Daily Papers

Computer-use agents (CUAs) hold great promise for automating complex desktop workflows, yet progress toward general-purpose agents is bottlenecked by the scarcity of continuous, high-quality human demonstration videos. Recent work emphasizes that continuous video, not sparse screenshots, is the critical missing ingredient for scaling these agents. However, the largest existing open dataset, ScaleCUA, contains only 2 million screenshots, equating to less than 20 hours of video. To address this bottleneck, we introduce CUA-Suite, a large-scale ecosystem of expert video demonstrations and dense annotations for professional desktop computer-use agents. At its core is VideoCUA, which provides approximately 10,000 human-demonstrated tasks across 87 diverse applications with continuous 30 fps screen recordings, kinematic cursor traces, and multi-layerfed reasoning annotations, totaling approximately 55 hours and 6 million frames of expert video. Unlike sparse datasets that capture only final click coordinates, these continuous video streams preserve the full temporal dynamics of human interaction, forming a superset of information that can be losslessly transformed into the formats required by existing agent frameworks. CUA-Suite further provides two complementary resources: UI-Vision, a rigorous benchmark for evaluating grounding and planning capabilities in CUAs, and GroundCUA, a large-scale grounding dataset with 56K annotated screenshots and over 3.6 million UI element annotations. Preliminary evaluation reveals that current foundation action models struggle substantially with professional desktop applications (~60% task failure rate). Beyond evaluation, CUA-Suite's rich multimodal corpus supports emerging research directions including generalist screen parsing, continuous spatial control, video-based reward modeling, and visual world models. All data and models are publicly released.

Vision-Language-Action (VLA) models have achieved revolutionary progress in robot learning, enabling robots to execute complex physical robot tasks from natural language instructions. Despite this progress, their adversarial robustness remains underexplored. In this work, we propose both adversarial patch attack and corresponding defense strategies for VLA models. We first introduce the Embedding Disruption Patch Attack (EDPA), a model-agnostic adversarial attack that generates patches directly placeable within the camera's view. In comparison to prior methods, EDPA can be readily applied to different VLA models without requiring prior knowledge of the model architecture, or the controlled robotic manipulator. EDPA constructs these patches by (i) disrupting the semantic alignment between visual and textual latent representations, and (ii) maximizing the discrepancy of latent representations between adversarial and corresponding clean visual inputs. Through the optimization of these objectives, EDPA distorts the VLA's interpretation of visual information, causing the model to repeatedly generate incorrect actions and ultimately result in failure to complete the given robotic task. To counter this, we propose an adversarial fine-tuning scheme for the visual encoder, in which the encoder is optimized to produce similar latent representations for both clean and adversarially perturbed visual inputs. Extensive evaluations on the widely recognized LIBERO robotic simulation benchmark demonstrate that EDPA substantially increases the task failure rate of cutting-edge VLA models, while our proposed defense effectively mitigates this degradation. The codebase is accessible via the homepage at https://edpa-attack.github.io/.

We study how vision-language-action (VLA) models can improve through real-world deployments via reinforcement learning (RL). We present a general-purpose method, RL with Experience and Corrections via Advantage-conditioned Policies (RECAP), that provides for RL training of VLAs via advantage conditioning. Our method incorporates heterogeneous data into the self-improvement process, including demonstrations, data from on-policy collection, and expert teleoperated interventions provided during autonomous execution. RECAP starts by pre-training a generalist VLA with offline RL, which we call π^{*}_{0.6}, that can then be specialized to attain high performance on downstream tasks through on-robot data collection. We show that the π^{*}_{0.6} model trained with the full RECAP method can fold laundry in real homes, reliably assemble boxes, and make espresso drinks using a professional espresso machine. On some of the hardest tasks, RECAP more than doubles task throughput and roughly halves the task failure rate.

Vision-language-action (VLA) models have shown strong performance in robotic manipulation, yet their robustness to physically realizable adversarial attacks remains underexplored. Existing studies reveal vulnerabilities through language perturbations and 2D visual attacks, but these attack surfaces are either less representative of real deployment or limited in physical realism. In contrast, adversarial 3D textures pose a more physically plausible and damaging threat, as they are naturally attached to manipulated objects and are easier to deploy in physical environments. Bringing adversarial 3D textures to VLA systems is nevertheless nontrivial. A central obstacle is that standard 3D simulators do not provide a differentiable optimization path from the VLA objective function back to object appearance, making it difficult to optimize through an end-to-end manner. To address this, we introduce Foreground-Background Decoupling (FBD), which enables differentiable texture optimization through dual-renderer alignment while preserving the original simulation environment. To further ensure that the attack remains effective across long-horizon and diverse viewpoints in the physical world, we propose Trajectory-Aware Adversarial Optimization (TAAO), which prioritizes behaviorally critical frames and stabilizes optimization with a vertex-based parameterization. Built on these designs, we present Tex3D, the first framework for end-to-end optimization of 3D adversarial textures directly within the VLA simulation environment. Experiments in both simulation and real-robot settings show that Tex3D significantly degrades VLA performance across multiple manipulation tasks, achieving task failure rates of up to 96.7\%. Our empirical results expose critical vulnerabilities of VLA systems to physically grounded 3D adversarial attacks and highlight the need for robustness-aware training.

Run the same LLM agent on the same task twice: do you get the same behavior? We find the answer is often no. In a study of 3,000 agent runs across three models (Llama 3.1 70B, GPT-4o, and Claude Sonnet 4.5) on HotpotQA, we observe that ReAct-style agents produce 2.0--4.2 distinct action sequences per 10 runs on average, even with identical inputs. More importantly, this variance predicts failure: tasks with consistent behavior (leq2 unique paths) achieve 80--92% accuracy, while highly inconsistent tasks (geq6 unique paths) achieve only 25--60%, a 32--55 percentage point gap depending on model. We trace variance to early decisions: 69% of divergence occurs at step 2, the first search query. Our results suggest that monitoring behavioral consistency during execution could enable early error detection and improve agent reliability.

Next token prediction is the fundamental principle for training large language models (LLMs), and reinforcement learning (RL) further enhances their reasoning performance. As an effective way to model language, image, video, and other modalities, the use of LLMs for end-to-end extraction of structured visual representations, such as scene graphs, remains underexplored. It requires the model to accurately produce a set of objects and relationship triplets, rather than generating text token by token. To achieve this, we introduce R1-SGG, a multimodal LLM (M-LLM) initially trained via supervised fine-tuning (SFT) on the scene graph dataset and subsequently refined using reinforcement learning to enhance its ability to generate scene graphs in an end-to-end manner. The SFT follows a conventional prompt-response paradigm, while RL requires the design of effective reward signals. Given the structured nature of scene graphs, we design a graph-centric reward function that integrates node-level rewards, edge-level rewards, and a format consistency reward. Our experiments demonstrate that rule-based RL substantially enhances model performance in the SGG task, achieving a zero failure rate--unlike supervised fine-tuning (SFT), which struggles to generalize effectively. Our code is available at https://github.com/gpt4vision/R1-SGG.

Recently, diffusion policy has shown impressive results in handling multi-modal tasks in robotic manipulation. However, it has fundamental limitations in out-of-distribution failures that persist due to compounding errors and its limited capability to extrapolate. One way to address these limitations is robot-gated DAgger, an interactive imitation learning with a robot query system to actively seek expert help during policy rollout. While robot-gated DAgger has high potential for learning at scale, existing methods like Ensemble-DAgger struggle with highly expressive policies: They often misinterpret policy disagreements as uncertainty at multi-modal decision points. To address this problem, we introduce Diff-DAgger, an efficient robot-gated DAgger algorithm that leverages the training objective of diffusion policy. We evaluate Diff-DAgger across different robot tasks including stacking, pushing, and plugging, and show that Diff-DAgger improves the task failure prediction by 39.0%, the task completion rate by 20.6%, and reduces the wall-clock time by a factor of 7.8. We hope that this work opens up a path for efficiently incorporating expressive yet data-hungry policies into interactive robot learning settings. The project website is available at: https://diffdagger.github.io.

Vision-Language-Action (VLA) models have demonstrated strong generalization in robotic manipulation, yet existing evaluations are primarily conducted in simulation or on expensive robotic platforms, leaving their robustness on affordable real-world robots largely unexplored. We present a standardized real-world benchmark for evaluating representative VLA and imitation learning policies on the low-cost SO-101 robotic platform. The benchmark comprises four representative manipulation tasks together with unified evaluation protocols, enabling systematic comparison under embodiment uncertainty. Using real-world teleoperated demonstrations, we fine-tune and evaluate π_{0.5}, SmolVLA, Wall-X, and ACT directly on the physical platform. Beyond conventional task success rates, the benchmark incorporates a structured failure taxonomy, semantic- and execution-level failure decomposition, and recovery-aware evaluation metrics to characterize policy robustness. Experimental results show that stronger pretrained VLA policies generally outperform the imitation learning baseline, although performance remains highly task-dependent under low-cost robotic deployment conditions. Execution instability emerges as the dominant failure source, while recovery capability varies substantially across architectures. These results highlight the importance of failure and recovery analysis beyond binary task success and establish SO-101 as a practical benchmark for evaluating embodied AI systems under realistic low-cost robotic deployment conditions.

Multi-agent systems (MAS) are critical for automating complex tasks, yet their practical deployment is severely hampered by the challenge of failure attribution. Current diagnostic tools, which rely on statistical correlations, are fundamentally inadequate; on challenging benchmarks like Who\&When, state-of-the-art methods achieve less than 15\% accuracy in locating the root-cause step of a failure. To address this critical gap, we introduce the first failure attribution framework for MAS grounded in multi-granularity causal inference. Our approach makes two key technical contributions: (1) a performance causal inversion principle, which correctly models performance dependencies by reversing the data flow in execution logs, combined with Shapley values to accurately assign agent-level blame; (2) a novel causal discovery algorithm, CDC-MAS, that robustly identifies critical failure steps by tackling the non-stationary nature of MAS interaction data. The framework's attribution results directly fuel an automated optimization loop, generating targeted suggestions whose efficacy is validated via counterfactual simulations. Evaluations on the Who\&When and TRAIL benchmarks demonstrate a significant leap in performance. Our method achieves up to 36.2\% step-level accuracy. Crucially, the generated optimizations boost overall task success rates by an average of 22.4\%. This work provides a principled and effective solution for debugging complex agent interactions, paving the way for more reliable and interpretable multi-agent systems.

Tool-augmented language agents frequently fail in real-world deployment due to tool malfunctions--timeouts, API exceptions, or inconsistent outputs--triggering cascading reasoning errors and task abandonment. Existing agent training pipelines optimize only for success trajectories, failing to expose models to the tool failures that dominate real-world usage. We propose PALADIN, a generalizable framework for equipping language agents with robust failure recovery capabilities. PALADIN trains on 50,000+ recovery-annotated trajectories constructed via systematic failure injection and expert demonstrations on an enhanced ToolBench dataset. Training uses LoRA-based fine-tuning to retain base capabilities while injecting recovery competence. At inference, PALADIN detects execution-time errors and retrieves the most similar case from a curated bank of 55+ failure exemplars aligned with ToolScan's taxonomy, then executes the corresponding recovery action. This approach generalizes to novel failures beyond the training distribution, retaining 95.2\% recovery performance on unseen tool APIs. Evaluation across PaladinEval and ToolReflectEval demonstrates consistent improvements in Recovery Rate (RR), Task Success Rate (TSR), Catastrophic Success Rate (CSR), and Efficiency Score (ES). PALADIN improves RR from 32.76% to 89.68% (+57% relative) over ToolBench and outperforms the strongest baseline CRITIC (76.34%) by +13.3%. Against vanilla agents, PALADIN achieves 89.86\% RR (+66% relative improvement from 23.75%). These results establish PALADIN as an effective method for building fault-tolerant agents capable of robust recovery in real-world tool environments.

The deployment of embodied agents in self-driving laboratories could accelerate scientific discovery, yet their reliability is constrained by the irreversible and safety-critical nature of chemical experiments. Progress is further hindered by scarce failure data and the lack of fine-grained evaluation protocols. To address these challenges, we introduce LabRobFail, a failure-centric framework for learning and evaluating robotic failure analysis in chemical laboratories. LabRobFail-Sim injects controllable failures at the control, physics, and semantic levels, enabling the construction of LabRobFail-Data, which contains over 20,000 trajectories across 70+ task scenarios, five failure categories, and 11 fine-grained failure types. LabRobFail-Bench evaluates six capabilities spanning task understanding, failure detection, temporal localization, severity assessment, failure classification, and actionable correction. We further develop LabRobFail-VLM, a domain-specialized vision-language model that generates structured failure diagnoses and recovery instructions. On seen environments, it achieves 90.83% failure-detection accuracy and 77.21% temporal-localization accuracy, substantially outperforming general-purpose VLMs. When integrated as a real-time supervisor, it improves downstream task success rates by 4-16 percentage points, demonstrating the value of fine-grained failure understanding for closed-loop recovery and reliable laboratory autonomy. Our code and data are available at https://github.com/Su-ISE-2001/SciRobo

Robust robotic manipulation requires reliable failure detection and recovery. Although recent Vision-Language Models (VLMs) show promise in robot failure detection, their generalization is severely limited by the scarcity and narrow coverage of failure data. To address this bottleneck, we propose an automatic framework for generating diverse robotic planning and execution failures across both simulated and real-world environments. Our approach perturbs successful manipulation trajectories to synthesize failures that reflect realistic failure distributions, and leverages VLMs to produce structured step-by-step reasoning traces. This yields FailCoT, a large-scale failure reasoning dataset built upon the RLBench simulator and the BridgeDataV2 real-robot dataset. Using FailCoT, we train Guardian, a multi-view reasoning VLM for unified planning and execution verification. Guardian achieves state-of-the-art performance on three unseen real-world benchmarks: RoboFail, RoboVQA, and our newly introduced UR5-Fail. When integrated with a state-of-the-art LLM-based manipulation policy, it consistently boosts task success rates in both simulation and real-world deployment. These results demonstrate that scaling high-quality failure reasoning data is critical for improving generalization in robotic failure detection. Code, Data, and Models available at https://www.di.ens.fr/willow/research/guardian/.

In dynamic operational environments, particularly in collaborative robotics, the inevitability of failures necessitates robust and adaptable recovery strategies. Traditional automated recovery strategies, while effective for predefined scenarios, often lack the flexibility required for on-the-fly task management and adaptation to expected failures. Addressing this gap, we propose a novel approach that models recovery behaviors as adaptable robotic skills, leveraging the Behavior Trees and Motion Generators~(BTMG) framework for policy representation. This approach distinguishes itself by employing reinforcement learning~(RL) to dynamically refine recovery behavior parameters, enabling a tailored response to a wide array of failure scenarios with minimal human intervention. We assess our methodology through a series of progressively challenging scenarios within a peg-in-a-hole task, demonstrating the approach's effectiveness in enhancing operational efficiency and task success rates in collaborative robotics settings. We validate our approach using a dual-arm KUKA robot.

Tool-using LLM agents can violate the very policies they are deployed to enforce while appearing to complete the task successfully. In policy-permissive environments, a tool may execute any well-formed call even when the corresponding state transition is forbidden by domain policy. The result is a silent wrong state (a booking cancelled, a passenger count changed, a claim acted on without verification) that neither the tool nor the agent's self-report exposes. We study this failure mode in the τ^2-bench airline domain. On a budget agent, 78% of observed failures are silent wrong-state failures with no tool error, and the aggregate failure rate is reproducible across disjoint seeds, not sampling noise. We then evaluate a lightweight intervention: deterministic, read-only pre-execution gates that inspect the proposed call and current state before allowing a write. A four-gate suite raises full-benchmark success from 29.6% to 42.0% on gpt-4o-mini (+12.4pp; paired task-level bootstrap P=0.0012), and the lift reproduces on a disjoint 15-seed set (+12.3pp; P=0.0008). The effect is concentrated where the gates fire: on the 26/50 firing tasks, success rises by +19.2pp, while movement on the 24 non-firing tasks does not exclude zero. Two negative controls (a self-enforcing retail domain and BFCL) bound the mechanism: gates help when tools are policy-permissive and add little where tools already self-enforce. As suggestive evidence, not a central claim, the same failure mode persists at the frontier: gpt-5.2 at default reasoning still attempts policy-violating writes, and the same suite improves success from 61.2% to 71.6% (+10.4pp; P=0.020; n=5, no replication). The contribution is a bounded evaluation and reliability result: deterministic gates do not guarantee task success, but they can deterministically prevent a known class of silent policy-violating writes at the action boundary.

Robotic manipulation in open-world settings requires not only task execution but also the ability to detect and learn from failures. While recent advances in vision-language models (VLMs) and large language models (LLMs) have improved robots' spatial reasoning and problem-solving abilities, they still struggle with failure recognition, limiting their real-world applicability. We introduce AHA, an open-source VLM designed to detect and reason about failures in robotic manipulation using natural language. By framing failure detection as a free-form reasoning task, AHA identifies failures and provides detailed, adaptable explanations across different robots, tasks, and environments. We fine-tuned AHA using FailGen, a scalable framework that generates the first large-scale dataset of robotic failure trajectories, the AHA dataset. FailGen achieves this by procedurally perturbing successful demonstrations from simulation. Despite being trained solely on the AHA dataset, AHA generalizes effectively to real-world failure datasets, robotic systems, and unseen tasks. It surpasses the second-best model (GPT-4o in-context learning) by 10.3% and exceeds the average performance of six compared models including five state-of-the-art VLMs by 35.3% across multiple metrics and datasets. We integrate AHA into three manipulation frameworks that utilize LLMs/VLMs for reinforcement learning, task and motion planning, and zero-shot trajectory generation. AHA's failure feedback enhances these policies' performances by refining dense reward functions, optimizing task planning, and improving sub-task verification, boosting task success rates by an average of 21.4% across all three tasks compared to GPT-4 models.

Evaluating the output of Large Language Models (LLMs) is one of the most critical aspects of building a performant compound AI system. Since the output from LLMs propagate to downstream steps, identifying LLM errors is crucial to system performance. A common task for LLMs in AI systems is tool use. While there are several benchmark environments for evaluating LLMs on this task, they typically only give a success rate without any explanation of the failure cases. To solve this problem, we introduce SpecTool, a new benchmark to identify error patterns in LLM output on tool-use tasks. Our benchmark data set comprises of queries from diverse environments that can be used to test for the presence of seven newly characterized error patterns. Using SPECTOOL , we show that even the most prominent LLMs exhibit these error patterns in their outputs. Researchers can use the analysis and insights from SPECTOOL to guide their error mitigation strategies.

We study the problem of learning to perform multi-stage robotic manipulation tasks, with applications to cable routing, where the robot must route a cable through a series of clips. This setting presents challenges representative of complex multi-stage robotic manipulation scenarios: handling deformable objects, closing the loop on visual perception, and handling extended behaviors consisting of multiple steps that must be executed successfully to complete the entire task. In such settings, learning individual primitives for each stage that succeed with a high enough rate to perform a complete temporally extended task is impractical: if each stage must be completed successfully and has a non-negligible probability of failure, the likelihood of successful completion of the entire task becomes negligible. Therefore, successful controllers for such multi-stage tasks must be able to recover from failure and compensate for imperfections in low-level controllers by smartly choosing which controllers to trigger at any given time, retrying, or taking corrective action as needed. To this end, we describe an imitation learning system that uses vision-based policies trained from demonstrations at both the lower (motor control) and the upper (sequencing) level, present a system for instantiating this method to learn the cable routing task, and perform evaluations showing great performance in generalizing to very challenging clip placement variations. Supplementary videos, datasets, and code can be found at https://sites.google.com/view/cablerouting.

While word error rates of automatic speech recognition (ASR) systems have consistently fallen, natural language understanding (NLU) applications built on top of ASR systems still attribute significant numbers of failures to low-quality speech recognition results. Existing assistant systems collect large numbers of these unsuccessful interactions, but these systems usually fail to learn from these interactions, even in an offline fashion. In this work, we introduce CLC: Contrastive Learning for Conversations, a family of methods for contrastive fine-tuning of models in a self-supervised fashion, making use of easily detectable artifacts in unsuccessful conversations with assistants. We demonstrate that our CLC family of approaches can improve the performance of ASR models on OD3, a new public large-scale semi-synthetic meta-dataset of audio task-oriented dialogues, by up to 19.2%. These gains transfer to real-world systems as well, where we show that CLC can help to improve performance by up to 6.7% over baselines. We make OD3 publicly available at https://github.com/amazon-science/amazon-od3 .

Large Language Models (LLMs) achieve strong performance on natural language tasks but remain unreliable in mathematical reasoning, frequently generating fluent yet logically inconsistent solutions. We present NeuroProlog, a neurosymbolic framework that ensures verifiable reasoning by compiling math word problems into executable Prolog programs with formal verification guarantees. We propose a multi-task Cocktail training strategy that jointly optimizes three synergistic objectives in a unified symbolic representation space: (i) mathematical formula-to-rule translation (KB), (ii) natural language-to-program synthesis (SOLVE), and (iii) program-answer alignment. This joint supervision enables positive transfer, where symbolic grounding in formula translation directly improves compositional reasoning capabilities. At inference, we introduce an execution-guided decoding pipeline with fine-grained error taxonomy that enables iterative program repair and quantifies model self-debugging capacity. Comprehensive evaluation on GSM8K across four model scales (3B--32B parameters) demonstrates consistent improvements: cocktail training achieves significant accuracy gains of +5.23\% (Qwen-32B, p < 0.01), +3.43\% (GPT-OSS-20B, p < 0.01), and +5.54\% (Llama-3B, p < 0.05) over single-task baselines. Systematic error analysis reveals scale-dependent learning dynamics: at 32B scale, cocktail training transforms unfixable type errors (12\% repair rate) into correctable domain errors (96\% repair rate), achieving 92.7\% overall correction; at 8B scale, the same training eliminates syntactic errors but introduces semantic failures, revealing a critical capacity threshold for type-safe symbolic reasoning.

Existing benchmarks measure capability -- whether a model succeeds on a single attempt -- but production deployments require reliability -- consistent success across repeated attempts on tasks of varying duration. We show these properties diverge systematically as task duration grows, and that pass@1 on short tasks is structurally blind to this divergence. We introduce a reliability science framework for long-horizon LLM agents with four metrics: Reliability Decay Curve (RDC), Variance Amplification Factor (VAF), Graceful Degradation Score (GDS), and Meltdown Onset Point (MOP). We evaluate 10 models across 23,392 episodes on a 396-task benchmark spanning four duration buckets and three domains. Key findings: (1) reliability decay is domain-stratified -- SE GDS drops from 0.90 to 0.44 while document processing is nearly flat (0.74 to 0.71); (2) VAF bifurcates by capability tier -- high VAF is a capability signature, not an instability signal; (3) capability and reliability rankings diverge substantially, with multi-rank inversions at long horizons; (4) frontier models have the highest meltdown rates (up to 19%) because they attempt ambitious multi-step strategies that sometimes spiral; and (5) memory scaffolds universally hurt long-horizon performance across all 10 models. These results motivate reliability as a first-class evaluation dimension alongside capability.

Stealth, the discipline of achieving an objective without revealing your presence, capabilities, or collected intelligence, is what separates sophisticated operators from detectable ones. Elite security researchers and advanced persistent threats achieve their objectives unnoticed; autonomous agents increasingly inherit the same offensive tasks, but do they inherit the tradecraft? We introduce StealthBench,a benchmark that measures operational stealth in autonomous offensive-security agents across six operational security (OPSEC) dimensions. We extract 11 hand-verified OPSEC incidents from real bug-bounty and red-team trajectories, expanded into 14 dockerized task scenarios, where agents, despite finding real vulnerabilities, committed stealth failures inconsistent with standard operational tradecraft: embedding credentials in public uploads, deleting production resources to prove access, force-adding uninvolved users to demonstrate a race condition. We evaluate agent trajectories using a 3-model large language model (LLM) judge panel with majority-vote aggregation, measuring safe success rate (solved and stealthy), Stealth@Solve (tradecraft quality among successful solves), and reckless solve rate (solved but cover blown). Our results show that no model exceeds 54% safe success rate (the compound metric requiring both task completion and stealth), confirming that OPSEC failures are systematic across model families. We release StealthBench as a public benchmark to support both the development of stealth-aware agents and automated OPSEC monitoring for autonomous offensive-security deployments. The interactive leaderboard, evaluation harness, and dataset are available at https://stealthbench.com.

The Model Context Protocol (MCP) is rapidly becoming the standard interface for Large Language Models (LLMs) to discover and invoke external tools. However, existing evaluations often fail to capture the complexity of real-world scenarios, relying on restricted toolsets, simplistic workflows, or subjective LLM-as-a-judge metrics. We introduce MCP-Atlas, a large-scale benchmark for evaluating tool-use competency, comprising 36 real MCP servers and 220 tools. It includes 1,000 tasks designed to assess tool-use competency in realistic, multi-step workflows. Tasks use natural language prompts that avoid naming specific tools or servers, requiring agents to identify and orchestrate 3-6 tool calls across multiple servers. We score tasks using a claims-based rubric that awards partial credit based on the factual claims satisfied in the model's final answer, complemented by internal diagnostics on tool discovery, parameterization, syntax, error recovery, and efficiency. Evaluation results on frontier models reveal that top models achieve pass rates exceeding 50%, with primary failures arising from inadequate tool usage and task understanding. We release the task schema, containerized harness, and a 500-task public subset of the benchmark dataset to facilitate reproducible comparisons and advance the development of robust, tool-augmented agents.

Observational studies can yield clinically actionable evidence at scale, but executing them on real-world databases is open-ended and requires coherent decisions across cohort construction, analysis, and reporting. Prior evaluations of LLM agents emphasize isolated steps or single answers, missing the integrity and internal structure of the resulting evidence bundle. To address this gap, we introduce RWE-bench, a benchmark grounded in MIMIC-IV and derived from peer-reviewed observational studies. Each task provides the corresponding study protocol as the reference standard, requiring agents to execute experiments in a real database and iteratively generate tree-structured evidence bundles. We evaluate six LLMs (three open-source, three closed-source) under three agent scaffolds using both question-level correctness and end-to-end task metrics. Across 162 tasks, task success is low: the best agent reaches 39.9%, and the best open-source model reaches 30.4%. Agent scaffolds also matter substantially, causing over 30% variation in performance metrics. Furthermore, we implement an automated cohort evaluation method to rapidly localize errors and identify agent failure modes. Overall, the results highlight persistent limitations in agents' ability to produce end-to-end evidence bundles, and efficient validation remains an important direction for future work. Code and data are available at https://github.com/somewordstoolate/RWE-bench.

As AI becomes more capable, we entrust it with more general and consequential tasks. The risks from failure grow more severe with increasing task scope. It is therefore important to understand how extremely capable AI models will fail: Will they fail by systematically pursuing goals we do not intend? Or will they fail by being a hot mess, and taking nonsensical actions that do not further any goal? We operationalize this question using a bias-variance decomposition of the errors made by AI models: An AI's incoherence on a task is measured over test-time randomness as the fraction of its error that stems from variance rather than bias in task outcome. Across all tasks and frontier models we measure, the longer models spend reasoning and taking actions, the more incoherent their failures become. Incoherence changes with model scale in a way that is experiment dependent. However, in several settings, larger, more capable models are more incoherent than smaller models. Consequently, scale alone seems unlikely to eliminate incoherence. Instead, as more capable AIs pursue harder tasks, requiring more sequential action and thought, our results predict failures to be accompanied by more incoherent behavior. This suggests a future where AIs sometimes cause industrial accidents (due to unpredictable misbehavior), but are less likely to exhibit consistent pursuit of a misaligned goal. This increases the relative importance of alignment research targeting reward hacking or goal misspecification.

How can "weak teacher models" such as average human annotators or existing AI systems, effectively supervise LLMs to improve performance on hard reasoning tasks, especially those that challenge and requires expertise or daily practice from the teacher models? In this paper, we seek for empirical answers to this question by investigating various data-driven strategies that offer supervision data at different quality levels upon tasks of varying complexity. Two intuitive strategies emerge for teacher models to provide supervision during alignment training: 1) using lower-quality supervision from complete tasks that match the difficulty of the target reasoning tasks, and 2) leveraging higher-quality supervision from easier subtasks that are less challenging. Interestingly, we find that even when the outcome error rate for hard task supervision is high (e.g., 90\%), training on such data can outperform perfectly correct supervision on easier subtasks on multiple hard math benchmarks. We further identify a more critical factor influencing training performance: step-wise error rates, which indicate the severity of errors in solutions. Specifically, training on hard task supervision with the same outcome error rates but disparate step-wise error rates can lead to a 30\% accuracy gap on MATH benchmark. Our results also reveal that supplementing hard task supervision with the corresponding subtask supervision can yield notable performance improvements than simply combining rephrased hard full task supervision, suggesting new avenues for data augmentation. Data and code are released at https://github.com/hexuan21/Weak-to-Strong.

The robot learning community has made great strides in recent years, proposing new architectures and showcasing impressive new capabilities; however, the dominant metric used in the literature, especially for physical experiments, is "success rate", i.e. the percentage of runs that were successful. Furthermore, it is common for papers to report this number with little to no information regarding the number of runs, the initial conditions, and the success criteria, little to no narrative description of the behaviors and failures observed, and little to no statistical analysis of the findings. In this paper we argue that to move the field forward, researchers should provide a nuanced evaluation of their methods, especially when evaluating and comparing learned policies on physical robots. To do so, we propose best practices for future evaluations: explicitly reporting the experimental conditions, evaluating several metrics designed to complement success rate, conducting statistical analysis, and adding a qualitative description of failures modes. We illustrate these through an evaluation on physical robots of several learned policies for manipulation tasks.

Can small language models achieve strong tool-use performance without complex adaptation mechanisms? This paper investigates this question through Meta-Tool, a controlled empirical study comparing hypernetwork-based LoRA adaptation against carefully designed few-shot prompting. Using a Llama-3.2-3B-Instruct backbone, we evaluate four adaptation mechanisms--few-shot prompting, documentation encoding, hypernetwork-generated LoRA weights, and value-guided beam search--across four diverse benchmarks: Gorilla APIBench, Spider 2.0, WebArena, and InterCode. Our central finding is a well-supported negative result: despite generating non-trivial weight matrices, the 227.8M-parameter hypernetwork provides no measurable improvement over few-shot prompting alone. Comprehensive ablation studies reveal that few-shot examples contribute +21.5% to performance and documentation contributes +5.0%, while the hypernetwork adds 0%. A 3B model with well-designed prompts achieves 79.7% of GPT-5's average performance at 10 times lower latency. Error analysis across 722 failure cases spanning all shot counts (0--5) shows that at the 5-shot configuration (106 failures), failure modes are task-dependent: schema-heavy tasks (Spider 2.0, WebArena) show near-zero format errors with remaining failures semantic, while format errors dominate on Gorilla (100%) and InterCode (70%). These findings redirect practitioners toward prompt engineering and example curation rather than complex adaptation architectures.

Computer-use agents have rapidly improved on real-world tasks such as web navigation, desktop automation, and software interaction, in some cases surpassing human performance. Yet even when the task and model are unchanged, an agent that succeeds once may fail on a repeated execution of the same task. This raises a fundamental question: if an agent can succeed at a task once, what prevents it from doing so reliably? In this work, we study the sources of unreliability in computer-use agents through three factors: stochasticity during execution, ambiguity in task specification, and variability in agent behavior. We analyze these factors on OSWorld using repeated executions of the same task together with paired statistical tests that capture task-level changes across settings. Our analysis shows that reliability depends on both how tasks are specified and how agent behavior varies across executions. These findings suggest the need to evaluate agents under repeated execution, to allow agents to resolve task ambiguity through interaction, and to favor strategies that remain stable across runs.

This study explores Large Language Models (LLMs) as autonomous agents for real-world tasks, including freelance software development. This work presents a new benchmark that evaluates LLMs on freelance programming and data analysis tasks derived from economic data. We construct the benchmark using synthetic tasks created from a Kaggle Freelancer dataset of job postings, with all job prices standardized to USD (median fixed-project price around 250, and an average of 306). Each task is accompanied by structured input-output test cases and an estimated price tag, enabling automated correctness checking and a monetary performance valuation. This approach is inspired by OpenAI's recent SWE-Lancer benchmark (1,400 real Upwork tasks worth 1M total). Still, our framework simplifies evaluation using programmatically testable tasks and predicted price values, making it highly scalable and repeatable. On this benchmark, we evaluate four modern LLMs - Claude 3.5 Haiku, GPT-4o-mini, Qwen 2.5, and Mistral. We report each model's accuracy (task success rate and test-case pass rate) and the total "freelance earnings" it achieves (sum of prices of solved tasks). Our results show that Claude 3.5 Haiku performs best, earning approximately 1.52 million USD, followed closely by GPT-4o-mini at 1.49 million, then Qwen 2.5 (1.33M) and Mistral ($0.70M). We analyze the distribution of errors per task and observe that the strongest models solve the most tasks and rarely fail completely on any project. We discuss the implications of these results for the feasibility of AI as a freelance developer, the advantages and limitations of our automated benchmark approach, and the gap between performance on structured tasks versus the true complexity of real-world freelance jobs.

Large language models often fail at simple counting tasks, even when the items to count are explicitly present in the prompt. We investigate whether this failure occurs because transformers do not represent counts internally, or because they cannot convert those representations into the correct output tokens. Across three model families, Pythia, Qwen3, and Mistral, ranging from 0.4B to 14B parameters, we find strong evidence for the second explanation. Linear probes recover the correct count from intermediate layers with near-perfect accuracy (R^2>0.99), showing that the information is present. However, the internal directions that encode counts are nearly orthogonal to the output-head rows for digit tokens (|cos|leq0.032). In other words, the model stores the count in a form that the digit logits do not naturally read out. We localize this failure with two interventions. Updating only the digit rows of the output head (36,864 parameters) substantially improves constrained next-token digit prediction (60.7 to 100.0% across four tasks), but it does not fix autoregressive generation. By contrast, a small LoRA intervention on attention Q/V weights (7.67M parameters) improves upstream routing and achieves 83.1% +/- 7.2% in true greedy autoregressive generation. Logit-lens measurements confirm the mechanism: the correct digit's vocabulary rank drops from 55,980 to 1, a 50,000x improvement. Additional norm, logit-lens, and cross-task analyses show that the bottleneck generalizes across character counting, addition, and list length, while remaining absent from broader multi-step reasoning benchmarks, including MMLU, GSM8K, and DROP. These results identify counting failure as a geometric readout bottleneck rather than a failure of internal representation: the model knows the count but the output pathway is geometrically misaligned with the tokens needed to express it.

While vision-language-action models (VLAs) have shown promising robotic behaviors across a diverse set of manipulation tasks, they achieve limited success rates when deployed on novel tasks out-of-the-box. To allow these policies to safely interact with their environments, we need a failure detector that gives a timely alert such that the robot can stop, backtrack, or ask for help. However, existing failure detectors are trained and tested only on one or a few specific tasks, while VLAs require the detector to generalize and detect failures also in unseen tasks and novel environments. In this paper, we introduce the multitask failure detection problem and propose SAFE, a failure detector for generalist robot policies such as VLAs. We analyze the VLA feature space and find that VLAs have sufficient high-level knowledge about task success and failure, which is generic across different tasks. Based on this insight, we design SAFE to learn from VLA internal features and predict a single scalar indicating the likelihood of task failure. SAFE is trained on both successful and failed rollouts, and is evaluated on unseen tasks. SAFE is compatible with different policy architectures. We test it on OpenVLA, pi_0, and pi_0-FAST in both simulated and real-world environments extensively. We compare SAFE with diverse baselines and show that SAFE achieves state-of-the-art failure detection performance and the best trade-off between accuracy and detection time using conformal prediction. More qualitative results can be found at https://vla-safe.github.io/.

We develop task scaling laws and model ladders to predict the individual task performance of pretrained language models (LMs) in the overtrained setting. Standard power laws for language modeling loss cannot accurately model task performance. Therefore, we leverage a two-step prediction approach: first use model and data size to predict a task-specific loss, and then use this task loss to predict task performance. We train a set of small-scale "ladder" models, collect data points to fit the parameterized functions of the two prediction steps, and make predictions for two target models: a 7B model trained to 4T tokens and a 13B model trained to 5T tokens. Training the ladder models only costs 1% of the compute used for the target models. On four multiple-choice tasks written in ranked classification format, we can predict the accuracy of both target models within 2 points of absolute error. We have higher prediction error on four other tasks (average absolute error 6.9) and find that these are often tasks with higher variance in task metrics. We also find that using less compute to train fewer ladder models tends to deteriorate predictions. Finally, we empirically show that our design choices and the two-step approach lead to superior performance in establishing scaling laws.

Autonomous agent systems powered by Large Language Models (LLMs) have demonstrated promising capabilities in automating complex tasks. However, current evaluations largely rely on success rates without systematically analyzing the interactions, communication mechanisms, and failure causes within these systems. To bridge this gap, we present a benchmark of 34 representative programmable tasks designed to rigorously assess autonomous agents. Using this benchmark, we evaluate three popular open-source agent frameworks combined with two LLM backbones, observing a task completion rate of approximately 50%. Through in-depth failure analysis, we develop a three-tier taxonomy of failure causes aligned with task phases, highlighting planning errors, task execution issues, and incorrect response generation. Based on these insights, we propose actionable improvements to enhance agent planning and self-diagnosis capabilities. Our failure taxonomy, together with mitigation advice, provides an empirical foundation for developing more robust and effective autonomous agent systems in the future.

Word error rate (WER) is the dominant metric for automatic speech recognition, yet it cannot detect a systematic failure mode: models that produce fluent output in the wrong writing system. We define Script Fidelity Rate (SFR), the fraction of hypothesis characters in the target script block, computable without reference transcriptions, and report the first systematic measurement of script collapse across six languages spanning four writing systems (Pashto, Urdu, Hindi, Bengali, Malayalam, Somali) and nine ASR models on FLEURS test sets. Across 53 evaluated model-language pairs, 18 (34%; 95% Wilson CI: 23-47%) exhibit script collapse (SFR < 10%); MMS-1B and SeamlessM4T-v2 maintain SFR above 99% on every language evaluated, confirming that SFR correctly identifies high fidelity where it is present. We identify three distinct collapse patterns: Latin phonetic substitution (smaller Whisper on Indic languages), Arabic substitution for Somali's Latin-script orthography, and Devanagari substitution where larger Whisper models treat all Indic audio as Hindi, a failure present even in Whisper large-v3.

The adoption of Large Language Models (LLMs) for code generation in data science offers substantial potential for enhancing tasks such as data manipulation, statistical analysis, and visualization. However, the effectiveness of these models in the data science domain remains underexplored. This paper presents a controlled experiment that empirically assesses the performance of four leading LLM-based AI assistants-Microsoft Copilot (GPT-4 Turbo), ChatGPT (o1-preview), Claude (3.5 Sonnet), and Perplexity Labs (Llama-3.1-70b-instruct)-on a diverse set of data science coding challenges sourced from the Stratacratch platform. Using the Goal-Question-Metric (GQM) approach, we evaluated each model's effectiveness across task types (Analytical, Algorithm, Visualization) and varying difficulty levels. Our findings reveal that all models exceeded a 50% baseline success rate, confirming their capability beyond random chance. Notably, only ChatGPT and Claude achieved success rates significantly above a 60% baseline, though none of the models reached a 70% threshold, indicating limitations in higher standards. ChatGPT demonstrated consistent performance across varying difficulty levels, while Claude's success rate fluctuated with task complexity. Hypothesis testing indicates that task type does not significantly impact success rate overall. For analytical tasks, efficiency analysis shows no significant differences in execution times, though ChatGPT tended to be slower and less predictable despite high success rates. This study provides a structured, empirical evaluation of LLMs in data science, delivering insights that support informed model selection tailored to specific task demands. Our findings establish a framework for future AI assessments, emphasizing the value of rigorous evaluation beyond basic accuracy measures.

We introduce NatureBench, a cross-discipline benchmark of 90 tasks distilled from peer-reviewed Nature-family publications, designed to evaluate whether AI coding agents can move beyond reproduction toward discovery on real scientific problems. NatureBench is built on NatureGym, an automated pipeline that constructs a standardized, per-task containerized environment from a source paper, addressing the environment-fragmentation problem that has limited the credibility of prior agent-on-research benchmarks. Evaluating ten frontier agent configurations under a strict web-search-disabled protocol, we find that the strongest model surpasses SOTA on only 17.8% of tasks under the g>0.1 criterion. Analysis of method pathways reveals that agents succeed primarily through methodological translation, converting scientific tasks into familiar supervised prediction problems, rather than through genuine scientific invention. Failures are dominated by wrong method choice and insufficient compute budget, not by task misunderstanding. We release the benchmark, the NatureGym pipeline, and a public leaderboard with maintainer-side reproduction. Code: https://github.com/FrontisAI/NatureBench

One of the main challenges developers face in the use of continuous integration (CI) and deployment pipelines is the occurrence of intermittent job failures, which result from unexpected non-deterministic issues (e.g., flaky tests or infrastructure problems) rather than regular code-related errors such as bugs. Prior studies developed machine learning (ML) models trained on large datasets of job logs to classify job failures as either intermittent or regular. As an alternative to costly manual labeling of large datasets, the state-of-the-art (SOTA) approach leveraged a heuristic based on non-deterministic job reruns. However, this method mislabels intermittent job failures as regular in contexts where rerunning suspicious job failures is not an explicit policy, and therefore limits the SOTA's performance in practice. In fact, our manual analysis of 2,125 job failures from 5 industrial and 1 open-source projects reveals that, on average, 32% of intermittent job failures are mislabeled as regular. To address these limitations, this paper introduces a novel approach to intermittent job failure detection using few-shot learning (FSL). Specifically, we fine-tune a small language model using a few number of manually labeled log examples to generate rich embeddings, which are then used to train an ML classifier. Our FSL-based approach achieves 70-88% F1-score with only 12 shots in all projects, outperforming the SOTA, which proved ineffective (34-52% F1-score) in 4 projects. Overall, this study underlines the importance of data quality over quantity and provides a more efficient and practical framework for the detection of intermittent job failures in organizations.

The continuous delivery of modern software requires the execution of many automated pipeline jobs. These jobs ensure the frequent release of new software versions while detecting code problems at an early stage. For TELUS, our industrial partner in the telecommunications field, reliable job execution is crucial to minimize wasted time and streamline Continuous Deployment (CD). In this context, flaky job failures are one of the main issues hindering CD. Prior studies proposed techniques based on machine learning to automate the detection of flaky jobs. While valuable, these solutions are insufficient to address the waste associated with the diagnosis of flaky failures, which remain largely unexplored due to the wide range of underlying causes. This study examines 4,511 flaky job failures at TELUS to identify the different categories of flaky failures that we prioritize based on Recency, Frequency, and Monetary (RFM) measures. We identified 46 flaky failure categories that we analyzed using clustering and RFM measures to determine 14 priority categories for future automated diagnosis and repair research. Our findings also provide valuable insights into the evolution and impact of these categories. The identification and prioritization of flaky failure categories using RFM analysis introduce a novel approach that can be used in other contexts.

Transferring learned patterns from pretrained neural language models has been shown to significantly improve effectiveness across a variety of language-based tasks, meanwhile further tuning on intermediate tasks has been demonstrated to provide additional performance benefits, provided the intermediate task is sufficiently related to the target task. However, how to identify related tasks is an open problem, and brute-force searching effective task combinations is prohibitively expensive. Hence, the question arises, are we able to improve the effectiveness and efficiency of tasks with no training examples through selective fine-tuning? In this paper, we explore statistical measures that approximate the divergence between domain representations as a means to estimate whether tuning using one task pair will exhibit performance benefits over tuning another. This estimation can then be used to reduce the number of task pairs that need to be tested by eliminating pairs that are unlikely to provide benefits. Through experimentation over 58 tasks and over 6,600 task pair combinations, we demonstrate that statistical measures can distinguish effective task pairs, and the resulting estimates can reduce end-to-end runtime by up to 40%.

We introduce the first version of KWBench (Knowledge Work Bench), a benchmark for unprompted problem recognition in large language models: can an LLM identify a professional scenario before attempting to solve it. Existing frontier benchmarks have saturated, and most knowledge-work evaluations to date reduce to extraction or task completion against a specification. KWBench targets the step before that: recognizing the governing structure of the situation from raw inputs alone. The benchmark contains 223 tasks sourced from practitioners across acquisitions, contract negotiations, clinical pharmacy, organizational politics, fraud analysis, and incentive design. Each task encodes a formal game-theoretic pattern (principal-agent conflict, signaling, mechanism design failure, strategic omission, coalitional dynamics, strategic interdependence) and carries structured ground truth recording the expert reading of the situation and the anticipated failure modes. Models receive raw data and a task prompt with no indication of problem type. Scoring is a three-tier rubric gated by a mandatory conjunctive check. Mandatory criteria encode the predicted wrong paths. We evaluate 16 models. The best model passes on 27.9% of tasks. The top two models agree on only 31.7% of their passes. Among the top 8, 44 tasks are solved by exactly one model; routing across the top 8 covers 50.7% of the benchmark, nearly double the best single model. Conditional on passing, quality scores converge (approx 83% across models); unconditional scores do not. Same models articulate the relevant game-theoretic concept correctly when asked, then fail to apply it unprompted. We release KWBench to shift how frontier models are evaluated on knowledge work, scoring them on whether they recognize the right problem from the situation alone, not only on how well they execute once the problem has been framed for them.

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.

As the focus in LLM-based coding shifts from static single-step code generation to multi-step agentic interaction with tools and environments, understanding which tasks will challenge agents and why becomes increasingly difficult. This is compounded by current practice: agent performance is typically measured by aggregate pass rates on benchmarks, but single-number metrics obscure the diversity of tasks within a benchmark. We present a framework for predicting success or failure on individual tasks tailored to the agentic coding regime. Our approach augments Item Response Theory (IRT) with rich features extracted from tasks, including issue statements, repository contexts, solutions, and test cases, and introduces a novel decomposition of agent ability into LLM and scaffold ability components. This parameterization enables us to aggregate evaluation data across heterogeneous leaderboards and accurately predict task-level performance for unseen benchmarks, as well as unseen LLM-scaffold combinations. Our methods have practical utility for benchmark designers, who can better calibrate the difficulty of their new tasks without running computationally expensive agent evaluations.

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.

Large language models are increasingly deployed as agents that solve tasks by interacting with external tool environments. Although recent tool-use benchmarks increasingly cover complex task settings, they still largely assume clean, stable, and trustworthy tool environments, leaving tool-environment unreliability insufficiently examined. We introduce ToolBench-X, a benchmark for evaluating agents under recoverable reliability hazards. ToolBench-X contains executable multi-step tasks across diverse domains and sequential, parallel, and mixed workflows, each paired with deterministic tools and a canonical final answer for automatic evaluation. Starting from clean tool environments, ToolBench-X injects five structured hazard types: Specification Drift, Invocation Error, Execution Failure, Output Drift, and Cross-source Conflict. Crucially, each injected instance remains solvable through at least one valid recovery path, such as retrying, fallback, verification, or cross-checking. Experiments reveal a substantial reliability gap: agents that perform well with reliable tools often fail under recoverable hazards. Further analysis shows that failures are driven less by tool-use volume or inference budget than by limited hazard diagnosis and ineffective recovery. Targeted recovery hints recover many failed tasks, while test-time scaling yields more limited gains. These results suggest that tool-use evaluation should move beyond function-call accuracy toward task completion under unreliable tool environments. The code and data is available at https://github.com/Foreverskyou/ToolBench-X.

Uncertainty in perception, actuation, and the environment often require multiple attempts for a robotic task to be successful. We study a class of problems providing (1) low-entropy indicators of terminal success / failure, and (2) unreliable (high-entropy) data to predict the final outcome of an ongoing task. Examples include a robot trying to connect with a charging station, parallel parking, or assembling a tightly-fitting part. The ability to restart after predicting failure early, versus simply running to failure, can significantly decrease the makespan, that is, the total time to completion, with the drawback of potentially short-cutting an otherwise successful operation. Assuming task running times to be Poisson distributed, and using a Markov Jump process to capture the dynamics of the underlying Markov Decision Process, we derive a closed form solution that predicts makespan based on the confusion matrix of the failure predictor. This allows the robot to learn failure prediction in a production environment, and only adopt a preemptive policy when it actually saves time. We demonstrate this approach using a robotic peg-in-hole assembly problem using a real robotic system. Failures are predicted by a dilated convolutional network based on force-torque data, showing an average makespan reduction from 101s to 81s (N=120, p<0.05). We posit that the proposed algorithm generalizes to any robotic behavior with an unambiguous terminal reward, with wide ranging applications on how robots can learn and improve their behaviors in the wild.

Software engineers develop, fine-tune, and deploy deep learning (DL) models. They use and re-use models in a variety of development frameworks and deploy them on a range of runtime environments. In this diverse ecosystem, engineers use DL model converters to move models from frameworks to runtime environments. However, errors in converters can compromise model quality and disrupt deployment. The failure frequency and failure modes of DL model converters are unknown. In this paper, we conduct the first failure analysis on DL model converters. Specifically, we characterize failures in model converters associated with ONNX (Open Neural Network eXchange). We analyze past failures in the ONNX converters in two major DL frameworks, PyTorch and TensorFlow. The symptoms, causes, and locations of failures (for N=200 issues), and trends over time are also reported. We also evaluate present-day failures by converting 8,797 models, both real-world and synthetically generated instances. The consistent result from both parts of the study is that DL model converters commonly fail by producing models that exhibit incorrect behavior: 33% of past failures and 8% of converted models fell into this category. Our results motivate future research on making DL software simpler to maintain, extend, and validate.

We construct evaluation tasks where extending the reasoning length of Large Reasoning Models (LRMs) deteriorates performance, exhibiting an inverse scaling relationship between test-time compute and accuracy. Our evaluation tasks span four categories: simple counting tasks with distractors, regression tasks with spurious features, deduction tasks with constraint tracking, and advanced AI risks. We identify five distinct failure modes when models reason for longer: 1) Claude models become increasingly distracted by irrelevant information; 2) OpenAI o-series models resist distractors but overfit to problem framings; 3) models shift from reasonable priors to spurious correlations; 4) all models show difficulties in maintaining focus on complex deductive tasks; and 5) extended reasoning may amplify concerning behaviors, with Claude Sonnet 4 showing increased expressions of self-preservation. These findings suggest that while test-time compute scaling remains promising for improving model capabilities, it may inadvertently reinforce problematic reasoning patterns. Our results demonstrate the importance of evaluating models across diverse reasoning lengths to identify and address these failure modes in LRMs.

We report a case study of four end-to-end attempts to autonomously generate ML research papers using a pipeline of six LLM agents mapped to stages of the scientific workflow. Of these four, three attempts failed during implementation or evaluation. One completed the pipeline and was accepted to Agents4Science 2025, an experimental inaugural venue that required AI systems as first authors, passing both human and multi-AI review. From these attempts, we document six recurring failure modes: bias toward training data defaults, implementation drift under execution pressure, memory and context degradation across long-horizon tasks, overexcitement that declares success despite obvious failures, insufficient domain intelligence, and weak scientific taste in experimental design. We conclude by discussing four design principles for more robust AI-scientist systems, implications for autonomous scientific discovery, and we release all prompts, artifacts, and outputs at https://github.com/Lossfunk/ai-scientist-artefacts-v1

Large language model (LLM) agents are increasingly deployed to automate productivity tasks (e.g., email, scheduling, document management), but evaluating them on live services is risky due to potentially irreversible changes. Existing benchmarks rely on simplified environments and fail to capture realistic, stateful, multi-service workflows. We introduce ClawsBench, a benchmark for evaluating and improving LLM agents in realistic productivity settings. It includes five high-fidelity mock services (Gmail, Slack, Google Calendar, Google Docs, Google Drive) with full state management and deterministic snapshot/restore, along with 44 structured tasks covering single-service, cross-service, and safety-critical scenarios. We decompose agent scaffolding into two independent levers (domain skills that inject API knowledge via progressive disclosure, and a meta prompt that coordinates behavior across services) and vary both to measure their separate and combined effects. Experiments across 6 models, 4 agent harnesses, and 33 conditions show that with full scaffolding, agents achieve task success rates of 39-64% but exhibit unsafe action rates of 7-33%. On OpenClaw, the top five models fall within a 10 percentage-point band on task success (53-63%), with unsafe action rates from 7% to 23% and no consistent ordering between the two metrics. We identify eight recurring patterns of unsafe behavior, including multi-step sandbox escalation and silent contract modification.

We introduce AutoMonitor-Bench, the first benchmark designed to systematically evaluate the reliability of LLM-based misbehavior monitors across diverse tasks and failure modes. AutoMonitor-Bench consists of 3,010 carefully annotated test samples spanning question answering, code generation, and reasoning, with paired misbehavior and benign instances. We evaluate monitors using two complementary metrics: Miss Rate (MR) and False Alarm Rate (FAR), capturing failures to detect misbehavior and oversensitivity to benign behavior, respectively. Evaluating 12 proprietary and 10 open-source LLMs, we observe substantial variability in monitoring performance and a consistent trade-off between MR and FAR, revealing an inherent safety-utility tension. To further explore the limits of monitor reliability, we construct a large-scale training corpus of 153,581 samples and fine-tune Qwen3-4B-Instruction to investigate whether training on known, relatively easy-to-construct misbehavior datasets improves monitoring performance on unseen and more implicit misbehaviors. Our results highlight the challenges of reliable, scalable misbehavior monitoring and motivate future work on task-aware designing and training strategies for LLM-based monitors.

Agent Skills package procedural instructions and checks for use by general-purpose agents, but loading a skill does not guarantee that every requirement remains active throughout a long tool-using trajectory. We study this problem in a production-derived, white-box code-audit workflow. Holding the task and 24 artifact checks fixed, we vary the surrounding context and classify where failures first become visible: lost requirements, editing drift, failed checking, or non-agent evaluator/runtime failures. Codex with gpt-5.4-mini passes 8/10 runs in a 10,991-character clean context but only 3/10 in both a 299,140-character relevant context and an equal-length irrelevant context. This 50-percentage-point difference is large but remains trend-level under two-sided Fisher tests (p = 0.0698). Requirement coverage nevertheless stays above 92% in both long conditions, showing that a few omissions can invalidate an otherwise complete artifact. A second task passes all clean and long runs, so the evidence does not support a universal context-length threshold. A detailed external checklist passes 10/10 runs, compared with 5/10 for a generic self-check (p = 0.0325). Coding-agent scaffolds may help by selecting a smaller working set, but they do not eliminate failures. We do not introduce context rot or a new general monitoring method; we provide a bounded failure classification and empirical case study for white-box code auditing.

AI agents are increasingly expected to complete long-horizon workflows that require sustained progress over hours, millions of tokens, and complex environments. Yet current agent benchmarks largely evaluate short-form tasks, such as single pull requests, small tickets, or 5-10 minute exercises, limiting our ability to measure agents' capabilities in planning, long-context understanding, and memory use. We introduce SWE-Marathon, a benchmark of 20 long-horizon tasks spanning software engineering and adjacent technical domains. Each task consists of a unique executable environment, a human-written reference solution, and a multi-layer verification suite. Logged agent attempts average 27.2M total tokens, making SWE-Marathon substantially longer-horizon than existing SWE and command-line agent benchmarks. Current frontier coding agents solve fewer than 30% of tasks. Failures often arise from poor self-verification, self-reported infeasibility, and premature termination. We also observe reward-hacking behavior in 13.8% of rollouts, where agents attempt to exploit the environment or verifier to bypass the intended workflow. SWE-Marathon includes adversarial review of test suites and execution environments, as well as multi-layer checks designed to prevent shortcut solutions. We release SWE-Marathon, evaluation code, and agent trajectories at https://swe-marathon.org/.

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.

Imitation learning (IL) with generative models, such as diffusion and flow matching, has enabled robots to perform complex, long-horizon tasks. However, distribution shifts from unseen environments or compounding action errors can still cause unpredictable and unsafe behavior, leading to task failure. Early failure prediction during runtime is therefore essential for deploying robots in human-centered and safety-critical environments. We propose FIPER, a general framework for Failure Prediction at Runtime for generative IL policies that does not require failure data. FIPER identifies two key indicators of impending failure: (i) out-of-distribution (OOD) observations detected via random network distillation in the policy's embedding space, and (ii) high uncertainty in generated actions measured by a novel action-chunk entropy score. Both failure prediction scores are calibrated using a small set of successful rollouts via conformal prediction. A failure alarm is triggered when both indicators, aggregated over short time windows, exceed their thresholds. We evaluate FIPER across five simulation and real-world environments involving diverse failure modes. Our results demonstrate that FIPER better distinguishes actual failures from benign OOD situations and predicts failures more accurately and earlier than existing methods. We thus consider this work an important step towards more interpretable and safer generative robot policies. Code, data and videos are available at https://tum-lsy.github.io/fiper_website.

Large language models (LLMs) are increasingly used for tasks that require complex reasoning. Most benchmarks focus on final outcomes but overlook the intermediate reasoning steps - such as planning, revision, and decision making under resource constraints. We argue that measuring these internal processes is essential for understanding model behavior and improving reliability. We propose using strategic games as a natural evaluation environment: closed, rule-based systems with clear states, limited resources, and automatic feedback. We introduce a framework that evaluates LLMs along three core dimensions: planning, revision, and resource-constrained decision making. To operationalize this, we define metrics beyond win rate, including overcorrection risk rate, correction success rate, improvement slope, and over-budget ratio. In 4320 adversarial rounds across 12 leading models, ChatGPT-o3-mini achieves the top composite score, with a win rate of 74.7 percent, a correction success rate of 78.6 percent, and an improvement slope of 0.041. By contrast, Qwen-Plus, despite an overcorrection risk rate of 81.6 percent, wins only 25.6 percent of its matches - primarily due to excessive resource use. We also observe a negative correlation between overcorrection risk rate and correction success rate (Pearson r = -0.51, p = 0.093), suggesting that more frequent edits do not always improve outcomes. Our findings highlight the value of assessing not only what LLMs decide but how they arrive at those decisions

When a phone-use agent avoids harm, does that show safety, or simply inability to act? Existing evaluations often cannot tell. A harmful outcome may be avoided because the agent recognized the risk and chose the safe action, or because it failed to understand the screen or execute any relevant action at all. These cases have different causes and call for different fixes, yet current benchmarks often merge them under task success, refusal, or final harmful outcome. We address this problem with PhoneSafety, a benchmark of 700 safety-critical moments drawn from real phone interactions across more than 130 apps. Each instance isolates the next decision at a risky moment and asks a simple question: does the model take the safe action, take the unsafe action, or fail to do anything useful? We evaluate eight representative phone-use agents under this framework. Our results reveal two main patterns. First, stronger general phone-use ability does not reliably imply safer choices at risky moments. Models that perform better on ordinary app tasks are not always the ones that behave more safely when the next action matters. Second, failures to do anything useful behave like a capability signal rather than a safety signal: they are concentrated in more visually and operationally demanding settings and remain stable when the evaluation protocol changes. Across models, failures split into two recurring patterns: unsafe choices in settings where the model can act but chooses wrongly, and inability to act in more visually and operationally demanding screens. Overall, a harmless outcome is not enough to count as evidence of safety. Evaluating phone-use agents requires separating unsafe judgment from inability to act.

Current benchmarks for computer-use agents evaluate models in impersonal environments. This leaves a gap between evaluation and deployment where personal assistants are expected to work across a user's whole digital life, including their context, historical data, and logged-in accounts. This gap is widest on web tasks, where live web evaluations cannot exercise sites that require logging in or personal information, the kind of site a real personal assistant has to drive. We introduce MyPCBench, which tests computer-use agents as personal assistants on a Linux desktop populated with 17 simulated real-world web applications and a full desktop stack, all seeded for one canonical persona, Michael Scott from The Office. We define 184 tasks in this environment, each inspired by a real request drawn from the OpenClaw community, and benchmark six closed and open-weight models with a uniform computer+bash tool surface. We find that the best model, Claude Opus 4.6, fully solves 55.4\% of the tasks, the only model above 50\%. Model failures cluster on tasks that span many applications and on long trajectories, where personalization stresses an assistant the most. We release the environment, task set, and agent harness at https://mypcbench.com.

We describe WebSuite, the first diagnostic benchmark for generalist web agents, designed to systematically evaluate why agents fail. Advances in AI have led to the rise of numerous web agents that autonomously operate a browser to complete tasks. However, most existing benchmarks focus on strictly measuring whether an agent can or cannot complete a task, without giving insight on why. In this paper, we 1) develop a taxonomy of web actions to facilitate identifying common failure patterns, and 2) create an extensible benchmark suite to assess agents' performance on our taxonomized actions. This benchmark suite consists of both individual tasks, such as clicking a button, and end-to-end tasks, such as adding an item to a cart, and is designed such that any failure of a task can be attributed directly to a failure of a specific web action. We evaluate two popular generalist web agents, one text-based and one multimodal, and identify unique weaknesses for each agent. Because WebSuite can disaggregate task failures into specific action failures, this enables granular identification of which UX flows an individual agent has trouble with and immediately highlights promising avenues for improvement. These findings highlight the need for more focused benchmarking on where web agents go wrong to effectively improve agents beyond their weaker performance today.

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.

Failure attribution for LLM-based agentic systems, i.e., identifying which steps in a failure trajectory caused the task to fail, is critical for debugging and improving these systems. Existing approaches either rely on prompting-based pipelines, which are computationally expensive, or require post-training on failure trajectories with step-level error annotations, which are costly to collect and difficult to scale. We argue that a practical failure attribution model should be lightweight and trainable without step-level supervision on failure data. To this end, we address unsupervised failure attribution, i.e., training exclusively on successful trajectories and identifying error steps at inference time given a failure trajectory. We propose OAT, which casts this problem as one-class learning with neural controlled differential equations, modeling the dynamical pattern of successful trajectories in latent space. At inference time, each step in a failure trajectory is assigned an anomaly score based on its deviation from the dynamics learned on successful trajectories, which is then used to form a set of error steps. With training on only 100 successful trajectories, experiments show that OAT is 200--5000 times faster than prompting-based baselines, and, at the same time, consistently outperforms them in both in-domain and out-of-distribution datasets with +20% and +7% F1 scores, respectively, demonstrating that OAT is a promising and efficient direction for diagnosing agentic system failures.

Long-term memory systems store what a user says in an external store and retrieve it when a related query arrives. This interface rests on an assumption so natural that it is rarely stated: a memory that is needed will resemble the query that needs it. World knowledge breaks the assumption. A tree-nut allergy should change the answer to a macaron request through their almond-flour ingredient, yet the two texts share no cue a retriever can see. We call this failure mode the implicit-association blind spot and introduce InMind, a 125-task, expert-verified benchmark spanning ten life domains, with 113 tasks grounded in citable public sources. Its paired controls separate three explanations that existing evaluations conflate: the fact was never stored, the model lacks the bridging knowledge, or the fact was stored and never surfaced. The verdict is clean. With the decisive memory placed in context, the backbone answers 84.0 percent of indirect queries; when the same memory must be retrieved, six vector, graph, and agentic memory systems reach at most 14.4 percent, even though they recall the same facts on demand at up to 100 percent. An embedding with eight times the dimensionality raises answer-blind target recall for every system yet leaves the gap essentially intact. A minimal diagnostic probe that keeps memory visible before the query arrives recovers most of the gap, locating the failure in the query-conditioned interface itself and pointing to routing, deciding which facts must stay visible, as the open problem InMind is built to score.

Hypernetwork-based methods such as Doc-to-LoRA internalize a document into an LLM's weights in a single forward pass, but they fail systematically on conflicts: when the document contradicts pretraining knowledge, accuracy collapses to 46.4% on the deepest facts. We show the failure is a magnitude problem rather than a representational one. The hypernetwork already targets the right layers, but its adapter margin is approximately constant across documents while the pretrained margin grows with training frequency, so deep conflicts lose by construction. The account predicts that failure should track prior strength: sorting 194 conflicts by the base model's log-probability on the contradicted fact, baseline accuracy falls from 68% on weak-prior questions to 16% on strong-prior ones, a 52 percentage-point gap. The cure is amplitude. Selective Layer Boosting scales the adapter at its top-norm layers, and Conflict-Aware Internalization triggers boosting only when the base model is confident. Both are training-free; together they raise deep-conflict accuracy from 46.4% to 71.0% on Gemma-2B and from 53.6% to 72.5% on Mistral-7B while preserving novel-knowledge recall, and beat vanilla retrieval-augmented generation on medium conflicts by 18 percentage points despite operating entirely in parameter space. We release KID-Bench, a 489-question benchmark that separates novel recall, cross-knowledge combination, and prior-graded conflicts.

Forecasts of explosive AI progress hinge on AI agents automating AI research. But evidence on whether agents can carry out open-ended AI research is thin. Current evaluations either test agents on narrow, verifiable tasks, which excludes open-ended research, or submit AI-generated papers to blind peer review, which is overstretched, stochastic, and suffers from poor review quality. We introduce a third way to measure progress towards AI R\&D automation. An agent takes on the central, open-ended research question of a high-quality unpublished paper, and the paper's original authors grade its output. We call these shadow evaluations. We ran shadow evaluations on two unpublished NeurIPS 2026 submissions, giving frontier agents six days and thousands of dollars of compute. The agents completed all of the engineering without human help, yet could not make substantial progress towards answering the research questions. As a result, both papers were unambiguously rejected by the authors. We identify five recurring failure modes: poor judgment about the bar for publishable research, uncreative responses to shortcomings in the research design, ineffective backtracking from dead ends, poor resource awareness, and instruction drift. A robustness check with a second model and scaffold reproduced these failures. We release the expert reviews, survey responses, agent repositories, and logs. Our results provide early evidence that today's agents can do the engineering of AI research, but struggle with critical parts of the research lifecycle.

The rapid proliferation of Large Language Models (LLMs) and diverse specialized benchmarks necessitates a shift from fragmented, task-specific metrics to a holistic, competitive ranking system that effectively aggregates performance across multiple ability dimensions. Primarily using static scoring, current evaluation methods are fundamentally limited. They struggle to determine the proper mix ratio across diverse benchmarks, and critically, they fail to capture a model's dynamic competitive fitness or its vulnerability when confronted with sequential, high-stakes tasks. To address this, we introduce the novel Competitive Swiss-System Dynamics (CSD) framework. CSD simulates a multi-round, sequential contest where models are dynamically paired across a curated sequence of benchmarks based on their accumulated win-loss record. And Monte Carlo Simulation (N=100,000 iterations) is used to approximate the statistically robust Expected Win Score (E[S_m]), which eliminates the noise of random pairing and early-round luck. Furthermore, we implement a Failure Sensitivity Analysis by parameterizing the per-round elimination quantity (T_k), which allows us to profile models based on their risk appetite--distinguishing between robust generalists and aggressive specialists. We demonstrate that CSD provides a more nuanced and context-aware ranking than traditional aggregate scoring and static pairwise models, representing a vital step towards risk-informed, next-generation LLM evaluation.

Existing benchmarks for tool-using LLM agents primarily report single-run success rates and miss reliability properties required in production. We introduce ReliabilityBench, a benchmark for evaluating agent reliability across three dimensions: (i) consistency under repeated execution using pass^k, (ii) robustness to semantically equivalent task perturbations at intensity ε, and (iii) fault tolerance under controlled tool/API failures at intensity λ. ReliabilityBench contributes a unified reliability surface R(k,ε,λ), action metamorphic relations that define correctness via end-state equivalence rather than text similarity, and a chaos-engineering-style fault injection framework (timeouts, rate limits, partial responses, schema drift). We evaluate two models (Gemini 2.0 Flash, GPT-4o) and two agent architectures (ReAct, Reflexion) across four domains (scheduling, travel, customer support, e-commerce) over 1,280 episodes. Perturbations alone reduce success from 96.9% at ε=0 to 88.1% at ε=0.2. Rate limiting is the most damaging fault in ablations. ReAct is more robust than Reflexion under combined stress, and Gemini 2.0 Flash achieves comparable reliability to GPT-4o at much lower cost. ReliabilityBench provides a systematic framework for assessing production readiness of LLM agents.

Memory-augmented LLM agents store and retrieve information from prior interactions, yet the relative importance of how memories are written versus how they are retrieved remains unclear. We introduce a diagnostic framework that analyzes how performance differences manifest across write strategies, retrieval methods, and memory utilization behavior, and apply it to a 3x3 study crossing three write strategies (raw chunks, Mem0-style fact extraction, MemGPT-style summarization) with three retrieval methods (cosine, BM25, hybrid reranking). On LoCoMo, retrieval method is the dominant factor: average accuracy spans 20 points across retrieval methods (57.1% to 77.2%) but only 3-8 points across write strategies. Raw chunked storage, which requires zero LLM calls, matches or outperforms expensive lossy alternatives, suggesting that current memory pipelines may discard useful context that downstream retrieval mechanisms fail to compensate for. Failure analysis shows that performance breakdowns most often manifest at the retrieval stage rather than at utilization. We argue that, under current retrieval practices, improving retrieval quality yields larger gains than increasing write-time sophistication. Code is publicly available at https://github.com/boqiny/memory-probe.

Reward modeling (RM), which captures human preferences to align large language models (LLMs), is increasingly employed in tasks such as model finetuning, response filtering, and ranking. However, due to the inherent complexity of human preferences and the limited coverage of available datasets, reward models often fail under distributional shifts or adversarial perturbations. Existing approaches for identifying such failure modes typically rely on prior knowledge about preference distributions or failure attributes, limiting their practicality in real-world settings where such information is unavailable. In this work, we propose a tractable, preference-distribution agnostic method for discovering reward model failure modes via reward guided controlled decoding. Building on this, we introduce REFORM, a self-improving reward modeling framework that enhances robustness by using the reward model itself to guide the generation of falsely scored responses. These adversarial examples are then used to augment the training data and patch the reward model's misaligned behavior. We evaluate REFORM on two widely used preference datasets Anthropic Helpful Harmless (HH) and PKU Beavertails and demonstrate that it significantly improves robustness without sacrificing reward quality. Notably, REFORM preserves performance both in direct evaluation and in downstream policy training, and further improves alignment quality by removing spurious correlations.

Large-scale AI training is now fundamentally a distributed systems problem, and hardware failures have become routine operating conditions rather than rare exceptions. Public operational evidence from production training clusters, however, remains scarce. This technical report presents an empirical analysis of a 63-node NVIDIA B200 production cluster (504 GPUs), using 55 days of Prometheus time-series data and 73 days of operational logs covering 224 multi-node training sessions. The cluster operates within a cross-organizational environment in which five parties (SKT, Upstage, Lablup, NVIDIA Korea, and VAST Data) share a unified monitoring pipeline. This arrangement enabled joint diagnosis of a 60-node-scale storage I/O bottleneck that did not appear at 2-4-node scale, a production-scale phenomenon no single team could isolate alone. Drawing on a months-long pre-training campaign, we perform three quantitative analyses yielding four findings. First, statistical analysis over 751 Prometheus metrics and 10 XID-identified GPU failures achieves a 10/10 detection rate (2/10 pre-XID) at ~0.84 false positives per day. No single metric is consistently dominant across failure types, motivating a multi-signal detection strategy. Second, profiling 523 checkpoint events along the GPU VRAM to NFS path attributes the "bandwidth paradox" (1.4-10.4% utilization of 200 Gbps RoCE) to saturation of the 128-slot NFS RPC layer. Third, multi-node failure response shows concentrated exclusions (top 3 of 63 nodes account for >50% of all exclusions) and an auto-retry chain success rate of 33.3% over 12 chains (73 attempts), 2.7x the 12.5% manual recovery rate; the median retry interval is 11 min (IQR 10-11). All analyses are grounded in production infrastructure providing session-level workload management, GPU-centric scheduling, and unified observability.

Frontier deep research agents (DRAs) plan a research task, synthesize across documents, and return a structured deliverable on demand. They are being deployed in enterprise workflows faster than they are being evaluated. Existing benchmarks measure factual recall, single-hop QA, or generic agentic skill, missing the multi-document, decision-grade work DRAs are deployed to produce. We introduce a benchmark targeting the structured analytical deliverables that fill a management consultant's typical week. We grade three frontier agents, namely Claude Opus 4.6 with web search, OpenAI o3-deep-research, and Google Gemini 3.1 Pro deep-research, on 42 SME-authored prompts. Each of the 126 responses is scored on two layers: deterministic ground-truth verifiers (mean 13.8 per task) and a five-criterion 0-3 SME rubric, composed into a Verifier-Rubric Score (VRS) on 0-100. Most prompts embed cognitive traps that penalize surface-pattern matching. Acceptance under our joint threshold (rubric mean >= 2.5 and verifier rate >= 80%) is uniformly low: Gemini 21.4%, o3 9.5%, Claude 9.5%. Mean VRS scores agree with published rubric-based benchmarks (our top 62.6 vs. APEX-v1 64.2, ProfBench 65.9, ResearchRubrics < 68%), validating the rubric construct. ACCEPT rates sit below APEX-Agents' MC-segment Pass@1 band (12.3-22.7%) on dedicated DR agents; our floor is three points lower despite the harness advantage, opened by stricter conjunctive grading and trap design. Each agent fails distinctively. Claude produces the deliverable most reliably (4.5x the others' rate on file-required tasks) but carries the highest fabrication signature. o3 has the cleanest reasoning average yet drops required sections and propagates arithmetic errors. Gemini is bimodal, with the highest acceptance rate alongside the most zero-scored rubric cells.

Rubric-based evaluation is widely used in LLM benchmarks and training pipelines for open-ended, less verifiable tasks. While prior work has demonstrated the effectiveness of rubrics using downstream signals such as reinforcement learning outcomes, there remains no principled way to diagnose how a rubric itself fails from such aggregated or downstream signals alone. To address this gap, we introduce RIFT: RubrIc Failure mode Taxonomy, a taxonomy for systematically characterizing failure modes in rubric composition and design. RIFT consists of eight failure modes organized into three high-level categories: Reliability Failures, Content Validity Failures, and Consequential Validity Failures. RIFT is developed using grounded theory by iteratively annotating rubrics drawn from five diverse data sources spanning general instruction following, code generation, creative writing, and expert-level deep research, until no new failure modes are identified. We evaluate the consistency of the taxonomy by measuring agreement among independent human annotators, observing fair agreement overall (87% pairwise agreement and 0.64 average Cohen's kappa). Finally, to support scalable diagnosis, we propose automated rubric quality metrics and show that they align with human failure-mode annotations, achieving up to 0.925 F1.

We introduce ARBIGRAPH, a benchmark generator for evaluating whether tool-assisted language agents can retain, update, compose, and discard task-relevant context across extended reasoning workflows. ARBIGRAPH represents each task as a natural-language problem with an executable Python solver, and composes tasks through typed intermediate states, instantiated here as scalar and list values. This design enables controllable task graphs whose length, dependency structure, distractor count, and value type can be varied while preserving exact automatic verification. We instantiate ARBIGRAPH with math, GSM-style word-problems, and Python-tracing task categories, and evaluate a Qwen3.5-27B tool-assisted agent across four topologies. The results show high accuracy on isolated tasks but substantial degradation on more complex dependent tasks: accuracy drops by up to 33.3% on branching chains of dependent math tasks. This shows that ARBIGRAPH exposes failures that are not visible from single-task evaluation alone. Our code, generated datasets, and evaluation results are available at https://github.com/pavelgolikov/ArbiGraph.git

Reinforcement learning holds great promise for improving robot policies beyond the limits of imitation learning. However, its practical adoption remains bottlenecked by the lack of reliable vision-language reward models that provide dense and informative feedback. Two key challenges remain: acquiring diverse failure data at scale and obtaining fine-grained reward signals beyond sparse trajectory-level success labels. Collecting failure trajectories typically requires laborious human effort, while pseudo-failures constructed by relabeling successful demonstrations fail to capture the diverse physical failure modes that arise during robot execution. Meanwhile, existing reward models often predict sparse binary or trajectory-level rewards, which provide limited guidance for efficient policy optimization. We introduce DenseReward, a dense robotic reward model that addresses both challenges. To train DenseReward, we develop an automated failure data generation pipeline that synthesizes physically realistic failure trajectories in simulation without human labeling, covering diverse failure modes such as collisions, missed grasps, object drops, and recovery behaviors. DenseReward predicts dense frame-level reward scores from visual observations and language instructions, enabling fine-grained estimation of task progress throughout an episode. Experiments show that DenseReward outperforms general-purpose VLMs and existing robotic reward models in dense reward prediction across both simulated and real-world manipulation. We further demonstrate that DenseReward provides effective reward guidance for downstream model predictive control and reinforcement learning. We release the dataset, trained reward models, and evaluation suite to support the development of failure-aware dense reward modeling for robot learning.

We present an empirical evaluation of various outputs generated by nine of the most widely-available large language models (LLMs). Our analysis is done with off-the-shelf, readily-available tools. We find a correlation between percentage of memorized text, percentage of unique text, and overall output quality, when measured with respect to output pathologies such as counterfactual and logically-flawed statements, and general failures like not staying on topic. Overall, 80.0% of the outputs evaluated contained memorized data, but outputs containing the most memorized content were also more likely to be considered of high quality. We discuss and evaluate mitigation strategies, showing that, in the models evaluated, the rate of memorized text being output is reduced. We conclude with a discussion on potential implications around what it means to learn, to memorize, and to evaluate quality text.

The advancement of large language model (LLM) based agents has shifted AI evaluation from single-turn response assessment to multi-step task completion in interactive environments. We present an empirical study evaluating frontier AI models on 150 workplace tasks within a realistic e-commerce RL environment from Surge. Our analysis reveals an empirically-derived hierarchy of agentic capabilities that models must master for real-world deployment: (1) tool use, (2) planning and goal formation, (3) adaptability, (4) groundedness, and (5) common-sense reasoning. Even the best-performing models fail approximately 40\% of the tasks, with failures clustering predictably along this hierarchy. Weaker models struggle with fundamental tool use and planning, whereas stronger models primarily fail on tasks requiring contextual inference beyond explicit instructions. We introduce a task-centric design methodology for RL environments that emphasizes diversity and domain expert contributions, provide detailed failure analysis, and discuss implications for agent development. Our findings suggest that while current frontier models can demonstrate coherent multi-step behavior, substantial capability gaps remain before achieving human-level task completion in realistic workplace settings.

AI chatbots are rapidly shaping how people encounter the news, yet no prior study has systematically measured how accurately these systems, with their proprietary search integrations and retrieval-synthesis pipelines, handle emerging facts across languages and regions. We present a 14-day (February 9-22, 2026) evaluation of six AI chatbots (Gemini 3 Flash and Pro, Grok 4, Claude 4.5 Sonnet, GPT-5 and GPT-4o mini) on 2,100 factual questions derived from same-day BBC News reporting across six regional services (US & Canada, Arabic, Afrique, Hindi, Russian, Turkish). The best systems achieve over 90% multiple-choice accuracy on questions about events reported hours earlier. The same systems, however, lose 11-13% under free-response evaluation, and 16-17% across the cohort. We further characterize three failure patterns. First, every model achieves its lowest accuracy on Hindi (79% vs. 89-91% elsewhere) and citations indicate an Anglophone retrieval bias (e.g., models answering Hindi queries cite English Wikipedia more than any Hindi outlet). Second, retrieval, not reasoning, failures drive over 70% of all errors. When models retrieve a correct source, they often extract the correct answer; the problem is to land on the right source in the first place. Third, models achieving 88-96% accuracy on well-formed questions drop to 19-70% when questions contain subtle false premises, with the most vulnerable model accepting fabricated facts 64% of the time. We also identify a detection-accuracy paradox: the best false-premise detector ranks second in adversarial accuracy (abstention rate), while a weaker detector ranks first, showing that premise detection and answer recovery are partially independent capabilities. Overall, these suggest that high accuracy can mask systematic regional inequity, near-total dependence on retrieval infrastructure, and vulnerability to imperfect queries real users pose.

Recent advances in large language models (LLMs) have led to the development of various evaluation benchmarks. These benchmarks typically rely on a single instruction template for evaluating all LLMs on a specific task. In this paper, we comprehensively analyze the brittleness of results obtained via single-prompt evaluations across 6.5M instances, involving 20 different LLMs and 39 tasks from 3 benchmarks. To improve robustness of the analysis, we propose to evaluate LLMs with a set of diverse prompts instead. We discuss tailored evaluation metrics for specific use cases (e.g., LLM developers vs. developers interested in a specific downstream task), ensuring a more reliable and meaningful assessment of LLM capabilities. We then implement these criteria and conduct evaluations of multiple models, providing insights into the true strengths and limitations of current LLMs.

Safety benchmark scores provide incomplete evidence of deployment readiness: aligned language models often adhere to rigid rules even when a situational update flips which action is safe. We term this failure brittle safety. To diagnose it, we introduce context-flip evaluation, testing 12 models across a safety benchmark (PacifAIst) and two commonsense controls using paired variants where the nominally safe action produces harm. Three findings emerge. First, brittle safety is safety-specific: all 12 models exhibit a safety-commonsense gap (mean +17.4 pp). Baseline accuracy fails to predict brittleness: among models above 90% baseline accuracy, brittleness rates range from 13.7% to 90.0%. Second, failures stem from policy override rather than miscomprehension: despite acknowledging the context change in every case, models persist via three distinct mechanisms that vary by update type and model family. Third, on a hand-audited probe of catastrophic consequence-flip scenarios, standard action-level guardrails catch none, while a state-aware validator catches all without false alarms on correct interventions. This indicates that action-level content moderation is systematically blind to consequence-flips, motivating state-aware architectural alternatives. We release our protocol, perturbed benchmarks, and deployment probe.

Deep Research Agents (DRAs) aim to automatically produce analyst-level reports through iterative information retrieval and synthesis. However, most existing DRAs were validated on question-answering benchmarks, while research on generating comprehensive reports remains overlooked. Worse, current benchmarks for report synthesis suffer from task complexity and subjective metrics -- this fails to reflect user demands and limits the practical utility of generated reports. To address these gaps, we present Fine-grained DEepResearch bench (FINDER), an enhanced benchmark consisting of 100 human-curated research tasks with 419 structured checklist items that standardize report structure, analytical depth, and factual grounding. Based on approximately 1,000 reports produced by mainstream DRAs, we further propose Deep rEsearch Failure Taxonomy (DEFT), the first failure taxonomy for deep research agents. DEFT contains 14 fine-grained failure modes across reasoning, retrieval, and generation, and is built upon grounded theory with human-LLM co-annotating and inter-annotator reliability validation. Our experimental findings reveal that current DRAs struggle not with task comprehension but with evidence integration, verification, and reasoning-resilient planning.

AI agents are increasingly deployed to execute important tasks. While rising accuracy scores on standard benchmarks suggest rapid progress, many agents still continue to fail in practice. This discrepancy highlights a fundamental limitation of current evaluations: compressing agent behavior into a single success metric obscures critical operational flaws. Notably, it ignores whether agents behave consistently across runs, withstand perturbations, fail predictably, or have bounded error severity. Grounded in safety-critical engineering, we provide a holistic performance profile by proposing twelve concrete metrics that decompose agent reliability along four key dimensions: consistency, robustness, predictability, and safety. Evaluating 14 agentic models across two complementary benchmarks, we find that recent capability gains have only yielded small improvements in reliability. By exposing these persistent limitations, our metrics complement traditional evaluations while offering tools for reasoning about how agents perform, degrade, and fail.

In long-horizon tasks, decision-relevant state is often scattered across an expanding trajectory, while the action agent must surface it and act. As trajectories grow, task requirements, environment facts, prior attempts, diagnoses, and open subgoals can be buried in the context window or pushed beyond it, failing to influence decisions when needed. We call this failure mode "behavioral state decay". We study memory as an active intervention mechanism rather than passive retrieval. A separate memory agent runs alongside an unmodified action agent, updating a structured memory bank from the recent trajectory and deciding whether to inject a memory-grounded reminder or remain silent. The module is plug-and-play with frontier action agents and existing agent harnesses. Across Terminal-Bench 2.0 and τ^2-Bench, it improves pass@1 for both weaker and stronger action agents, with gains of +8.3 pp on Terminal-Bench and +6.8 pp on τ^2-Bench. Ablations show that selective intervention outperforms passive bank exposure, always-on injection, advisor-only guidance, and general retrieval. As an early step toward open-weight memory policies, we train Qwen3.5-27B on SETA using SFT and GRPO, improving validation reward and achieving partial transfer to Terminal-Bench.

Large language models exhibit safety degradation in non-English languages. Standard evaluation relies on Jailbreak Success Rate (JSR), which confounds several safety-driving factors into one, obscuring the specific cause(s) of safety failure. We introduce a latent variable model, a Multi-Group Item Response Theory (IRT) framework, that decouples safety-driving factors such as language-agnostic safety robustness (θ), intrinsic prompt hardness (β), global language processing difficulty (γ), and a prompt-specific cross-lingual safety gap (τ). Using the MultiJail dataset, we evaluate the safety robustness of 61 model configurations across 5 closed-model families and 10 languages of varying resource, aggregating a dataset of 1.9 million rows. Exploratory Factor Analysis shows safety is primarily unidimensional: models refuse different harm types mainly through a shared mechanism. Contrary to the expected trend that safety degrades largely in low-resource languages, 22 model configurations are more vulnerable in English than in low-resource languages. Low-resource languages produce more uncertain responses (high entropy) than high-resource languages. Also, high-τ prompts cluster in physical harm categories like Theft and Weapons and lower-resource languages, trends validated through cross-dataset generalization. While global translation quality shows low correlation with τ, severe mistranslations drive high-bias outliers, as validated by native speakers. Cultural and conceptual grounding mismatches also contribute to τ. In predictive validation, the IRT framework achieves AUC = 0.940, outperforming simpler baselines in predicting safe refusal of unsafe prompts. Our framework reveals concept-language vulnerabilities that aggregate metrics obscure, enabling fairer cross-lingual safety evaluation and targeted improvements in dataset construction.

LLM agents are expected to complete end-to-end units of work across software tools, business services, and local workspaces. Yet many agent benchmarks freeze a curated task set at release time and grade mainly the final response, making it difficult to evaluate agents against evolving workflow demand or verify whether a task was executed. We introduce Claw-Eval-Live, a live benchmark for workflow agents that separates a refreshable signal layer, updated across releases from public workflow-demand signals, from a reproducible, time-stamped release snapshot. Each release is constructed from public workflow-demand signals, with ClawHub Top-500 skills used in the current release, and materialized as controlled tasks with fixed fixtures, services, workspaces, and graders. For grading, Claw-Eval-Live records execution traces, audit logs, service state, and post-run workspace artifacts, using deterministic checks when evidence is sufficient and structured LLM judging only for semantic dimensions. The release contains 105 tasks spanning controlled business services and local workspace repair, and evaluates 13 frontier models under a shared public pass rule. Experiments reveal that reliable workflow automation remains far from solved: the leading model passes only 66.7% of tasks and no model reaches 70%. Failures are structured by task family and execution surface, with HR, management, and multi-system business workflows as persistent bottlenecks and local workspace repair comparatively easier but unsaturated. Leaderboard rank alone is insufficient because models with similar pass rates can diverge in overall completion, and task-level discrimination concentrates in a middle band of tasks. Claw-Eval-Live suggests that workflow-agent evaluation should be grounded twice, in fresh external demand and in verifiable agent action.

Analyzing memes on the internet has emerged as a crucial endeavor due to the impact this multi-modal form of content wields in shaping online discourse. Memes have become a powerful tool for expressing emotions and sentiments, possibly even spreading hate and misinformation, through humor and sarcasm. In this paper, we present the overview of the Memotion 3 shared task, as part of the DeFactify 2 workshop at AAAI-23. The task released an annotated dataset of Hindi-English code-mixed memes based on their Sentiment (Task A), Emotion (Task B), and Emotion intensity (Task C). Each of these is defined as an individual task and the participants are ranked separately for each task. Over 50 teams registered for the shared task and 5 made final submissions to the test set of the Memotion 3 dataset. CLIP, BERT modifications, ViT etc. were the most popular models among the participants along with approaches such as Student-Teacher model, Fusion, and Ensembling. The best final F1 score for Task A is 34.41, Task B is 79.77 and Task C is 59.82.

Interactive agent benchmarks face a tension between scalable construction and realistic workflow evaluation. Hand-authored tasks are expensive to extend and revise, while static prompt evaluation misses failures that only appear when agents operate over persistent state. Existing interactive benchmarks have advanced agent evaluation significantly, but most initialize tasks from clean state and do not systematically test how agents handle pre-existing partial, stale, or conflicting artifacts. We present ClawForge, a generator-backed benchmark framework for executable command-line workflows under state conflict. The framework compiles scenario templates, grounded slots, initialized state, reference trajectories, and validators into reproducible task specifications, and evaluates agents step by step over persistent workflow surfaces using normalized end state and observable side effects rather than exact trajectory matching. We instantiate this framework as the ClawForge-Bench (17 scenarios, 6 ability categories). Results across seven frontier models show that the best model reaches only 45.3% strict accuracy, wrong-state replacement remains below 17\% for all models, and the widest model separation (17% to 90%) is driven by whether agents inspect existing state before acting. Partial-credit and step-efficiency analyses further reveal that many failures are near-miss closures rather than early breakdowns, and that models exhibit qualitatively different failure styles under state conflict.

Proactive interventions by LLM critic models are often assumed to improve reliability, yet their effects at deployment time are poorly understood. We show that a binary LLM critic with strong offline accuracy (AUROC 0.94) can nevertheless cause severe performance degradation, inducing a 26 percentage point (pp) collapse on one model while affecting another by near zero pp. This variability demonstrates that LLM critic accuracy alone is insufficient to determine whether intervention is safe. We identify a disruption-recovery tradeoff: interventions may recover failing trajectories but also disrupt trajectories that would have succeeded. Based on this insight, we propose a pre-deployment test that uses a small pilot of 50 tasks to estimate whether intervention is likely to help or harm, without requiring full deployment. Across benchmarks, the test correctly anticipates outcomes: intervention degrades performance on high-success tasks (0 to -26 pp), while yielding a modest improvement on the high-failure ALFWorld benchmark (+2.8 pp, p=0.014). The primary value of our framework is therefore identifying when not to intervene, preventing severe regressions before deployment.

E-commerce risk management requires aggregating diverse, deeply embedded web data through multi-step, stateful interactions, which traditional scraping methods and most existing Graphical User Interface (GUI) agents cannot handle. These agents are typically limited to single-step tasks and lack the ability to manage dynamic, interactive content critical for effective risk assessment. To address this challenge, we introduce RISK, a novel framework designed to build and deploy GUI agents for this domain. RISK integrates three components: (1) RISK-Data, a dataset of 8,492 single-step and 2,386 multi-step interaction trajectories, collected through a high-fidelity browser framework and a meticulous data curation process; (2) RISK-Bench, a benchmark with 802 single-step and 320 multi-step trajectories across three difficulty levels for standardized evaluation; and (3) RISK-R1, a R1-style reinforcement fine-tuning framework considering four aspects: (i) Output Format Constraint, (ii) Single-step and (iii) Multi-step Level Reward, and (iv) Task Level Reweight. Experiments show that RISK-R1 achieves a 6.8% improvement in offline single-step and an 8.8% improvement in offline multi-step, using only 7.2% of the parameters of the SOTA baseline. Moreover, it attains a top task success rate of 70.5% in online evaluation. RISK provides a scalable, domain-specific solution for automating complex web interactions in e-commerce risk management. The code is available at https://github.com/RenqiChen/RISK-GUI.

AI systems fail silently far more often than they fail visibly. In an analysis of 100K human-AI interactions from the WildChat dataset, we find that 79% of AI failures are invisible: something went wrong but the user gave no overt indication that there was a problem. These invisible failures cluster into eight archetypes that help us characterize where and how AI systems are failing to meet users' needs. In addition, the archetypes show systematic co-occurrence patterns indicating higher-level failure types. To address the question of whether these archetypes will remain relevant as AI systems become more capable, we also created and annotated a counterfactual dataset in which WildChat's 2024-era responses are replaced by those from three present-day frontier LMs. This analysis indicates that failure rates have dropped substantially, but that the vast majority of failures remain invisible in our sense, and the distribution of failure archetypes seems stable. Finally, we illustrate how the archetypes help us to identify systematic and variable AI limitations across different usage domains. Overall, we argue that our invisible failure taxonomy can be a key component in reliable failure monitoring for product developers, scientists, and policy makers. Our code and data are available at https://github.com/bigspinai/bigspin-invisible-failure-archetypes

Collecting high-quality training examples for language model fine-tuning is expensive, with practical budgets limiting the amount of data that can be procured. We investigate whether example difficulty affects GRPO training effectiveness by comparing selection strategies (easy, medium, hard, random) across multiple models and reasoning tasks. Training on the hardest 10\% of examples (those where the base model fails most often) yields dramatic performance gains up to 47\%, while easy examples produce minimal improvements of 3-15\%. This occurs because GRPO requires outcome variance to generate learning signals; hard examples maintain mixed success/failure outcomes throughout training while easy examples quickly converge to consistent success, eliminating learning opportunities. Moreover, models trained on hard examples show superior out-of-distribution generalization, with only hard-trained models achieving meaningful gains on the AIME2025 benchmark. Our findings provide clear guidance: when budget-constrained, prioritize collecting and annotating examples where your base model struggles, as these drive nearly all learning value in GRPO fine-tuning

We propose a new test to measure a text model's multitask accuracy. The test covers 57 tasks including elementary mathematics, US history, computer science, law, and more. To attain high accuracy on this test, models must possess extensive world knowledge and problem solving ability. We find that while most recent models have near random-chance accuracy, the very largest GPT-3 model improves over random chance by almost 20 percentage points on average. However, on every one of the 57 tasks, the best models still need substantial improvements before they can reach expert-level accuracy. Models also have lopsided performance and frequently do not know when they are wrong. Worse, they still have near-random accuracy on some socially important subjects such as morality and law. By comprehensively evaluating the breadth and depth of a model's academic and professional understanding, our test can be used to analyze models across many tasks and to identify important shortcomings.

Recent human-computer interaction (HCI) research has revealed a widespread misalignment between how developers design workplace artificial intelligence (AI) systems, and what workers actually need from them. Yet, little research has examined the effects of this gap, or how it may cause harm. We analyzed 1,524 reports of incidents in which AI systems were used to perform 171 occupational tasks across 12 industry sectors. Using an Large Language Model (LLM)-as-an-expert approach, we extracted the main traits of the AI systems involved in those incidents using an established framework of twelve traits. We then compared them with the traits that 202 workers highly familiar with those tasks would have preferred. We found that as many as 83\% of workplace incidents stem from worker-AI misalignments. In most cases, workers wanted systems that are precise, insightful, or personal, but instead received systems that are basic, simple, or general. Over the years, fast AI caused a considerable number of incidents, yet these declined, and imaginative AI, with the mass introduction of generative AI, started to cause incidents. We also compared the traits causing the incidents with the traits that 197 developers building AI systems for those tasks would have preferred. If the traits causing the incidents were the same as those designed by developers, then developers may be responsible for those incidents. We found that 74\% of task misalignments could be attributed to developers who tended to overfocus on efficiency and speed, especially for systems performing tasks in people-facing occupations such as those in the human resources sector. Our results call for design interventions that better align AI development with workers' needs, as without such corrections, workplace AI incidents are likely to persist, causing the invisible erosion of worker agency and organizational productivity.

Language-model agents are increasingly used as persistent coworkers that assist users across multiple working days. During such workflows, the surrounding environment may change independently of the agent: new emails arrive, calendar entries shift, knowledge-base records are updated, and evidence appears across images, scanned PDFs, audio, video, and spreadsheets. Existing benchmarks do not adequately evaluate this setting because they typically run within a single static episode and remain largely text-centric. We introduce , a benchmark for coworker agents built around multi-turn multi-day tasks, a stateful sandboxed service environment whose state evolves between turns, and rule-based verification. The current release contains 100 tasks across 13 professional scenarios, executed against five stateful sandboxed services (filesystem, email, calendar, knowledge base, spreadsheet) and scored by 1537 deterministic Python checkers over post-execution service state; no LLM-as-judge is invoked during scoring. We benchmark seven frontier agent systems. The strongest model reaches 75.8 weighted score, but the best strict Task Success is only 20.0\%, indicating that partial progress is common while complete end-to-end workflow completion remains rare. Turn-level analysis shows that performance drops after the first exogenous environment update, highlighting adaptation to changing state as a key open challenge. We release the benchmark, evaluation harness, and construction pipeline to support reproducible coworker-agent evaluation.

Recent advances in large language models (LLMs) have increased the demand for comprehensive benchmarks to evaluate their capabilities as human-like agents. Existing benchmarks, while useful, often focus on specific application scenarios, emphasizing task completion but failing to dissect the underlying skills that drive these outcomes. This lack of granularity makes it difficult to deeply discern where failures stem from. Additionally, setting up these environments requires considerable effort, and issues of unreliability and reproducibility sometimes arise, especially in interactive tasks. To address these limitations, we introduce the Massive Multitask Agent Understanding (MMAU) benchmark, featuring comprehensive offline tasks that eliminate the need for complex environment setups. It evaluates models across five domains, including teal{Tool-use}, teal{Directed Acyclic Graph (DAG) QA}, teal{Data Science and Machine Learning coding}, teal{Contest-level programming} and teal{Mathematics}, and covers five essential capabilities: orange{Understanding}, orange{Reasoning}, orange{Planning}, orange{Problem-solving}, and orange{Self-correction}. With a total of 20 meticulously designed tasks encompassing over 3K distinct prompts, MMAU provides a comprehensive framework for evaluating the strengths and limitations of LLM agents. By testing 18 representative models on MMAU, we provide deep and insightful analyses. Ultimately, MMAU not only sheds light on the capabilities and limitations of LLM agents but also enhances the interpretability of their performance. Datasets and evaluation scripts of MMAU are released at https://github.com/apple/axlearn/docs/research/mmau.

When deploying large language models (LLMs), it is important to ensure that these models are not only capable, but also reliable. Many benchmarks have been created to track LLMs' growing capabilities, however there has been no similar focus on measuring their reliability. To understand the potential ramifications of this gap, we investigate how well current benchmarks quantify model reliability. We find that pervasive label errors can compromise these evaluations, obscuring lingering model failures and hiding unreliable behavior. Motivated by this gap in the evaluation of reliability, we then propose the concept of so-called platinum benchmarks, i.e., benchmarks carefully curated to minimize label errors and ambiguity. As a first attempt at constructing such benchmarks, we revise examples from fifteen existing popular benchmarks. We evaluate a wide range of models on these platinum benchmarks and find that, indeed, frontier LLMs still exhibit failures on simple tasks such as elementary-level math word problems. Analyzing these failures further reveals previously unidentified patterns of problems on which frontier models consistently struggle. We provide code at https://github.com/MadryLab/platinum-benchmarks

Frontier coding agents solve complex tasks when given complete context but collapse when specifications are incomplete or ambiguous. The bottleneck is not raw capability, but judgment: knowing when to act autonomously and when to ask for help. Current benchmarks are blind to this failure mode. They supply unambiguous detailed instructions and solely reward execution correctness, so an agent that makes a lucky guess for a missing requirement will score identically to one that would have asked to be certain. We present HiL-Bench (Human-in-the-Loop Benchmark) to measure this selective escalation skill. Each task contains human-validated blockers (missing information, ambiguous requests, contradictory information) that surface only through progressive exploration, not upfront inspection. Our core metric, Ask-F1, the harmonic mean of question precision and blocker recall, captures the tension between over-asking and silent guessing; its structure architecturally prevents gaming through question spam. Evaluation across SWE and text-to-SQL domains reveals a large universal judgment gap: no frontier model recovers more than a fraction of its full-information performance when deciding whether to ask. Failure analysis identifies three key help-seeking patterns: overconfident wrong beliefs with no gap detection; high uncertainty detection yet persistent errors; broad, imprecise escalation without self-correction. These consistent patterns confirm poor help-seeking is a model-level flaw, not task-specific. RL training on shaped Ask-F1 reward shows judgment is trainable: a 32B model improves both help-seeking quality and task pass rate, with gains that transfer across domains. The model does not learn domain-specific heuristics for when to ask; it learns to detect unresolvable uncertainty and act on it.

Despite widespread deployment of Large Language Models, systematic evaluation of instruction-following capabilities remains challenging. While comprehensive benchmarks exist, focused assessments that quickly diagnose specific instruction adherence patterns are valuable. As newer models may be trained on existing benchmarks, novel evaluation approaches are needed to assess genuine capabilities rather than memorized performance. This paper presents a streamlined evaluation framework using twenty carefully designed prompts to assess LLM instruction-following across diverse task categories. We demonstrate this framework through a large-scale empirical study conducted on October 14, 2025, testing 256 verified working models from 331 available via OpenRouter. To ensure methodological rigor and prevent selection bias, we first verified each model's basic functionality before inclusion. Unlike large-scale benchmarks requiring extensive computational resources, our approach offers a practical diagnostic tool researchers and practitioners can readily apply. Our methodology builds upon verifiable instructions while introducing a compact test suite balancing comprehensiveness with efficiency. Each prompt targets distinct aspects of instruction following, including format compliance, content constraints, logical sequencing, and multi-step task execution. We evaluate models from major providers (OpenAI, Anthropic, Google, Meta, Mistral) and emerging implementations (Qwen, DeepSeek, community models), providing comparative performance analysis. Our findings reveal consistent failure modes and identify specific instruction types posing particular challenges. This work contributes both a practical evaluation tool and one of the most comprehensive empirical analyses of instruction-following capabilities across the contemporary LLM landscape.

When an LLM agent reads a confidential file, then writes a summary, then emails it externally, no single step is unsafe, but the sequence is a data leak. We call this safety drift: individually safe actions compounding into violations. Prior work has measured this problem; we predict it. SafetyDrift models agent safety trajectories as absorbing Markov chains, computing the probability that a trajectory will reach a violation within a given number of steps via closed form absorption analysis. A consequence of the monotonic state design is that every agent will eventually violate safety if left unsupervised (absorption probability 1.0 from all states), making the practical question not if but when, and motivating our focus on finite horizon prediction. Across 357 traces spanning 40 realistic tasks in four categories, we discover that "points of no return" are sharply task dependent: in communication tasks, agents that reach even a mild risk state have an 85% chance of violating safety within five steps, while in technical tasks the probability stays below 5% from any state. A lightweight monitor built on these models detects 94.7% of violations with 3.7 steps of advance warning at negligible computational cost, outperforming both keyword matching (44.7% detection, 55.9% false positive rate) and per step LLM judges (52.6% detection, 38.2% false positive rate) while running over 60,000x faster.

Foundation models have revolutionized general-purpose problem-solving, offering rapid task adaptation through pretraining, meta-training, and finetuning. Recent crucial advances in these paradigms reveal the importance of challenging task prioritized sampling to enhance adaptation robustness under distribution shifts. However, ranking task difficulties over iteration as a preliminary step typically requires exhaustive task evaluation, which is practically unaffordable in computation and data-annotation. This study provides a novel perspective to illuminate the possibility of leveraging the dual importance of adaptation robustness and learning efficiency, particularly in scenarios where task evaluation is risky or costly, such as iterative agent-environment interactions for robotic policy evaluation or computationally intensive inference steps for finetuning foundation models. Firstly, we introduce Model Predictive Task Sampling (MPTS), a framework that bridges the task space and adaptation risk landscape, providing a theoretical foundation for robust active task sampling. MPTS employs a generative model to characterize the episodic optimization process and predicts task-specific adaptation risk via posterior inference. The resulting risk learner amortizes the costly evaluation of task adaptation performance and provably approximates task difficulty rankings. MPTS seamlessly integrates into zero-shot, few-shot, and supervised finetuning settings. Empirically, we conduct extensive experiments in pattern recognition using foundation models and sequential decision-making. Our results demonstrate that MPTS significantly enhances adaptation robustness for tail or out-of-distribution (OOD) tasks and improves learning efficiency compared to state-of-the-art (SOTA) methods. The code is available at the project site https://github.com/thu-rllab/MPTS.

We study how far a deliberately simple behavioral-cloning policy can progress in a visually rich first-person game before adding reinforcement learning or explicit memory. Cortex is a compact Quake policy with 10.98 million trainable parameters in a six-layer transformer over a frozen DINOv3 encoder. It is trained on the Quake subset of the public Pixels2Play corpus: 6,849 recordings (about 474.7 hours), represented as 17.09 million cached decision frames with keyboard and mouse actions. One sampled training epoch uses 517,048 four-frame windows and takes 3.3 minutes of policy-head optimization on one RTX 5080, excluding one-time feature extraction. We evaluate two independent batches of 20 stochastic, 120-second episodes on Quake E1M1. Cortex does not complete the level, but every episode reaches the opening door, button room, and gate descent; 19 of 20 episodes in each batch record at least one kill. Under the same time-controlled harness, released P2P-150M and NitroGen checkpoints remain shallower in five matched-duration episodes each. These comparisons are limited by small reference samples and different native interfaces. Ablations show that denser visual tokens improve combat and survival, while longer optimization and naive action history improve offline metrics without consistently improving play. The remaining failures are consistent with covariate shift and motivate targeted corrective data. We release the policy implementation, checkpoint, and a representative rollout.

AI agents often fail in ways that are difficult to localize because executions are probabilistic, long-horizon, multi-agent, and mediated by noisy tool outputs. We address this gap by manually annotating failed agent runs and release a novel benchmark of 115 failed trajectories spanning structured API workflows, incident management, and open-ended web/file tasks. Each trajectory is annotated with a critical failure step and a category from a grounded-theory derived, cross-domain failure taxonomy. To mitigate the human cost of failure attribution, we present AGENTRX, an automated domain-agnostic diagnostic framework that pinpoints the critical failure step in a failed agent trajectory. It synthesizes constraints, evaluates them step-by-step, and produces an auditable validation log of constraint violations with associated evidence; an LLM-based judge uses this log to localize the critical step and category. Our framework improves step localization and failure attribution over existing baselines across three domains.

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.

Language Model (LM) agents for cybersecurity that are capable of autonomously identifying vulnerabilities and executing exploits have the potential to cause real-world impact. Policymakers, model providers, and other researchers in the AI and cybersecurity communities are interested in quantifying the capabilities of such agents to help mitigate cyberrisk and investigate opportunities for penetration testing. Toward that end, we introduce Cybench, a framework for specifying cybersecurity tasks and evaluating agents on those tasks. We include 40 professional-level Capture the Flag (CTF) tasks from 4 distinct CTF competitions, chosen to be recent, meaningful, and spanning a wide range of difficulties. Each task includes its own description, starter files, and is initialized in an environment where an agent can execute bash commands and observe outputs. Since many tasks are beyond the capabilities of existing LM agents, we introduce subtasks, which break down a task into intermediary steps for more gradated evaluation; we add subtasks for 17 of the 40 tasks. To evaluate agent capabilities, we construct a cybersecurity agent and evaluate 7 models: GPT-4o, Claude 3 Opus, Claude 3.5 Sonnet, Mixtral 8x22b Instruct, Gemini 1.5 Pro, Llama 3 70B Chat, and Llama 3.1 405B Instruct. Without guidance, we find that agents are able to solve only the easiest complete tasks that took human teams up to 11 minutes to solve, with Claude 3.5 Sonnet and GPT-4o having the highest success rates. Finally, subtasks provide more signal for measuring performance compared to unguided runs, with models achieving a 3.2\% higher success rate on complete tasks with subtask-guidance than without subtask-guidance. All code and data are publicly available at https://cybench.github.io

Frontier large language models are increasingly deployed as orchestration backbones for biological research workflows, yet no shared evidence base exists for comparing their refusal behaviour on legitimate research prompts. RefusalBench, introduced here, is a matched-triple benchmark of 141 prompts in 47 bundles that holds task framing constant while varying only biological risk tier (benign, borderline, dual-use), enabling tier-conditioned comparisons robust to subdomain confounding. A 15-prompt should-refuse positive-control module establishes per-model calibration floors; three models fail to refuse even these prompts. Across 19 frontier models in the May 2026 snapshot, strict refusal rates span 0.1% to 94.6% on identical prompts. Jurisdiction does not predict refusal in this snapshot (Mann-Whitney U, p = 0.393; EU n = 1, US bimodal); provider identity does, with Anthropic's API stack predicting refusal at OR = 21.03 (95% CI: 14.58-30.34 prompt-clustered; 5.70-77.55 under model-clustered GEE). This effect is best read as access-path-level rather than model-weight-level: 99.8% of Anthropic's strict refusals carry the same safety_policy adjudicated reason code, consistent with a small set of canonical refusal templates rather than case-by-case model reasoning. Strict refusal rate misranks safety calibration: Grok 4.20 achieves the highest tier discrimination (Youden's J = 0.787) while ranking only seventh by overall refusal rate, and Claude Opus 4.7's J dropped 65% from prior versions with no improvement in dual-use detection. Nine of 18 frontier models exhibit a hedge-but-help partial-compliance pattern at dual-use tier that binary refusal metrics cannot detect.

Language model agents are increasingly effective in solving realistic tasks through multi-turn tool use. However, training reliable tool-using agents remains challenging in practice. While reinforcement learning provides an on-policy paradigm for improving agents from their own environment interactions, its effectiveness depends heavily on the training task distribution. When tasks are fixed before training, the task distribution can become increasingly mismatched with the policy's evolving capabilities, causing many rollouts to be spent on uninformative tasks. We propose SENTINEL, a failure-driven reinforcement learning framework that turns the Solver's rollout failures into targeted training tasks. SENTINEL follows a Controller--Proposer--Solver loop: the Controller analyzes failed trajectories and summarizes recurring error patterns, the Proposer generates executable tasks that stress these weaknesses, and the Solver is trained on the targeted tasks. On Tau2-Bench Retail with Qwen3-4B-Thinking-2507, SENTINEL improves Pass\^{}1 from 66.4 to 74.9 and outperforms RL on general synthetic tasks across Pass\^{}k metrics. These results demonstrate that model failures provide an effective and scalable source of targeted training signal for improving tool-using language model agents.

Does continued scaling of large language models (LLMs) yield diminishing returns? Real-world value often stems from the length of task an agent can complete. We start this work by observing the simple but counterintuitive fact that marginal gains in single-step accuracy can compound into exponential improvements in the length of a task a model can successfully complete. Then, we argue that failures of LLMs when simple tasks are made longer arise from mistakes in execution, rather than an inability to reason. We propose isolating execution capability, by explicitly providing the knowledge and plan needed to solve a long-horizon task. We find that larger models can correctly execute significantly more turns even when small models have 100\% single-turn accuracy. We observe that the per-step accuracy of models degrades as the number of steps increases. This is not just due to long-context limitations -- curiously, we observe a self-conditioning effect -- models become more likely to make mistakes when the context contains their errors from prior turns. Self-conditioning does not reduce by just scaling the model size. In contrast, recent thinking models do not self-condition, and can also execute much longer tasks in a single turn. We conclude by benchmarking frontier thinking models on the length of task they can execute in a single turn. Overall, by focusing on the ability to execute, we hope to reconcile debates on how LLMs can solve complex reasoning problems yet fail at simple tasks when made longer, and highlight the massive benefits of scaling model size and sequential test-time compute for long-horizon tasks.

We investigate large language model performance across five orders of magnitude of compute scaling in eleven recent model architectures. We show that average benchmark performance, aggregating over many individual tasks and evaluations as in the commonly-used BIG-Bench dataset, is decently predictable as a function of training compute scale. Specifically, when extrapolating BIG-Bench Hard performance across one order of magnitude in compute, we observe average absolute errors of 6 percentage points (pp). By contrast, extrapolation for individual BIG-Bench tasks across an order of magnitude in compute yields higher average errors of 18pp. Nonetheless, individual task performance remains significantly more predictable than chance. Overall, our work suggests compute scaling provides a promising basis to forecast AI capabilities in diverse benchmarks, though predicting performance in specific tasks poses challenges.

Large language model (LLM) agents increasingly rely on external memory systems to remain consistent across long-horizon interactions, but little empirical work has been done to understand the specific failure modes and design choices that these systems present. Existing benchmarks report aggregate question-answering accuracy and treat memory systems as black boxes, making it impossible to attribute an incorrect answer to a particular failure mode of the system. We introduce MemFail, a diagnostic benchmark that isolates the failure modes of modern LLM memory systems. We begin by formalizing memory systems as the composition of three canonical operations -- summarization, storage, and retrieval -- and identify the potential failure modes induced by each. Based on these hypothesized failure modes, we construct five datasets spanning four tasks, each adversarially designed to test a specific operation of a memory system. Using these datasets, we evaluate four state-of-the-art memory systems on MemFail and demonstrate how MemFail can be used to empirically understand the tradeoffs induced by differences in memory system architectures.

In this paper, we introduce a novel theoretical framework for multi-task regression, applying random matrix theory to provide precise performance estimations, under high-dimensional, non-Gaussian data distributions. We formulate a multi-task optimization problem as a regularization technique to enable single-task models to leverage multi-task learning information. We derive a closed-form solution for multi-task optimization in the context of linear models. Our analysis provides valuable insights by linking the multi-task learning performance to various model statistics such as raw data covariances, signal-generating hyperplanes, noise levels, as well as the size and number of datasets. We finally propose a consistent estimation of training and testing errors, thereby offering a robust foundation for hyperparameter optimization in multi-task regression scenarios. Experimental validations on both synthetic and real-world datasets in regression and multivariate time series forecasting demonstrate improvements on univariate models, incorporating our method into the training loss and thus leveraging multivariate information.

Reflexion-style agents rely on self-generated reflections as memory, implicitly assuming that agents can accurately diagnose their own failures. We show that this assumption can fail systematically: across ALFWorld and HumanEval, agents store confident but incorrect interpretations of the task and continue acting on them across trials, even though the environment resets to the correct task each time. We call this failure mode memory confabulation and introduce the Reflection Repetition Rate (RRR), a log-based metric that detects repeated reliance on incorrect reflective content. Using RRR, we identify 16 frozen environments in ALFWorld, where 0 of 121 reflections mention the correct target object, and 4 analogous cases in HumanEval. Our mitigation replaces open-ended self-diagnosis with programmatic extraction of trajectory-level failure signals, increasing correct object mention from 0% to 86%, reducing RRR from 0.64 to 0.10, and solving 3 of 16 frozen ALFWorld environments, suggesting that reflective memory can reinforce false beliefs rather than correct them.

Recent advances in autonomous LLM agents demonstrate their ability to improve performance through iterative interaction with the environment. We define this paradigm as Test-Time Improvement (TTI). However, the mechanisms under how and why TTI succeed or fail remain poorly understood, and existing evaluation metrics fail to capture their task optimization efficiency, behavior adaptation after erroneous actions, and the specific utility of working memory for task completion. To address these gaps, we propose Test-time Improvement Diagnostic Evaluation (TIDE), an agent-agnostic and environment-agnostic framework that decomposes TTI into three comprehensive and interconnected dimensions. The framework measures (1) the overall temporal dynamics of task completion and (2) identifies whether performance is primarily constrained by recursive looping behaviors or (3) by burdensome accumulated memory. Through extensive experiments across diverse agents and environments, TIDE highlights that improving agent performance requires more than scaling internal reasoning, calling for explicitly optimizing the interaction dynamics between the agent and the environment.

Computer-use agents (CUAs) automate on-screen work, as illustrated by GPT-5.4 and Claude. Yet their reliability on complex, low-frequency interactions is still poor, limiting user trust. Our analysis of failure cases from advanced models suggests a long-tail pattern in GUI operations, where a relatively small fraction of complex and diverse interactions accounts for a disproportionate share of task failures. We hypothesize that this issue largely stems from the scarcity of data for complex interactions. To address this problem, we propose a new benchmark CUActSpot for evaluating models' capabilities on complex interactions across five modalities: GUI, text, table, canvas, and natural image, as well as a variety of actions (click, drag, draw, etc.), covering a broader range of interaction types than prior click-centric benchmarks that focus mainly on GUI widgets. We also design a renderer-based data-synthesis pipeline: scenes are automatically generated for each modality, screenshots and element coordinates are recorded, and an LLM produces matching instructions and action traces. After training on this corpus, our Phi-Ground-Any-4B outperforms open-source models with fewer than 32B parameters. We will release our benchmark, data, code, and models at https://github.com/microsoft/Phi-Ground.git

Vision-Language-Action (VLA) models enable robots to follow natural language instructions and generalize across diverse tasks, but they remain vulnerable to execution failures that compromise reliability in real-world deployment. Detecting such failures during execution is therefore critical for the robust deployment of embodied systems. Existing failure detection methods either rely on expensive action resampling or external models, while alternatives propagate trajectory-level labels uniformly across every timestep, obscuring localized failure signals. In this paper, we propose Hide-and-Seek, a framework that formulates VLA failure detection as a coarsely supervised learning problem. By combining inter-trajectory and intra-trajectory contrastive objectives, Hide-and-Seek localizes failure-indicative actions and induces temporally structured failure signals from trajectory-level supervision alone, without any step-level annotation. We evaluate Hide-and-Seek on LIBERO, VLABench, and a real-world robotic platform across three representative VLA policies: OpenVLA, π_0, and π_{0.5}.Our method achieves state-of-the-art multi-task failure detection performance with a practical accuracy--timeliness trade-off under conformal prediction, and generalizes well to both seen and unseen tasks.

Current LLM safety benchmarks are predominantly English-centric and often rely on translation, failing to capture country-specific harms. Moreover, they rarely evaluate a model's ability to detect culturally embedded sensitivities as distinct from universal harms. We introduce XL-SafetyBench. a suite of 5,500 test cases across 10 country-language pairs, comprising a Jailbreak Benchmark of country-grounded adversarial prompts and a Cultural Benchmark where local sensitivities are embedded within innocuous requests. Each item is constructed via a multi-stage pipeline that combines LLM-assisted discovery, automated validation gates, and dual independent native-speaker annotators per country. To distinguish principled refusal from comprehension failure, we evaluate Attack Success Rate (ASR) alongside two complementary metrics we introduce: Neutral-Safe Rate (NSR) and Cultural Sensitivity Rate (CSR). Evaluating 10 frontier and 27 local LLMs reveals two key findings. First, jailbreak robustness and cultural awareness do not show a coupled relationship among frontier models, so a composite safety score obscures per-axis variation. Second, local models exhibit a near-linear ASR-NSR trade-off (r = -0.81), indicating that their apparent safety reflects generation failure rather than genuine alignment. XL-SafetyBench enables more nuanced, cross-cultural safety evaluation in the multilingual era.

Large Language Models (LLMs) have achieved remarkable success in code generation, and the race to improve their performance has become a central focus of AI research. Benchmarks and leaderboards are increasingly popular, offering quantitative rankings of LLMs. However, they provide limited insight into the tasks that LLMs consistently fail to solve - information that is crucial for understanding current limitations and guiding the development of more capable models. To address this gap, we examined code generation tasks across four popular benchmarks, identifying those that major LLMs are most likely to fail. To understand the causes of these failures, we investigated whether the static complexity of solution code contributes to them, followed by a systematic inspection of 114 tasks that LLMs consistently struggled with. Our analysis revealed four recurring patterns of weaknesses in LLMs, as well as common complications within benchmark tasks that most often lead to failure.

Multi-Agent Systems (MAS) built on large language models typically solve complex tasks by coordinating multiple agents through workflows. Existing approaches generates workflows either at task level or query level, but their relative costs and benefits remain unclear. After rethinking and empirical analyses, we show that query-level workflow generation is not always necessary, since a small set of top-K best task-level workflows together already covers equivalent or even more queries. We further find that exhaustive execution-based task-level evaluation is both extremely token-costly and frequently unreliable. Inspired by the idea of self-evolution and generative reward modeling, we propose a low-cost task-level generation framework SCALE, which means \textbf{S}elf prediction of the optimizer with few shot \textbf{CAL}ibration for \textbf{E}valuation instead of full validation execution. Extensive experiments demonstrate that SCALE maintains competitive performance, with an average degradation of just 0.61\% compared to existing approach across multiple datasets, while cutting overall token usage by up to 83\%.

Computer-use agents, which leverage multimodal large language models (MLLMs) to operate computers and complete tasks, have attracted significant attention for their utility and versatility. A major challenge in developing these agents is collecting large-scale, high-quality trajectories. The standard approach generates synthetic data through a self-improving loop: an agent is placed in a verifiable environment and iteratively fine-tuned on its successful trajectories. Despite its effectiveness, this paradigm exploits only successful trajectories and discards the failed ones, even though failures carry rich information about a model's weaknesses. In this work, we explore a complementary failure-driven self-improvement loop, a data-centric paradigm that turns failed trajectories into agent improvements. Specifically, we employ an LLM to diagnose failure modes, propose inference-time solutions, and generate code patches -- lightly verified by humans -- that upgrade the agent. We validate this approach with the state-of-the-art OpenCUA-72B model on the OSWorld benchmark, improving the success rate from 42.3% to 48.9%, a gain of 6.6 percentage points, without any additional training cost and with only modest inference overhead. Our results demonstrate that failure-driven self-improvement is a viable complement to success-based pipelines, enabling more efficient agent improvement.

Language models (LMs) can produce errors that are hard to detect for humans, especially when the task is complex. RLHF, the most popular post-training method, may exacerbate this problem: to achieve higher rewards, LMs might get better at convincing humans that they are right even when they are wrong. We study this phenomenon under a standard RLHF pipeline, calling it "U-SOPHISTRY" since it is Unintended by model developers. Specifically, we ask time-constrained (e.g., 3-10 minutes) human subjects to evaluate the correctness of model outputs and calculate humans' accuracy against gold labels. On a question-answering task (QuALITY) and programming task (APPS), RLHF makes LMs better at convincing our subjects but not at completing the task correctly. RLHF also makes the model harder to evaluate: our subjects' false positive rate increases by 24.1% on QuALITY and 18.3% on APPS. Finally, we show that probing, a state-of-the-art approach for detecting Intended Sophistry (e.g. backdoored LMs), does not generalize to U-SOPHISTRY. Our results highlight an important failure mode of RLHF and call for more research in assisting humans to align them.

Agentic coding assistants are increasingly given extra capabilities, such as browser based testing tools and design oriented system prompts, on the assumption that more capability yields better software. This study tested that assumption directly. Ninety independent agent runs built the same application, a real time retrospective board, from one detailed specification, each scored on a fixed 14 criterion functional rubric (42 point maximum) and a visual quality review. The runs spanned several model generations, two agent harnesses, two reasoning effort levels, a testing tool, and two design oriented prompts. Capability tier dominated: frontier models clustered near the ceiling while a low cost local model fell to 24 to 37 points. A criterion level analysis revealed what run totals conceal. Container deployment was the dominant defect, failing first try in 44 percent of runs, with its failure rate shifting sharply across model generations while mean totals moved less than a point. The testing tool raised cost by 42 to 68 percent without improving functional score or reliability, even on interface visible criteria. Raising reasoning effort from High to xHigh lifted first try perfect runs from 28 percent to 89 percent and cut corrective prompts about five fold, for 9 to 29 percent more cost. A design oriented prompt raised visual quality, 4.5 versus 3.0 on a 5 point scale, without lifting function, and a one paragraph paraphrase of its directive reproduced the entire lift. The practical lesson is to match the fix to the failure: most first run failures came from weak reasoning, which a stronger model or more effort prevents, not from visible flaws a checking tool would catch.

Failure attribution in LLM multi-agent systems-identifying the agent and step responsible for task failures-provides crucial clues for systems debugging but remains underexplored and labor-intensive. In this paper, we propose and formulate a new research area: automated failure attribution for LLM multi-agent systems. To support this initiative, we introduce the Who&When dataset, comprising extensive failure logs from 127 LLM multi-agent systems with fine-grained annotations linking failures to specific agents and decisive error steps. Using the Who&When, we develop and evaluate three automated failure attribution methods, summarizing their corresponding pros and cons. The best method achieves 53.5% accuracy in identifying failure-responsible agents but only 14.2% in pinpointing failure steps, with some methods performing below random. Even SOTA reasoning models, such as OpenAI o1 and DeepSeek R1, fail to achieve practical usability. These results highlight the task's complexity and the need for further research in this area. Code and dataset are available at https://github.com/mingyin1/Agents_Failure_Attribution

Outcome-reward reinforcement learning (RL) has proven effective at improving the reasoning capabilities of large language models (LLMs). However, standard RL assigns credit only at the level of the final answer, penalizing entire reasoning traces when the outcome is incorrect and uniformly reinforcing all steps when it is correct. As a result, correct intermediate steps may be discouraged in failed traces, while spurious steps may be reinforced in successful ones. We refer to this failure mode as the problem of credit assignment. While a natural remedy is to train a process reward model, accurately optimizing such models to identify corrective reasoning steps remains challenging. We introduce Intervention Training (InT), a training paradigm in which the model performs fine-grained credit assignment on its own reasoning traces by proposing short, targeted corrections that steer trajectories toward higher reward. Using reference solutions commonly available in mathematical reasoning datasets and exploiting the fact that verifying a model-generated solution is easier than generating a correct one from scratch, the model identifies the first error in its reasoning and proposes a single-step intervention to redirect the trajectory toward the correct solution. We then apply supervised fine-tuning (SFT) to the on-policy rollout up to the point of error concatenated with the intervention, localizing error to the specific step that caused failure. We show that the resulting model serves as a far better initialization for RL training. After running InT and subsequent fine-tuning with RL, we improve accuracy by nearly 14% over a 4B-parameter base model on IMO-AnswerBench, outperforming larger open-source models such as gpt-oss-20b.

Coding agents are increasingly used as iterative development partners, but most benchmarks still evaluate one specification followed by one final assessment. This leaves out a basic question: can an agent keep its own codebase working as requirements change? We introduce EvoCode-Bench, a benchmark of 26 stateful coding tasks and 227 evaluated rounds. Each task preserves the agent's workspace for 5-15 rounds, states requirements through observable behavior, and uses cumulative executable tests to check new requirements and still-active prior ones. We evaluate 13 coding agents with two metrics: MT@4, a four-attempt fail-stop multi-round score, and SR, a single-round score from a reference-completed prior state. For most agents, SR exceeds MT@4 by 22-40 points. The gap also changes rankings: the highest-SR agent (78.9) ranks only third in persistent execution (44.0 MT@4). Even the strongest agents achieve only about 50% success on multi-turn metrics, and aggregate pass rate drops below half of round-1 performance by round 5. Failure analysis shows tier-dependent behavior: weaker agents fail early, while stronger agents survive long enough to expose specification-tracking and regression failures. We release the benchmark data and Harbor multi-turn infrastructure.