engage in sabotage. We evaluate Gemini models across 17 simulated agentic deployment scenarios that incentivize sabotage. We find Gemini models misbehave in about 2-3% of our simulated trajectories. Many of these cases are explained by "overeagerness" in Gemini models resulting in both excessive role-playing and goal-seeking behavior. In contrast to other alignment auditing approaches, Gram is designed to specifically evaluate misalignment and intentional sabotage in agentic coding and research agents. We additionally introduce an experimental investigator agent pipeline which enables fine-grained targeted experiments to identify the drivers of misbehavior. We find that increasing realism of environments and removing nudges to misbehave tends to reduce sabotage rates close to zero.

Large language models in Agentic AI systems consume tool schemas and execution results and emit tool invocations as structured data. The default language for that exchange, JSON, was designed for application-to-application interchange rather than token efficiency, so its structural elements impose substantial token overhead. Recent work proposes token-optimized alternatives such as TOON (Token-Oriented Object Notation) and TRON (Token Reduced Object Notation) as more compact replacements, but these formats have been evaluated only on isolated comprehension or generation tasks. Whether their token reductions hold inside end-to-end agentic loops therefore remains an open question. We evaluate TOON and TRON on four agentic benchmarks (BFCL, MCPToolBenchPP, MCP-Universe, StableToolBench) and five open-weight LLMs, decoupling input compression from output compression to measure comprehension and generation independently. TRON reduces tokens by up to 27% with accuracy within 14pp of the JSON baseline. TOON achieves up to 18% reduction at a similar 9pp accuracy cost, but additionally cascades on multi-turn parsing failures and collapses parallel tool-call output for most models. The code is available at: https://github.com/lkutschka/notation-matters

Multimodal large language models are increasingly deployed as long-horizon agents, where memory must do more than recall: it must track an evolving world, revise what has gone stale, and surface the right evidence at decision time. Existing benchmarks measure recall over static dialogue, collapse memory into a single end-of-task accuracy, and reduce visual observations to captions, leaving us unable to localize failures to writing, maintenance, retrieval, or use. The rise of agent harnesses that author their own memory sharpens this gap, since we have no principled way to compare hand-designed pipelines with self-managing alternatives. To close these gaps, we formulate multimodal agent memory as an Action-World Interaction Loop with an observable four-stage lifecycle, and instantiate it in WorldMemArena: 400 multi-session multimodal tasks spanning Lifelong Evolution (evolving personal and task states) and Agentic Execution (memory from real observations, actions, and feedback), annotated with gold memory points, updates, distractors, and evidence chains for stage-level diagnosis. This enables the first head-to-head comparison of long-context, manually designed (RAG and external memory systems), and harness-based memory agents. Results show that: (1) better memory writing and storage do not guarantee better performance; (2) multimodal memory still struggles to fully use visual evidence; (3) systems are unstable across domains and degrade on realistic agentic trajectories; and (4) harness memory is more flexible but remains costly and less reliable.

DeFi investment agents, systems that use AI for autonomous on-chain trading, have attained over USD 3 billion in combined token valuations since late 2024. We survey over 1,900 AI-tagged crypto projects, filter to investment-focused agents, and curate 10 representative projects spanning strategy and observability dimensions. We then conduct a deep-dive architectural analysis of two prominent agent frameworks, ElizaOS and Virtuals Protocol, and a quantitative on-chain performance analysis of 11 Solana-based agent treasuries with publicly attributable trading activity, covering 925,323 token holders. We find that current deployments remain early and heterogeneous: (1) in our sample, many projects do not yet provide clear evidence of autonomous trade execution, and developer interviews suggest that many visible deployments remain basic API integrations; (2) agent treasuries retain over USD 30M in paper gains while token holders collectively lost USD 191.7M, with the top 1% of wallets capturing 81.4% of all gains (USD 1.81B); (3) token valuations are weakly connected to treasury fundamentals, with market-cap-to-AUM ratios exceeding 10,000x versus below 1x for established DeFi protocols; and (4) aggregate user gains peaked at USD 2.4B before declining to net losses, with median returns negative on every platform and tokens declining 93% on average from all-time highs. We interpret these outcomes as characteristic of a permissionless, first-generation market in which open infrastructure enables rapid experimentation but also allows naive or speculative agents to launch before robust standards for autonomy, performance, and stakeholder alignment emerge. We therefore propose a maturity framework along three dimensions: autonomous execution, risk-adjusted profitability, and stakeholder alignment, to characterize the gap between current deployments and future investment-grade agent systems.

We report a comparison of two state-of-the-art agentic AI systems, Claude Code (Anthropic) and Codex (OpenAI), tasked with autonomously executing a simple end-to-end gravitational wave data analysis pipeline on a shared computing infrastructure without human intervention. The pipeline comprises power spectral density estimation from raw Einstein Telescope simulated noise, geometric template bank generation, matched filter recovery of 100 binary black hole signal injections, automated results generation, and large language model-assisted production of a manuscript formatted in the style of Physical Review D. Both agents received identical written specifications and identical compute resources. The experiment was run twice: a first run with unrealistically loud injections, and a second run with signals rescaled to a physically motivated SNR range. The scientific results converged in both runs. However, the agents exhibited substantially different behaviors and computational costs: Claude Code completed the pipeline in ~3.4 minutes with silent deviations from the specification, while Codex required ~16 minutes across explicit self-correcting restarts, including an unsolicited performance optimization of the matched filter inner loop. The autonomously generated manuscripts also diverged in length, details, and quality. In the second run, a subtle difference in the interpretation of the SNR range instruction led to a genuine scientific divergence: Claude Code silently reinterpreted the instructions, while Codex followed the specification literally. We discuss the implications of these behavioral differences, such as speed versus auditability, silent versus transparent error handling, instruction interpretation, and the criticality of intermediate data representations in multi-model pipelines, for the deployment of agentic AI in scientific computing workflows.

AI research agents now support large-scale AI-assisted scientific discovery. We examine whether AI-generated ideas broaden scientific exploration or primarily reinforce existing work. Using five agent frameworks and five large language models, we generate 219,655 ideas for different scientific fields. Across experiments, four consistent patterns emerge. First, AI-generated ideas are more concentrated than human-authored papers within the same research area. Second, they remain much closer to starting literature than later human follow-on work does. Third, AI-generated ideas align less with future human research. Last, AI-generated ideas are located in lower-impact regions of the historical scientific landscape. Overall, current AI research agents appear better suited to local elaboration than to broadening scientific exploration.

As AI agents increasingly operate in open, real-world environments, they require a deep synergy of multimodal perception, tool invocation with multi-hop reasoning, and dynamic interaction with users. However, existing benchmarks fail to jointly evaluate these capabilities due to challenges in designing strictly coupled multi-capability tasks, simulating natural and task-constrained user feedback, and ensuring objective evaluation of dynamic interaction. To bridge this gap, we introduce EgoBench, the first interactive multimodal benchmark for tool-using agents. EgoBench comprises 1,590 egocentric-video-grounded tasks covering five daily scenarios, along with a user-agent-tool interactive environment for evaluation. We implement a three-stage synergistic pipeline through which each task is designed to enforce the joint application of visual perception and tool-augmented multi-hop reasoning. We additionally develop a multi-agent simulated user to evaluate agents' interaction capabilities, which generates high-fidelity, task-aligned responses to agents. Furthermore, we establish a deterministic joint validation framework that guarantees objective assessment through process-based and result-based equivalence. Benchmarking eight video-MLLM agents on EgoBench reveals a severe performance ceiling: the best-performing model achieves only a 34.95% average Joint Success Rate across the three interaction modes. Finally, we conduct a multi-dimensional error analysis to disentangle failure modes, exposing capability bottlenecks for advancing future AI agents.

Large language models (LLMs) have evolved into interactive agents that collaborate with users in real-world tasks. Effective collaboration in such settings increasingly depends on understanding the user beyond what is explicitly stated, as user intent is often reflected in fragmented daily interactions and requires both personalized modeling and proactive interaction. However, existing agent benchmarks primarily evaluate reasoning and tool use, largely overlooking the challenges of inferring and leveraging user preferences in realistic scenarios. To address this gap, we introduce VitaBench 2.0, a benchmark for evaluating personalized and proactive agent behavior in long-term user interactions. In VitaBench 2.0, tasks are organized as temporally ordered sequences for individual users, where preferences are embedded in fragmented and heterogeneous interactions. Successful completion of tasks requires the agent to continuously extract, utilize, and update user preferences from these interactions. We further evaluate proactiveness through tasks that require agents to recognize missing information and actively acquire it from users or environments before making decisions. To support systematic analysis, we provide an extensible memory interface that enables controlled comparison across different memory architectures. We benchmark a diverse set of frontier proprietary and open-source LLMs. Results show that real-world personalization remains highly challenging even for state-of-the-art models, revealing a substantial gap between current capabilities and practical requirements. Extensive analysis further reveals the failure modes and capability bottlenecks of current agents in real-world personalized decision-making, providing insights for future model improvements.

While AI agents demonstrate remarkable capabilities in reasoning and tool use, they remain fundamentally reactive: they compute responses only after explicit user prompts. This paradigm ignores a critical opportunity: the idle time between interactions is largely wasted, leaving agents unable to prepare for future user needs. To bridge this gap, we introduce ProAct, a proactive agent architecture that leverages idle-time compute to anticipate and fulfill likely upcoming user needs. By analyzing evolving dialogue history together with persistent memory, ProAct predicts upcoming needs and iteratively acquires information, allowing the agent to resolve knowledge gaps and prepare evidence before the user initiates a query. To rigorously evaluate proactive capabilities, we also introduce ProActEval, a comprehensive benchmark comprising 200 scenarios across 40 domains, featuring predictable need chains and diverse user cognitive profiles. Empirical results demonstrate significant advantages over reactive baselines. ProAct accelerates task completion by reducing required turns by 14.8%, decreases user effort by 11.7%, and cuts hallucination rates by 28.1% on ProActEval. Furthermore, MemBench evaluations confirm that ProAct achieves state-of-the-art reflective accuracy, underscoring its sustained and robust performance.

We systematically investigate all 25 public ARC-AGI-3 games and find that every one is reachable through non-intelligent strategies: 10 in a single blind step, 5 after one probing action, 1 via repeated ACTION1 presses, 1 via diverse exploration, and 8 via single repeated actions with sufficient budget (50-200 steps). A library-level null-coordinate vulnerability additionally bypasses 18 games in 1 step. This benchmark critique implies the public evaluation set cannot discriminate intelligent exploration from trivial heuristics - the private 55-game evaluation is the only genuine intelligence test. Against this backdrop, we present AERA (Adaptive Epistemic Reasoning Agent), a three-phase (EXPLORE / VERIFY / PLAN) agent achieving RHAE=0.2116 (4/25 solved) on these 25 games with Qwen2.5-0.5B, while random and no-explore baselines score 0.0000. We formalise AERA through a Speed--Depth trade-off framework: under a convexity assumption (proved for a class of environments in the Appendix), RHAE's quadratic form emerges as a second-order penalty for deviating from the Pareto frontier between action efficiency and information gain. Contributions: (i) a benchmark validity analysis showing that current interactive reasoning benchmarks fail to measure the exploration they claim to require, and (ii) the EXPLORE-before-PLAN framework and model-capability x exploration interaction. The linked code track entry achieves RHAE=0.30 on the full 55-game private evaluation. Code: CC0.

Agent-to-Agent (A2A) networks enable autonomous AI agents to collaborate by sharing reusable problem-solving instructions. However, how these decentralized ecosystems operate in practice remains largely unexplored. We present the first large-scale empirical study of EvoMap, a prominent A2A collaboration network. By analyzing over 1.5M assets and 128K agents, we show how design choices that prioritize scalable growth introduce trade-offs in reusability, evolution, and auditability. First, EvoMap's credit economy rewards agents for publishing valuable assets. Although this design encourages participation at scale, rewards are tied primarily to publication rather than adoption. This leads agents to mass-produce assets to accumulate credits. As a result, 98% of assets are never reused, while rewards become highly concentrated among a small fraction of agents. Second, EvoMap employs an algorithm (referred to as GDI) to score and rank the quality of these shared assets. We demonstrate that this scoring system is flawed: rather than measuring objective performance, an asset's rank is heavily dictated by unverified, self-reported metadata (e.g., claimed lines of code modified). This allows agents to trivially manipulate their asset's scores. Finally, EvoMap relies on agents to provide local execution logs as evidence that uploaded assets function correctly. Because these validations are not independently verified, over 84% of approved assets bypass quality checks using vacuous tests (e.g., console$.$log()). Our findings show that future A2A collaboration networks cannot rely on unverified self-reporting alone. Scalable collaboration requires mechanisms that balance open participation with verifiable execution and trustworthy evaluation.

Autonomous AI agents increasingly issue side-effect-bearing actions: database mutations, refunds, payments, external commitments. We propose the Actuarial Action Interface (AAI), a deterministic runtime contract that prices each such action against a contractually fixed safe default under a time-consistent risk mapping, and gates execution against a per-boundary reserve capital budget. We then develop the Authority Frontier, an evaluation primitive measuring how much autonomous authority the runtime releases at each level of reserve capital. The framework provides (i) a deterministic quote-bind-commit protocol with toll-bounded capability tokens; (ii) a universal seven-class action taxonomy mapping heterogeneous tool calls to comparable authority units; (iii) replay determinism and pathwise reserve coverage under alpha-spending; (iv) cross-domain normalization via full reserve demand C_full and capital metrics Capital@k. We instantiate AAI across four agentic environments (database mutation, customer-service refund, and the public tau-bench retail and airline tool-use traces) and report a live Postgres panel in which three Azure-hosted models propose actions through the same contract. The frontier exhibits a common low-reserve refusal and intermediate-release pattern across domains, with saturation only where the budget grid reaches full reserve demand; required reserve capital varies by 22x (Capital@50 from 289 to 6457). The framework does not force domains into the same shape; it surfaces each domain's actuarial geometry. In the live panel the contract prevents realized loss across all three models at low budget while differing in underwriting persistence under denial: model identity is an actuarial underwriting variable. The contribution is a benchmark-ready evaluation framework for runtime actuarial control of autonomous-agent side effects.

Time series research is moving beyond fixed forecasting benchmarks toward realistic tasks that combine prediction, contextual reasoning, tool use, and structured decision support. Most benchmarks are built around clean data and short evaluation loops; agents alone may miss temporal constraints, evidence checks, or review before finalizing outputs. We first formalize next-generation time series tasks as three-component tuples consisting of a task file, a workspace, and a validation interface. We then present AION, a time series harness built from six component groups: agents, skills, rules, memory, evaluation, and protocols. In this harness, we use three design principles: temporal grounding, temporal knowledge-grounded reasoning, and reliability mechanisms such as post-experiment analysis and layered review. A Kaggle Store Sales case study shows that the harness produces more detailed process traces, more artifacts, and more review steps than the same base agent operating in OpenCode direct build mode. Taken together, these results argue for a paradigm shift from fixed tasks to realistic ones under real-world constraints.

We work towards measuring both AI adoption and the capability of AI to perform discrete labor tasks across various occupations. To measure adoption, we develop an open-source economic index that uses publicly available user-LLM chat data and ONET tasks to replicate studies produced by frontier AI labs, finding that occupations in the finance, computer science, and arts sectors are those with the highest adoption rates. To measure capabilities, we build a system that generates benchmark scenarios grounded in ONET occupations, tasks, and model-context-protocol (MCP) servers. We test Kimi-k2.5 with an OpenAI agents SDK harness on scenarios across 9 occupations that appear frequently in our index, finding that AI correctly executes high-level workflows but often errs in the granular details (such as specific tool calls used).

We introduce AvalancheBench, a benchmark for evaluating enterprise data agents through \emph{latent world recovery}. AvalancheBench improves on existing benchmarks in three ways. First, it evaluates analytical understanding rather than pipeline completion: systems are scored on whether they recover the segments, drivers, temporal events, and relationships that explain the data, not merely on whether they execute a workflow or produce a plausible report. Second, it provides ground truth for goal-driven analytics by generating observations from a known latent world, enabling partial credit for incomplete but valid recoveries. Third, it exposes how early analytical mistakes propagate into later conclusions: missed segments, merged events, or wrong attributions can lead to systematically wrong recommendations. In this sense, AvalancheBench complements real-data benchmarks by providing a controlled setting for diagnosing whether agents recover the analytical structure behind enterprise data. On a first e-commerce use case, the strongest configuration of a leading coding agent recovers only 26% of the rubric, with failures concentrated in generic customer segmentations and merged temporal events.

AI agents are entering high-risk production settings, where they use tools, retain context, follow policies, handle private data, and interact with users over multiple turns. Yet many evaluation methods still judge isolated outputs or static tasks, missing failures that emerge through trajectory, pressure, and adversarial interaction. We introduce ProofAgent Harness, open infrastructure for scalable, auditable, and adversarial AI agent evaluation. The harness provides evaluation infrastructure around an agent: it curates evaluation intelligence, runs adversarial multi-turn trials, captures behavioral traces, applies post-hoc multi-juror scoring, resolves disagreement, and produces evidence-linked reports. Its open design allows developers and researchers to extend domains, traps, metrics, juror personas, scoring rules, and reporting formats. At its core is Adversarial Multi-Juror Scoring with Turn-Level Audit, which evaluates completed agent behavior under pressure using calibrated juror personas, consensus checks, and turn-level evidence. Experiments across customer support, medical triage, privacy and security, and code generation agents show that strong agents fail selectively through weak metrics, fragile turns, unsafe reframing, and manipulation paths. We also find that a small quantized local Harness LLM can challenge production agents powered by best-in-class large LLMs, suggesting that evaluation capability emerges from the full harness pipeline rather than model scale alone. ProofAgent Harness turns AI agent evaluation from a static score into scalable adversarial evaluation infrastructure: repeatable, evidence-backed, extensible, and actionable before deployment.

The development of LLM agents has led to a growing body of work on knowledge-work AI, including coding, research, and healthcare. However, current knowledge-work evaluation and benchmark design still largely follow the logic of traditional NLP tasks. As a result, higher benchmark performance does not reliably show that a system can carry out knowledge work in real-world deployment settings. This paper contributes a three-step approach for making explicit how benchmarked tasks represent the work claims attached to their scores: defining the work activity under evaluation, specifying the tested setting, and scoring the appropriate work product. We review work studies showing that knowledge work is organized through roles and responsibilities, local materials and tools, and artifacts that must remain usable in downstream workflows. We then translate these concerns into benchmark design and reporting guidance, covering how tasks should be mapped to work activities, how tested settings should specify materials, tools, roles, and constraints, and how scoring should focus on the work product left by the system. To name the work activity being evaluated and distinguish it from common benchmark tasks, we derive an inventory of 18 work activities from the O{*}NET occupational task database. We demonstrate the approach through three benchmark case analyses: GDPval, a non-code occupational deliverable benchmark; OfficeQA Pro, a grounded document-analysis benchmark scored by final answers; and APEX-SWE, a software-engineering benchmark with executable scored products. These cases show how benchmark design choices shape the strongest work claim a score can support, and where gaps arise between the benchmarked task, tested setting, scored product, and broader work claim.

Empirical claims about autonomous Kubernetes operations agents are largely unfalsifiable. Published work reports observational results without controlled comparisons against an agent-disabled baseline, selection bias is endemic, pre-registered decision matrices are absent, and samples are typically too small for the noise level of the underlying scoring system. The cause is the same gap that limits the agents themselves: code agents have a verification substrate that turns "did it work" into a fast, falsifiable, ground-truth signal, and operations has nothing equivalent. We present agent-breakage, a closed-loop measurement framework that injects faults into a target Kubernetes cluster, observes how an autonomous agent responds, scores the response on four axes against ground truth, and accumulates outcome-labeled (state, action, outcome) tuples. The framework distinguishes framework error from reasoning error, supports a true off-condition control via a deterministic-embedder mechanism, and enforces pre-registered decision matrices. We use it as a case study to test whether retrieval over past postmortems compounds an agent's capability. The methodological payload is three confounds the substrate caught during that case study, each of which would have produced a wrong published claim on a less instrumented version of the same work: a pgvector index bug, a +19% selection-bias artifact, and small-sample estimates that overstated effects by roughly 3x. The retrieval result itself is a partial falsification: 1 of 3 dense-corpus scenarios significant at p<0.05, pooled effect +3.9 percentage points, not significant at n=60. A within-scenario corpus-density sweep at 360 runs shows that mechanistic alignment of near-neighbors dominates raw count. The framework is released open source.

Large language models (LLMs) increasingly excel at mathematical reasoning, but their unreliability limits their utility in mathematics research. A mitigation is using LLMs to generate formal proofs in languages like Lean. We perform the first large-scale evaluation of this method's ability to solve open problems. Our most capable agent autonomously resolved 9 of 353 open Erdős problems at the per-problem cost of a few hundred dollars, proved 44/492 OEIS conjectures, and is being deployed in combinatorics, optimization, graph theory, algebraic geometry, and quantum optics research. A basic agent alternating LLM-based generation with Lean-based verification replicated the Erdős successes but proved costlier on the hardest problems. These findings demonstrate the power of AI-aided formal proof search and shed light on the agent designs that enable it.

The benchmarks used to evaluate AI agents in security-critical roles suffer from crucial weaknesses. Building on recent empirical evidence, we characterize three core challenges that undermine security evaluations: benchmark vulnerabilities, temporal staleness, and runtime uncertainty. We then outline practical directions toward building more robust and trustworthy evaluation frameworks.

Harness optimization enables automated agent creation by having an optimizer agent iteratively update the harness of target agents. Despite its success, current studies evaluate optimizers solely by observing target agents' performance gains. This indirect end-improvement evaluation neglects optimizers' actions at intermediate steps, which are often erroneous and hinder agent performance in deployment. Therefore, it is unclear whether harness optimization is driven by optimizers' informed update actions or simply trial-and-error. This necessitates direct evaluation of harness optimizers. However, evaluating harness optimizers directly is non-trivial and costly due to the lack of oracle harnesses. To address this, we present a simple, low-cost design to directly evaluate them, namely priority ranking. By asking harness optimizers to rank components (e.g., tools) in a given harness by their potential to improve/hinder agent performance when updated, our design quantifies optimizer ability at the step level without expensive rollouts or manual examination. More importantly, optimizers' ranking performance correlates with their ability to improve agents in actual multi-step harness optimization, establishing priority ranking as a reliable indicator of optimization ability. Priority ranking is enabled by SHOR, a collection of 182 human-verified optimization scenarios spanning across domains, designs, and time stages. Code and data: https://github.com/k59118/Harness_Optimizer_Evaluation.