Daily Papers

A research idea may be novel, coherent, and scientifically plausible, yet its proposed method may remain insufficiently specified for faithful implementation. We study the codification readiness of implementation-facing research-method specifications, defined by whether they provide sufficient methodological information for a competent implementer or coding agent to construct the intended method without unsupported assumptions. We construct evidence-grounded specifications and their supported resolutions from papers, codebases, issue threads, and reproduction artifacts. We introduce IdeaAMBIG, a benchmark of 660 evidence-grounded instances: 163 real-world gaps from reproducibility reports and GitHub issues, and 497 controlled synthetic gaps injected into codification-ready references. IdeaAMBIG evaluates three capabilities: codification-readiness assessment, defect localization, and clarification action generation. Defect localization receives only the specification, whereas clarification additionally receives the annotated defect. Across 13 LLMs, the best model achieves 9.6% Macro Defect Recovery Rate on real-world instances but 80.6% Macro Clarification Action Success Rate when given the defect. In an oracle study, supplying the gold resolution raises the downstream codification-ready rate from 14% to 98%. Across all evaluated models, defect localization is the main bottleneck, with stronger clarification given the defect.

Despite rapid progress in LLM agents, performance on long-horizon, tool-using tasks remains fragile. To better understand this fragility, we ask a simple question: do all actions contribute equally to failure? Analyzing execution traces on τ-Bench (Airline/Retail) and SWE-Bench Verified, we decompose trajectories into mutating (environment-changing) vs.\ non-mutating steps and formalize decisive deviations, earliest action, level divergences that flip success to failure. A logistic regression reveals that each additional deviation in a mutating action reduces the odds of success by upto 92% on Airline and upto 96% on Retail for SoTA models. In contrast, deviations in non-mutating actions have little to no effect. Errors also grow with context length as agents drift from role and act on stale constraints. Motivated by these observations, we introduce , a model-agnostic, gradient-free, test-time safeguard that (i) adds mutation-gated verification, (ii) injects Targeted Reflection before mutating steps, and (iii) performs block-based context cleaning. delivers consistent gains, e.g., Qwen3-Thinking: +28\% relative on Airline, +11\% on Retail, and +7\% on SWE-Bench Verified; Claude: +9\%/+7\%. We further identify ceiling effects in τ-Bench, where annotation errors and underspecified tasks artificially cap model performance. To address this, we release τ-Bench Verified, which restores benchmark headroom through targeted revisions. Our results argue for action-level analysis, targeted safeguards, and reliable evaluations as prerequisites for robust multi-turn agents.

Evolving documents, such as laws, tax codes, and software documentation, are amended, replaced, and sometimes reverted over time, so a question has different correct answers at different dates. In contrast to encyclopedic knowledge, where an old fact is simply overwritten, an amendment is itself an official text that states what it replaces and when it takes effect, and the earlier version stays correct for its validity period. The central challenge is therefore version resolution, that is, identifying the version in force on the queried date. Existing temporal QA datasets treat time only as an annotation, so version resolution stays untested. We present TIDE, an expert-verified benchmark of 3,050 QA pairs over 644 official customs instruments issued between 1969 and 2025 by the Government of Bangladesh, covering eight task types over deeply code-mixed documents that are heterogeneous in layout and dated in two calendars. In addition, we evaluate nine recent LLMs under a single protocol across parametric, gold-context, and retrieval access, scored by a three-judge LLM council with a hard date gate separating correct meaning from correct time. The best macro-averaged accuracy is only 68.5%. Resolving a version from an implicit date reaches 59.7%, and detecting that the supplied version does not govern the query reaches only 26.7%. Models are more likely to find correct versions than to reject incorrect ones, and they tend to follow a confident parametric answer over the supplied authoritative text. All code and data are available at https://github.com/icsetepa44/TIDE

We present our submission to the EgoProactive track of the ECCV 2026 Wearable AI Challenge, which ranked first in the large-model division and second in the <=2B division. The task requires a wearable assistant to decide after each eight-second segment of egocentric video whether to intervene or remain silent. Our approach has two main components. First, we reformulate intervention timing as single-token classification. Rather than generating either interrupt<utterance> or silent, the model predicts yes or no, and we derive the decision from the renormalised probabilities of these two tokens. This formulation improved macro-F1 by 0.249 and G-mean by 0.30 over free-form generation. Second, because labelled data were limited to the released validation set, we generated additional supervision using a tool-calling video agent that inspects each clip and assigns intervention timestamps. A narration-only alternative was four times larger and ten times cheaper, but transferred worse than supervision from an unrelated real corpus, suggesting that visual grounding is more important than annotation volume for this task.

This paper presents a multi-label topic model for financial texts like ad-hoc announcements, 8-K filings, finance related news or annual reports. I train the model on a new financial multi-label database consisting of 3,044 German ad-hoc announcements that are labeled manually using 20 predefined, economically motivated topics. The best model achieves a macro F1 score of more than 85%. Translating the data results in an English version of the model with similar performance. As application of the model, I investigate differences in stock market reactions across topics. I find evidence for strong positive or negative market reactions for some topics, like announcements of new Large Scale Projects or Bankruptcy Filings, while I do not observe significant price effects for some other topics. Furthermore, in contrast to previous studies, the multi-label structure of the model allows to analyze the effects of co-occurring topics on stock market reactions. For many cases, the reaction to a specific topic depends heavily on the co-occurrence with other topics. For example, if allocated capital from a Seasoned Equity Offering (SEO) is used for restructuring a company in the course of a Bankruptcy Proceeding, the market reacts positively on average. However, if that capital is used for covering unexpected, additional costs from the development of new drugs, the SEO implies negative reactions on average.

Multiple-choice QA benchmarks usually evaluate small language models (SLMs) as direct answerers, but deployed language-model systems increasingly rely on external scaffolds such as tools, code, and repeated model calls. We introduce Code-Guided Reasoning (CGR), an evaluation protocol and generated-program resource for measuring when executable reasoning scaffolds improve SLM performance on MCQA tasks. CGR standardizes six components: a normalized item interface, a direct solver prompt, a generator prompt, a Python scaffold, solver-call and extraction helpers, and a three-channel result record. On 20,498 retained result rows from a locally prepared MCQA bundle and six metadata-registered solver models, the observed non-zero-baseline partition shows 66.21% macro assisted accuracy versus 38.11% direct accuracy, a +28.10 percentage-point difference with a pair-bootstrap interval of [20.32, 36.43]. Under a stricter Ab > 30% direct-signal gate, the macro difference is +14.11 points. These estimates are descriptive. Assisted inference uses a larger solver-call budget, answer extraction is brittle, Time-MQA contains the observed regressions, and some generated programs violate the no-hard-coding instruction. CGR provides the trace package needed to interpret these results, including direct, assisted, and generator-side answers, partition definitions, generated programs, response metadata, and audits.

We introduce MacroBench, a code-first benchmark that evaluates whether LLMs can synthesize reusable browser-automation programs (macros) from natural-language goals by reading HTML/DOM and emitting Selenium. MacroBench instantiates seven self-hosted sites covering 681 tasks across interaction complexity and targeting difficulty. Our end-to-end protocol validates generated code via static checks, sandboxed execution, and outcome verification (DOM assertions, database snapshots), and includes a safety suite for scraping, spam/abuse, and credential/privacy prompts. Across 2,636 model-task runs, we observe stratified success: GPT-4o-mini (96.8%), GPT-4o (95.3%), Gemini (89.0%), DeepSeek (83.4%). Models handle simple tasks reliably (91.7%) but fail on complex workflows (0.0%), and none meet production-quality coding practices despite functional completion. We release our complete benchmark pipeline, evaluation framework, and experimental results at https://github.com/hyunjun1121/MacroBench to enable reproducible assessment of macro synthesis for web automation.

Loop Engineering is emerging as a practice for organizing development work around coding agents. Instead of writing each prompt by hand, practitioners design loops that monitor progress, assign work, run checks, and decide what the agent should do next. Even with a capable coding agent, a loop may trust a stale progress note, skip needed verification, spend its budget in the wrong direction, or stop before the task is safe to submit. Yet the final outcome of one end-to-end run cannot tell whether success or failure reflects the loop's guidance or the coding agent's ability to carry out the task. We introduce LoopArena, a benchmark for evaluating how well one model can guide a separate coding agent through a long-running task. The model under evaluation is the Controller: after each coding round, it receives a structured summary of the run and instructs a separate, fixed coding agent, the Worker, on what to do or verify next, or decides whether to stop. LoopArena evaluates this ability in three complementary settings that differ in execution scope and cost. Type I scores next-step Loop Contract selection through execution-validated questions without running the Worker at evaluation time. Type II executes repeated control over a selected slice of a full task, while Type III evaluates the paired full task from its original state. On full tasks, the best observed Strict Success Rate is 24.69\%, leaving substantial room for improvement in long-horizon loop control. Across Controllers, the paired reduction in estimated inference cost averages 64.4\%, and Type II produces a similar ordering under the main Core criterion (Spearman's \(ρ=0.9747\)). We release the benchmark data and evaluation code at https://github.com/AMAP-ML/LoopArena .

While making investment decisions by reading financial documents, investors need to differentiate between in-claim and outof-claim numerals. In this paper, we present a tool which does it automatically. It extracts context embeddings of the numerals using one of the transformer based pre-trained language model called BERT. After this, it uses a Logistic Regression based model to detect whether the numerals is in-claim or out-of-claim. We use FinNum-3 (English) dataset to train our model. After conducting rigorous experiments we achieve a Macro F1 score of 0.8223 on the validation set. We have open-sourced this tool and it can be accessed from https://github.com/sohomghosh/FiNCAT_Financial_Numeral_Claim_Analysis_Tool

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.

A coding agent combines a model with a harness, which decides what the model sees, which tools it can use, and how the work continues. We ask whether changing the harness changes the result when the model and task stay fixed. We compare two configurations of the same harness on three coding benchmarks. The control supplies the full conversation in time order, while the treatment keeps the same record but mechanically shortens older tool results as the context fills and responds to repeated or stalled work. Under tight context, the treatment raises mean per-task fail-to-pass fraction (F2PF) in all three pressure comparisons and increases complete solutions on SWE-bench Verified and SWE-bench Pro. The tight-window Verified comparison uses 169 tasks, a 20,480-token window, and a fixed 480-second attempt endpoint; on this cohort, treatment raises mean per-task F2PF from 28 percent to 49 percent and complete solutions from 43 to 72. Without model-specific retuning, the same frozen treatment also raises both endpoints on the same cohort for three additional models with different designs. In the wide-window Qwen3.6 comparisons, observed arm outcomes are close on Verified and Pro, while FeatureBench retains a higher mean per-task F2PF under treatment. On the wide-window Verified cohort, treatment also serves fewer prompt tokens per turn. Because changing the harness changed what unchanged model weights could accomplish, coding-agent evaluations should treat the model and harness together as the tested solver.

Clarification-seeking behavior is widely regarded as a desirable property of LLM agents, enabling them to resolve ambiguity before acting on underspecified tasks. However, the security implications of this interaction pattern remain unexplored. We investigate whether the transition from standard execution to a clarification-seeking state increases an agent's susceptibility to prompt injection attacks. We introduce ASPI (Ambiguous-State Prompt Injection), a benchmark of 728 task-attack scenarios that isolates clarification as a distinct agent state and measures how this state transition affects vulnerability under controlled conditions. Each benchmark instance is evaluated under matched execution and clarification settings: in the execution setting, the agent acts on a fully specified instruction and encounters adversarial content only through tool-returned data; in the clarification setting, the agent must first request and incorporate additional user input before acting. We evaluate ten frontier LLMs and find that clarification-seeking consistently and substantially amplifies vulnerability. For instance, attack success rises from 1.8% to 34.0% for o3 and from 2.2% to 35.7% for Gemini-3-Flash. A decomposition analysis reveals that this gap reflects both a state-dependent shift in how models process incoming content and a channel-specific effect arising from the agent-solicited clarification interface. These findings demonstrate that standard execution-time security evaluation systematically underestimates the attack surface of interactive agents, and that robustness under fully specified tasks does not translate to robustness under ambiguity. For reproducibility, our data and source code are available at https://github.com/scaleapi/aspi.

We introduce the Agent GPA (Goal-Plan-Action) framework: an evaluation paradigm based on an agent's operational loop of setting goals, devising plans, and executing actions. The framework includes five evaluation metrics: Goal Fulfillment, Logical Consistency, Execution Efficiency, Plan Quality, and Plan Adherence. Logical Consistency checks that an agent's actions are consistent with its prior actions. Execution Efficiency checks whether the agent executes in the most efficient way to achieve its goal. Plan Quality checks whether an agent's plans are aligned with its goals; Plan Adherence checks if an agent's actions are aligned with its plan; and Goal Fulfillment checks that agent's final outcomes match the stated goals. Our experimental results on two benchmark datasets - the public TRAIL/GAIA dataset and an internal dataset for a production-grade data agent - show that this framework (a) provides a systematic way to cover a broad range of agent failures, including all agent errors on the TRAIL/GAIA benchmark dataset; (b) supports LLM-judges that exhibit strong agreement with human annotation, covering 80% to over 95% errors; and (c) localizes errors with 86% agreement to enable targeted improvement of agent performance.

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.

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.

When AI agents operating with access to sensitive information encounter a conflict between completing an assigned task and following rules or ethical constraints, they can resort to unsanctioned behaviour. Existing inference time safety work addresses this primarily through monitoring and access restriction. We investigate a complementary and under-explored layer: environmental controls that act on the agent's decision context at the point of conflict, making it more likely that the agent takes an authorised alternative path rather than an unsanctioned one. Drawing on Situational Crime Prevention (SCP), a framework used in human insider risk management to make harmful actions less rewarding and compliant actions more viable by design choices in the environment, we design and evaluate escalation channels as a concrete instantiation of this control class. An escalation channel provides an agent with a formal, out-of-band route to surface a conflict to an independent authority. We evaluate two designs: a simple email escalation and an instrumentally credible channel that guarantees a 30-minute pause and independent review, making the authorised path genuinely useful for goal achievement rather than merely nominally available. Across 10 frontier LLMs using the agentic task-rule conflict scenario of Lynch et al. (2025), we find that without any control the harmful action rate is 38.73%. A simple escalation channel reduces this to 5.92%; the instrumentally credible channel reduces it further to 1.21%, a statistically significant improvement observed in all 10 models tested across 24,000 samples. Our results suggest that the instrumental credibility of the authorised alternative matters considerably, and that environmental control design is a productive and largely unexplored addition to the defence-in-depth toolkit for agentic AI systems.

Wu et al. (2026) showed that most frontier large language models (LLMs) recommend a sponsored, roughly twice-as-expensive flight when their system prompt contains a soft sponsorship cue. We reproduce their evaluation on ten open-weight chat models plus the two of their twenty-three models that are still reachable today (gpt-3.5-turbo, gpt-4o). All reported rates in this paper are produced under the same judge the original paper used (gpt-4o); we additionally store every label under an open-weight (gpt-oss-120b) and a smaller proprietary (gpt-4o-mini) judge for an ablation. Three findings emerge. First, a prose description of an LLM evaluation pipeline is not, on its own, sufficient for accurate reproduction: we surfaced three silent implementation failures that each shifted a reported rate by tens of percentage points. Second, the central claims do generalise - the gpt-3.5-turbo logistic-regression intercept of alpha = 0.81 is within four points of the original alpha = 0.86, and 200 of 200 trials on gpt-3.5-turbo and gpt-4o promote a payday lender to a financially distressed user. Third, a thirty-token user prompt that asks the assistant for a neutral comparison table first cuts sponsored recommendation from 46.9% to 1.0% averaged across our ten open-source models, and from 53.0% to 0% averaged across the two OpenAI models. AI literacy and price-comparison portals are likely market-level mitigations; the harmful-product cell is bounded by neither. Raw data, labels and analysis scripts are at https://github.com/akmaier/Paper-LLM-Ads .

A central aspect of machine learning research is experimentation, the process of designing and running experiments, analyzing the results, and iterating towards some positive outcome (e.g., improving accuracy). Could agents driven by powerful language models perform machine learning experimentation effectively? To answer this question, we introduce MLAgentBench, a suite of 13 tasks ranging from improving model performance on CIFAR-10 to recent research problems like BabyLM. For each task, an agent can perform actions like reading/writing files, executing code, and inspecting outputs. We then construct an agent that can perform ML experimentation based on ReAct framework. We benchmark agents based on Claude v1.0, Claude v2.1, Claude v3 Opus, GPT-4, GPT-4-turbo, Gemini-Pro, and Mixtral and find that a Claude v3 Opus agent is the best in terms of success rate. It can build compelling ML models over many tasks in MLAgentBench with 37.5% average success rate. Our agents also display highly interpretable plans and actions. However, the success rates vary considerably; they span from 100% on well-established older datasets to as low as 0% on recent Kaggle challenges created potentially after the underlying LM was trained. Finally, we identify several key challenges for LM-based agents such as long-term planning and reducing hallucination. Our code is released at https://github.com/snap-stanford/MLAgentBench.

The paper describes the RuSentNE-2023 evaluation devoted to targeted sentiment analysis in Russian news texts. The task is to predict sentiment towards a named entity in a single sentence. The dataset for RuSentNE-2023 evaluation is based on the Russian news corpus RuSentNE having rich sentiment-related annotation. The corpus is annotated with named entities and sentiments towards these entities, along with related effects and emotional states. The evaluation was organized using the CodaLab competition framework. The main evaluation measure was macro-averaged measure of positive and negative classes. The best results achieved were of 66% Macro F-measure (Positive+Negative classes). We also tested ChatGPT on the test set from our evaluation and found that the zero-shot answers provided by ChatGPT reached 60% of the F-measure, which corresponds to 4th place in the evaluation. ChatGPT also provided detailed explanations of its conclusion. This can be considered as quite high for zero-shot application.

Evaluating the quality of multi-turn chatbot interactions remains challenging, as most existing methods assess interactions at the turn level without addressing whether a user's overarching goal was fulfilled. A ``goal'' here refers to an information need or task, such as asking for policy information or applying for leave. We propose a comprehensive framework for goal-oriented evaluation of multi-agent systems (MAS), introducing the Goal Success Rate (GSR) to measure the percentage of fulfilled goals, and a Root Cause of Failure (RCOF) taxonomy to identify reasons for failure in multi-agent chatbots. Our method segments conversations by user goals and evaluates success using all relevant turns. We present a model-based evaluation system combining teacher LLMs, where domain experts define goals, set quality standards serving as a guidance for the LLMs. The LLMs use ``thinking tokens'' to produce interpretable rationales, enabling explainable, data-efficient evaluations. In an enterprise setting, we apply our framework to evaluate AIDA, a zero-to-one employee conversational agent system built as a ground-up multi-agent conversational agent, and observe GSR improvement from 63\% to 79\% over six months since its inception. Our framework is generic and offers actionable insights through a detailed defect taxonomy based on analysis of failure points in multi-agent chatbots, diagnosing overall success, identifying key failure modes, and informing system improvements.

Automatic evaluation of large language model (LLM) responses requires not only factual correctness but also clarity, particularly in political question-answering. While recent datasets provide human annotations for clarity and evasion, the impact of prompt design on automatic clarity evaluation remains underexplored. In this paper, we study prompt-based clarity evaluation using the CLARITY dataset from the SemEval 2026 shared task. We compare a GPT-3.5 baseline provided with the dataset against GPT-5.2 evaluated under three prompting strategies: simple prompting, chain-of-thought prompting, and chain-of-thought with few-shot examples. Model predictions are evaluated against human annotations using accuracy and class-wise metrics for clarity and evasion, along with hierarchical exact match. Results show that GPT-5.2 consistently outperforms the GPT-3.5 baseline on clarity prediction, with accuracy improving from 56 percent to 63 percent under chain-of-thought with few-shot prompting. Chain-of-thought prompting yields the highest evasion accuracy at 34 percent, though improvements are less stable across fine-grained evasion categories. We further evaluate topic identification and find that reasoning-based prompting improves accuracy from 60 percent to 74 percent relative to human annotations. Overall, our findings indicate that prompt design reliably improves high-level clarity evaluation, while fine-grained evasion and topic detection remain challenging despite structured reasoning prompts.

Automated fact checking with large language models (LLMs) offers a scalable alternative to manual verification. Evaluating fact checking is challenging as existing benchmark datasets often include post claim analysis and annotator cues, which are absent in real world scenarios where claims are fact checked immediately after being made. This limits the realism of current evaluations. We present Politi Fact Only (PFO), a 5 class benchmark dataset of 2,982 political claims from politifact.com, where all post claim analysis and annotator cues have been removed manually. This ensures that models are evaluated using only the information that would have been available prior to the claim's verification. Evaluating LLMs on PFO, we see an average performance drop of 22% in terms of macro f1 compared to PFO's unfiltered version. Based on the identified challenges of the existing LLM based fact checking system, we propose RAV (Recon Answer Verify), an agentic framework with three agents: question generator, answer generator, and label generator. Our pipeline iteratively generates and answers sub questions to verify different aspects of the claim before finally generating the label. RAV generalizes across domains and label granularities, and it outperforms state of the art approaches on well known baselines RAWFC (fact checking, 3 class) by 25.28%, and on HOVER (encyclopedia, 2 class) by 1.54% on 2 hop, 4.94% on 3 hop, and 1.78% on 4 hop, sub categories respectively. RAV shows the least performance drop compared to baselines of 16.3% in macro f1 when we compare PFO with its unfiltered version.

Though safety alignment has been applied to most large language models (LLMs), LLM service providers generally deploy a subsequent moderation as the external safety guardrail in real-world products. Existing moderators mainly practice a conventional full detection, which determines the harmfulness based on the complete LLM output, causing high service latency. Recent works pay more attention to partial detection where moderators oversee the generation midway and early stop the output if harmfulness is detected, but they directly apply moderators trained with the full detection paradigm to incomplete outputs, introducing a training-inference gap that lowers the performance. In this paper, we explore how to form a data-and-model solution that natively supports partial detection. For the data, we construct FineHarm, a dataset consisting of 29K prompt-response pairs with fine-grained annotations to provide reasonable supervision for token-level training. Then, we propose the streaming content monitor, which is trained with dual supervision of response- and token-level labels and can follow the output stream of LLM to make a timely judgment of harmfulness. Experiments show that SCM gains 0.95+ in macro F1 score that is comparable to full detection, by only seeing the first 18% of tokens in responses on average. Moreover, the SCM can serve as a pseudo-harmfulness annotator for improving safety alignment and lead to a higher harmlessness score than DPO.

Multimodal automated fact-checking (MAFC) verifies claims by retrieving and reasoning over external evidence. However, most existing static benchmarks risk contamination: they primarily consist of outdated claims verifiable using an LLM's internal knowledge without external evidence. This can inflate performance estimates and fail to reflect true capability on novel claims that require up-to-date information. To address this, emerging dynamic benchmarks collect claims published after LLMs' knowledge cut-off dates, assuming they are uncontaminated. This work revisits this assumption by empirically studying contamination risks in both the state-of-the-art (SOTA) static AVeriTeC benchmark and our newly constructed dynamic ClaimReview2025Q4 benchmark, as well as their impact on MAFC evaluation. Our experiments yield 16 findings, highlighting three key results: (1) Dynamic evaluation reduces but does not eliminate contamination risks, as 17.09\%--29.30\% of post-cut-off claims remain potentially contaminated; (2) Many newly published claims can be verified either directly or by synthesizing multiple pieces of public knowledge available before the cut-off; and (3) Contamination can induce statistically significant inflation in MAFC performance, increasing Macro-F1 by up to 11.34 points and distorting system rankings. In light of these findings, we re-evaluate SOTA LLMs under a strictly contamination-controlled setting. Our study provides practical guidelines for trustworthy MAFC evaluation.

As LLM-based agents are deployed in production systems, understanding their behavioral consistency (whether they produce similar action sequences when given identical tasks) becomes critical for reliability. We study consistency in the context of SWE-bench, a challenging software engineering benchmark requiring complex, multi-step reasoning. Comparing Claude~4.5~Sonnet, GPT-5, and Llama-3.1-70B across 50 runs each (10 tasks times 5 runs), we find that across models, higher consistency aligns with higher accuracy: Claude achieves the lowest variance (CV: 15.2\%) and highest accuracy (58\%), GPT-5 is intermediate (CV: 32.2\%, accuracy: 32\%), and Llama shows the highest variance (CV: 47.0\%) with lowest accuracy (4\%). However, within a model, consistency can amplify both correct and incorrect interpretations. Our analysis reveals a critical nuance: consistency amplifies outcomes rather than guaranteeing correctness. 71\% of Claude's failures stem from "consistent wrong interpretation": making the same incorrect assumption across all runs. Interestingly, GPT-5 achieves similar early strategic agreement as Claude (diverging at step 3.4 vs.\ 3.2) but exhibits 2.1times higher variance, suggesting that divergence timing alone does not determine consistency. These findings suggest that for production deployment, interpretation accuracy matters more than execution consistency, with implications for agent evaluation and training.

Agentic systems are evaluated on benchmarks where agents interact with environments to solve tasks. Most papers report a pass@1 score computed from a single run per task, assuming this gives a reliable performance estimate. We test this assumption by collecting 60,000 agentic trajectories on SWE-Bench-Verified, spanning three models and two scaffolds. We find substantial variance: single-run pass@1 estimates vary by 2.2 to 6.0 percentage points depending on which run is selected, with standard deviations exceeding 1.5 percentage points even at temperature 0. This variance has critical implications: reported improvements of 2--3 percentage points may reflect evaluation noise rather than genuine algorithmic progress. Through token-level analysis, we show that trajectories diverge early, often within the first few percent of tokens, and that these small differences cascade into different solution strategies. To enable reliable evaluation of agentic systems, we recommend three concrete practices: (1) estimate pass@1 from multiple independent runs per task, especially when measuring small improvements, (2) use statistical power analysis to determine the number of runs needed to detect expected effect sizes, and (3) consider metrics like pass@k (optimistic bound) and pass^k (pessimistic bound) with k>1 to better characterize the full performance envelope. While these practices increase evaluation cost, they are essential for distinguishing genuine scientific progress from statistical noise.

Large language model (LLM) agents increasingly rely on skills, structured documents that specify when to act, which procedure to follow, and which tools are allowed. Existing evaluations mostly judge the quality of a skill or its contribution to task success, leaving unexamined whether an agent can recognize a relevant skill and apply it on its own. We introduce Skill-Use, a benchmark that evaluates skill use under progressive disclosure, where an agent sees only a skill's name and short description and must retrieve the full procedure before following it. Skill-Use separates three facets of skill use. Trigger measures whether the agent invokes the relevant skill, Compliance measures how faithfully it follows the prescribed procedure, and Boundary measures whether it avoids forbidden operations. A Skill-Use (SU) score combines the three and credits execution only after the skill is triggered. Skill-Use pairs 79 real skills with 177 executable tasks across nine domains, each grounded in real files, run in an isolated Docker sandbox, and scored by a trajectory-based rubric. Evaluating eight LLMs under two agent harnesses, we find that reliable skill use remains out of reach, as the strongest configuration reaches an SU of only 0.613. Triggering and procedural compliance fail as independent bottlenecks, and both scores and model rankings shift with the harness, so skill use behaves as a capability conditioned on the harness rather than a fixed property of the model.

We post-train Qwen3.8-27B for Korean response style -- verbosity, list and markdown usage, discourse structure and register -- and measure two behaviours the objective never targets: abstention on ambiguous social questions in KoBBQ, where the benchmark-correct answer is UNKNOWN, and unprompted disclosure in securities guidance. Both move, and the changes are expressed primarily through the model's emission policy: how often it answers and how much it says. Matched target-form controls show that answer propensity depends on the training target, not the prompt set or recipe alone. Holding prompts, recipe, data volume and serving fixed and changing only the target text, three style seeds give positive answer-rate point estimates (mean +0.82 pp) and three neutral seeds negative ones (mean -1.53 pp); the observed seed ranges do not overlap and the means differ by 2.34 pp. A length-matched arm lies between them, and a fourth arm that stays short while preserving hedging is unstable across seeds, so which feature of the form is responsible is unresolved. For absolute stereotyped exposure the decomposition into an answer-propensity term and a conditional-composition term is an algebraic identity, not a finding; its empirical content is where the movement went. Across the trained checkpoints the changes are dominated by answer propensity while the composition term stays small, and because that term is evaluated on treatment-dependent answered subsets we do not read it as evidence about latent preference. Two measurement results follow. A between-arm contrast in conditional stereotyped share does not identify a change in conditional content preference when answer status is treatment-dependent. And agreement between two rule detectors for the same construct runs from 0.44 to 0.99 depending on which checkpoint produced the text -- observable without any reference labels.

Practitioners make three prompt-design decisions with almost no controlled evidence behind them: how to format instructions and context (markdown, plain text, prose, or tabular), how many simultaneous instructions a system prompt can carry before compliance degrades, and how much context a model can hold before recall and honesty degrade. We report two controlled experiments crossing all three factors on one held, contamination-free synthetic corpus (the "Book of Veyra," 8,780 uniquely-named entities, deterministically regenerable from a fixed seed), evaluated across five models. Experiment 1 (960 calls/model) measures instruction-following decay as rule count N grows from 10 to 160, crossed with four formats and system-prompt vs. user-turn placement. Perfect-response rate collapses to zero by N=80 for every model, format, and placement. Placement produces effects at least as large as format at N=160 in most models, but the direction is model-specific. No model shows a reliable markdown advantage; one 35B model favors plain text instead. Experiment 2 (5,520 calls/model) measures recall accuracy, false-premise sycophancy, and absent-fact fabrication across a 2k-to-512k-token context ladder in the same four formats. Recall stays near ceiling through 64-128k tokens, then degrades sharply and format-dependently: one model's accuracy spread reaches 48 points at 128k tokens. Fabrication never occurs (0/5,760 probes), and sycophancy stays negligible (<=8.3%). What rises sharply near each model's context ceiling is outright refusal to answer (0% to 79-90%), distinct from sycophancy or fabrication. Neither pre-registered format ordering holds, and token overhead (+22% to +37% over plain text) further changes which format is preferable where accuracy spread is genuine. We release the full harness, corpus generator, and raw results (VeyraBench): https://github.com/iNetanel/veyrabench

Large Language Models (LLMs) are evolving beyond their classical role of providing information within dialogue systems to actively engaging with tools and performing actions on real-world applications and services. Today, humans verify the correctness and appropriateness of the LLM-generated outputs (e.g., code, functions, or actions) before putting them into real-world execution. This poses significant challenges as code comprehension is well known to be notoriously difficult. In this paper, we study how humans can efficiently collaborate with, delegate to, and supervise autonomous LLMs in the future. We argue that in many cases, "post-facto validation" - verifying the correctness of a proposed action after seeing the output - is much easier than the aforementioned "pre-facto validation" setting. The core concept behind enabling a post-facto validation system is the integration of an intuitive undo feature, and establishing a damage confinement for the LLM-generated actions as effective strategies to mitigate the associated risks. Using this, a human can now either revert the effect of an LLM-generated output or be confident that the potential risk is bounded. We believe this is critical to unlock the potential for LLM agents to interact with applications and services with limited (post-facto) human involvement. We describe the design and implementation of our open-source runtime for executing LLM actions, Gorilla Execution Engine (GoEX), and present open research questions towards realizing the goal of LLMs and applications interacting with each other with minimal human supervision. We release GoEX at https://github.com/ShishirPatil/gorilla/.

Multi-agent LLM systems that generate structured workflows from natural-language requests are now deployed in production across cloud automation, DevOps, and enterprise process orchestration. Operating such systems exposes a recurring change-management problem. Routine updates, such as re-running the same input, swapping the underlying LLM, or refactoring an agent's prompt or orchestration code, frequently produce workflows that differ substantially from previously validated references. Engineers are then left without a principled way to decide whether a change is safe to ship. Automatic workflow evaluation is the natural tool for answering this question. In practice, however, metric scores are poorly calibrated, and a numeric change rarely communicates the severity of the underlying degradation. We introduce WorkflowPerturb, a controlled benchmark for studying workflow evaluation metrics by applying realistic, graded perturbations to golden workflows. WorkflowPerturb contains 4,973 golden workflows and 44,757 perturbed variants across three perturbation types (Missing Steps, Compressed Steps, and Description Changes), each applied at severity levels of 10%, 30%, and 50%. We benchmark multiple metric families and analyze their sensitivity and calibration using expected score trajectories and residuals. Our results characterize systematic differences across metric families and support severity-aware interpretation of workflow evaluation scores in change-management settings. Our dataset will be released upon acceptance.

Multimodal retail agents should not only recognize what a customer is doing, but also decide whether and how to assist before an explicit request is made. We study this setting through the See--Infer--Intervene (SII) framework, where a device must see pre-interaction behavior, infer latent customer intent, and act by selecting an appropriate service intervention or choosing to wait. We instantiate SII with the Proactive Intent World Model (PIWM), which represents customer state with AIDA (Attention, Interest, Desire, Action) purchasing phases and BDI (belief, desire, intention) psychological fields, predicts action-conditioned intent transitions, and selects from five response classes: Greet, Elicit, Inform, Recommend, and Hold. We further construct GuidanceSalesBench, a smart-retail benchmark containing state manifests, pre-interaction videos, candidate responses, action-conditioned outcomes, and best-action labels. When conditioned on ground-truth customer state to isolate action selection, PIWM achieves 0.641 macro F1 on 30 held-out target videos, outperforming a zero-shot Qwen2.5-VL-7B baseline and training variants without balanced action supervision; end-to-end video-only selection drops to 0.295, below the 5-class balanced random baseline of 0.414, identifying video-to-state grounding as the dominant deployment-time bottleneck. A preliminary staged real-store pilot (recorded with paid participants performing scripted customer behaviors) reaches 0.579 action macro F1 on 20 fully annotated videos, with 10 additional accessible videos released with index-level labels.

Large language model (LLM) agents increasingly operate over evolving software repositories, where success depends on repository-specific procedural knowledge: which APIs to call, which scripts to run, and which conventions the current release expects. Agent skills externalize this knowledge into reusable units, and prior work shows that they can improve agent performance. What remains unclear is whether that improvement is durable. The same version specificity that makes a skill useful also makes it fragile: after a release, it may become stale without raising any explicit signal, while continuing to provide obsolete guidance. Externalizing knowledge into a skill can therefore make its decay invisible. We study whether agents can keep this externalized knowledge current. Repo2Skill-Evo casts each release transition as a skill-maintenance task: given a V1 skill set and the official V1-to-V2 patch, an agent must update obsolete skill content while preserving guidance that remains valid. Across 57 real-world repositories and 105 selected release transitions, every evaluated transition invalidates part of the V1 skill set. Yet six frontier agents reach only 29.9%-69.7% avg@3 macro F1 under a patch-grounded removal metric that balances stale-content recall against over-editing precision. Across runs, two opposing errors dominate: incomplete coverage of affected files in the skill set leaves stale content untouched, while overbroad editing is associated with higher recall but lower precision. Repository skills go stale in silence, and even frontier agents cannot reliably maintain them.

Macros are building block tasks of our everyday smartphone activity (e.g., "login", or "booking a flight"). Effectively extracting macros is important for understanding mobile interaction and enabling task automation. These macros are however difficult to extract at scale as they can be comprised of multiple steps yet hidden within programmatic components of mobile apps. In this paper, we introduce a novel approach based on Large Language Models (LLMs) to automatically extract semantically meaningful macros from both random and user-curated mobile interaction traces. The macros produced by our approach are automatically tagged with natural language descriptions and are fully executable. We conduct multiple studies to validate the quality of extracted macros, including user evaluation, comparative analysis against human-curated tasks, and automatic execution of these macros. These experiments and analyses show the effectiveness of our approach and the usefulness of extracted macros in various downstream applications.

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.

Monitoring complex industrial assets relies on engineer-authored symbolic rules that trigger based on sensor conditions and prompt technicians to perform corrective actions. The bottleneck is not detection but response: translating rules into maintenance steps requires asset-specific knowledge gained through years of practice. We investigate whether LLMs can serve as decision support for this rule-to-action step and introduce , a benchmark of 6{,}690 expert-validated multiple-choice questions from 118 rule-action pairs across 16 asset types. We contribute (i) a symbolic-to-MCQA pipeline normalizing rules to Disjunctive Normal Form with embedding-based distractor sampling, (ii) five variants probing distinct failure modes (Pro, Pert, Verbose, Aug, Rationale), and (iii) a benchmark of 29 LLMs and 4 embedding baselines. A human evaluation (9 practitioners, mean 45.0\%) confirms requires specialist knowledge beyond operational experience. Three findings stand out. The frontier has closed: the top three LLMs lie within one Macro point, with Bradley-Terry Elo placing claude-opus-4-6 30 points above the next model. Yet \,Pro exposes brittleness, with every model losing 13--60\% relative accuracy under distractor expansion. \,Aug exposes pattern-matching: under condition inversion, frontier models still select the original answer 49--63\% of the time. The deployment bottleneck is not capability but calibration: frontier models handle template-style fault detection but break under structural perturbation.

LLM coding agents are increasingly deployed to act autonomously on real production infrastructure. They execute shell commands, modify repositories, and call operational APIs. However, completing a task is not sufficient for safety. A wrong action can cause severe consequences. Existing agent benchmarks largely emphasize task completion, leaving open how agents behave under benign but underspecified instructions. We present UnderSpecBench, a benchmark for measuring action-boundary violations in coding agents (i.e., Claude Code, Codex, and OpenCode) on DevOps tasks. UnderSpecBench includes 69 task families grounded in documented incidents, CVEs, or tool behavior and organized across four DevOps capability domains and nine operational control surfaces. To isolate underspecification from task difficulty, each task keeps the same environment and ground-truth safe action while varying the instruction along three axes: intent clarity, target certainty, and blast radius. The resulting 2,208 prompt variants are evaluated with deterministic, side-effect-based oracles that separate Safe Success, Wrong Target, and OverScope outcomes; non-action runs are further classified as clarification, refusal, or deferment. Across five agent x model configurations using OpenCode, Claude Code, and Codex, the evaluation results show that underspecification does not mainly make agents fail; it makes them guess. 55.8-67.8% of runs violate at least one boundary. Target underspecification sharply degrades action quality, while blast-radius cues barely reduce action propensity. These findings show that completion-centric evaluation can overstate safe autonomy and motivate mitigations at the model, harness, and system layer.

From professional research to everyday planning, many tasks are bottlenecked by wide-scale information seeking, which is more repetitive than cognitively complex. With the rapid development of Large Language Models (LLMs), automated search agents powered by LLMs offer a promising solution to liberate humans from this tedious work. However, the capability of these agents to perform such "wide-context" collection reliably and completely remains largely unevaluated due to a lack of suitable benchmarks. To bridge this gap, we introduce WideSearch, a new benchmark engineered to evaluate agent reliability on these large-scale collection tasks. The benchmark features 200 manually curated questions (100 in English, 100 in Chinese) from over 15 diverse domains, grounded in real user queries. Each task requires agents to collect large-scale atomic information, which could be verified one by one objectively, and arrange it into a well-organized output. A rigorous five-stage quality control pipeline ensures the difficulty, completeness, and verifiability of the dataset. We benchmark over 10 state-of-the-art agentic search systems, including single-agent, multi-agent frameworks, and end-to-end commercial systems. Most systems achieve overall success rates near 0\%, with the best performer reaching just 5\%. However, given sufficient time, cross-validation by multiple human testers can achieve a near 100\% success rate. These results demonstrate that present search agents have critical deficiencies in large-scale information seeking, underscoring urgent areas for future research and development in agentic search. Our dataset, evaluation pipeline, and benchmark results have been publicly released at https://widesearch-seed.github.io/

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.

Large language models (LLMs) have advanced rapidly from conversational problem solving to addressing real-world tasks involving tool use, such as software engineering (SWE). Recent LLM-powered toolkits, such as OpenAI Codex and Cursor, have offered end-to-end automation of the software development process. However, building effective SWE agents remains challenging due to the lack of high-quality training data and effective test cases. To address this issue, we present SWE-Dev, an SWE agent built upon open-source LLMs. First, we develop a robust pipeline to synthesize test cases for patch evaluation. Second, we scale up agent trajectories to construct the training data for building SWE-Dev. Experiments on the SWE-bench-Verified benchmark show that the SWE-Dev models can achieve top performance among all open SWE agents. Specifically, the success rates of the SWE-Dev 7B and 32B parameter models reach 23.4% and 36.6%, respectively, outperforming state-of-the-art open-source models. All code, models, and datasets are publicly available at https://github.com/THUDM/SWE-Dev.

Language-model agents are increasingly deployed under standing instructions: a system prompt, a policy file, or a skills document is placed in context, and the agent is trusted to let that document govern every action that follows. Existing benchmarks rarely test this deployment pattern directly; they measure whether an agent can complete a task, not whether a long, binding policy document constrains its behavior over an extended tool-use horizon. We present HANDBOOK_md, a benchmark of 65 agentic tasks modeled on how employees follow company handbooks. Each task places an agent in a self-contained company environment (a file workspace with mock email, chat, calendar, issue-tracking, and commerce services exposed over the Model Context Protocol) and instructs it to carry out routine professional work governed by an expert-written standard operating procedure of 20-124 pages. Tasks span five domains (finance, medical billing, insurance, logistics, and HR) and 10 fictional companies. To resist memorization, every task modifies one of 10 base handbooks, altering the specific rules and thresholds on which grading depends, so no two tasks share the same set of policies. Grading is fully deterministic: each task carries a rubric of programmatic criteria (824 in total) that check both that required actions occurred and that prohibited actions did not. Under strict grading, where a trial passes only if every criterion is satisfied, the strongest evaluated model passes 36.2% of trials, and most frontier models remain below 25%. Failures follow consistent patterns: agents let a plausible but unauthorized in-environment request override the standing policy, perform a required check and then act against its result, lose rule details over long horizons, and report compliance they did not achieve. We release the tasks, environments, and evaluation harness.

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

Dexterous manipulation policies learned by imitation are typically evaluated for robustness to variation in scenes, objects, or instructions, but their performance across task execution speeds is less often examined. This leaves open how much temporal robustness a learner retains relative to the expert it imitates. We compare an expert and learner under the same task conditions, initial-condition draws, and speedup factors. We instantiate the evaluation in ParcelStow, a contact-rich task in which the robot acquires, reorients, and inserts a parcel. The demonstrations span the speedup range for the manipulation phases after parcel acquisition. A scripted expert and an Action Chunking with Transformers (ACT) policy trained from the expert's demonstrations both achieve 100 percent task success at nominal speed. Their success rates diverge within the demonstrated range: at its maximum, expert success is 84 percent and ACT success is 53 percent. Two ACT policies with different parameter initializations show similar degradation, decreasing by 34 and 48 percentage points from nominal speed to the maximum demonstrated speed, compared with 16 points for the expert. Stage-level analysis shows that 35 of ACT's 47 failures at the maximum demonstrated speed are insertion misalignments. Under the relative-motion handoff, every ACT acquisition retains the parcel through reorientation and transfer in free space, but only 64 percent complete the overall task, compared with 95 percent after expert acquisition. Across all evaluated policies and speeds, none of the 414 acquisitions without force closure completes the task. Equal nominal task success therefore does not imply preservation of expert performance across execution speeds. Code, data, and evaluation scripts are available at https://github.com/coenwerem/parcelstow.

Large language models (LLMs) have raised hopes for automated end-to-end fact-checking, but prior studies report mixed results. As mainstream chatbots increasingly ship with reasoning capabilities and web search tools -- and millions of users already rely on them for verification -- rigorous evaluation is urgent. We evaluate 15 recent LLMs from OpenAI, Google, Meta, and DeepSeek on more than 6,000 claims fact-checked by PolitiFact, comparing standard models with reasoning- and web-search variants. Standard models perform poorly, reasoning offers minimal benefits, and web search provides only moderate gains, despite fact-checks being available on the web. In contrast, a curated RAG system using PolitiFact summaries improved macro F1 by 233% on average across model variants. These findings suggest that giving models access to curated high-quality context is a promising path for automated fact-checking.

Interactive dialogue games test a capability that static benchmarks largely leave implicit: a model must carry state across turns, interpret feedback, and choose valid actions under changing constraints. We study this setting in the LM Playschool Challenge with a 2B open-weight model, and find that many failures are not only broad knowledge failures but also local decision failures: repeated guesses, malformed actions, and violations of feedback that the model has just seen. These diagnostics motivate a training recipe organized around three steps: acquire broad game participation through supervised fine-tuning, repair mechanically verifiable failures within one targeted dialogue-game family using turn-local preference pairs, and preserve general capabilities beyond these dialogue games. In the official final evaluation, our submission improves public clemscore from 10.67 to 38.92 and closed in-domain score from 13.41 to 41.17, while approximately preserving aggregate static performance (44.14 vs. 44.24 for the baseline). Out-of-domain clemscore remains low at 7.88, with the largest gains concentrated in unseen variants of the targeted family. Our results suggest that broad SFT brings most of the model's capability improvement; turn-local supervision can be effective when failure detection is precise, with observed transfer concentrated primarily within-family.

We study how runtime enforcement against unsafe actions affects end-to-end task performance in multi-step tool using large language model (LLM) agents. Using tau-bench across Airline and Retail domains, we compare baseline Tool-Calling, planning-integrated (TRIAD), and policy-mediated (TRIAD-SAFETY) architectures with GPT-OSS-20B and GLM-4-9B. We identify model dependent interaction horizons (15 to 30 turns) and decompose outcomes into overall success rate (SR), safe success rate (SSR), and unsafe success rate (USR). Our results reveal a persistent Safety Capability Gap. While safety mediation can intercept up to 94 percent of non-compliant actions, it rarely translates into strictly safe goal attainment (SSR below 5 percent in most settings). We find that high unsafe success rates are primarily driven by Integrity Leaks, where models hallucinate user identifiers to bypass mandatory authentication. Recovery rates following blocked actions are consistently low, ranging from 21 percent for GPT-OSS-20B in simpler procedural tasks to near zero in complex Retail scenarios. These results demonstrate that runtime enforcement imposes a significant verifier tax on conversational length and compute cost without guaranteeing safe completion, highlighting the critical need for agents capable of grounded identity verification and post-intervention reasoning.

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 (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

Personal AI agents like OpenClaw run with elevated privileges on users' local machines, where a single successful prompt injection can leak credentials, redirect financial transactions, or destroy files. This threat goes well beyond conventional text-level jailbreaks, yet existing safety evaluations fall short: most test models in isolated chat settings, rely on synthetic environments, and do not account for how the agent framework itself shapes safety outcomes. We introduce CLAWSAFETY, a benchmark of 120 adversarial test scenarios organized along three dimensions (harm domain, attack vector, and harmful action type) and grounded in realistic, high-privilege professional workspaces spanning software engineering, finance, healthcare, law, and DevOps. Each test case embeds adversarial content in one of three channels the agent encounters during normal work: workspace skill files, emails from trusted senders, and web pages. We evaluate five frontier LLMs as agent backbones, running 2,520 sandboxed trials across all configurations. Attack success rates (ASR) range from 40\% to 75\% across models and vary sharply by injection vector, with skill instructions (highest trust) consistently more dangerous than email or web content. Action-trace analysis reveals that the strongest model maintains hard boundaries against credential forwarding and destructive actions, while weaker models permit both. Cross-scaffold experiments on three agent frameworks further demonstrate that safety is not determined by the backbone model alone but depends on the full deployment stack, calling for safety evaluation that treats model and framework as joint variables. Code and data will be available at: https://weibowen555.github.io/ClawSafety/.

Large Language Model (LLM)-powered multi-agent systems (MAS) have rapidly advanced collaborative reasoning, tool use, and role-specialized coordination in complex tasks. However, reliability-critical deployment remains hindered by a systemic failure mode: hierarchical compliance under instruction conflicts (system-user, peer-peer), where agents misprioritize system-level rules in the presence of competing demands. Moreover, widely used macro-level metrics (e.g., pass@k) obscure these micro-level violations and offer little actionable guidance for remedy. In this work, we present a full-stack, three-stage framework: (1) Diagnose - Contextualized Role Adherence Score (CRAS), a query-wise, context-aware scoring metric that decomposes role adherence into four measurable dimensions; (2) Localize - attention drift analysis revealing that instruction conflicts are resolved by attention heads that are largely concentrated in middle layers; (3) Align - Surgical Alignment of Instruction Layers (SAIL), which installs LoRA only on the localized focal layers and optimizes a token-weighted DPO-style preference objective that credits tokens by their focal attentional contribution. Across standard benchmarks and MAS frameworks, our surgical approach improves instruction hierarchy compliance (e.g., +5.60% with AutoGen on MedQA) without full-model finetuning.

Skills are becoming the capability layer through which LLM agents turn plans into actions, but their use introduces security risks such as data leakage, unauthorized operations, and tool misuse. Existing vetting usually evaluates each skill in isolation, while real agent tasks often invoke multiple skills in a shared execution context. This creates Skill Composition Risk (SCR): a skill that appears benign alone can become harmful when its outputs, trust signals, authorization cues, or side effects influence later invocations along an activated path. We introduce SCR-Bench to evaluate this risk in controlled, sandboxed skill environments. Rather than relying only on textual intent or surface behavior, SCR-Bench records downstream state changes and path-level outcomes across composed skill executions. It contains three sub-benchmarks: SCR-CapFlow for capability-flow composition, SCR-TrustLift for trust-transfer composition, and SCR-AuthBlur for authorization-confusion composition. Across SCR-Bench, composed paths expose risks that are largely absent under isolated evaluation. In SCR-CapFlow, attack success rate reaches 33.6 percent under composition, compared with near-zero isolated baselines. In SCR-TrustLift, attack success rate exceeds 96.5 percent on four of five backends. In SCR-AuthBlur, the risky-approval rate increases by 71.8 percent relative to the L0 isolated baseline under the L1 context setting. These results show that agent skill security should be assessed at the level of activated paths rather than isolated artifacts. SCR and SCR-Bench provide a foundation for path-aware risk evaluation and defense in LLM agent skill ecosystems. Benchmark: https://github.com/saint-viperx/SCR_Bench.

Evaluating LLM agents is essential for guiding their development, yet it has grown prohibitively expensive: a single pass of a frontier model over an agentic benchmark can cost hundreds to thousands of dollars, a price paid repeatedly across iterative development cycles. Prior efforts, centered on benchmark distillation, reduce the number of evaluation tasks but leave the cost of executing each retained task untouched. In this work, we introduce early outcome prediction, a complementary axis of efficiency that instead cuts cost within each task. Our key insight is that an agent's final outcome is often evident from its intermediate behavior well before execution completes. We instantiate this idea in EarlyEval, a lightweight framework that trains a pair of LightGBM success and failure classifiers over behavioral, textual, and reference-solution features, and halts an agent run the moment either classifier crosses a calibrated confidence threshold, adding negligible per-step overhead. Across three benchmarks, SWE-bench Verified, TerminalBench, and Toolathlon, EarlyEval can eliminate 13%-26% of agent steps and up to 44.1% input tokens and 29.4% output tokens at 89%-97% prediction accuracy, while perturbing per-agent resolve rates by only one to two percentage points on average.

Standard accuracy benchmarks are designed to test how closely large language models (LLMs) approach correct answers, but are not suitable for testing whether LLMs stick with a correct answer when that answer is challenged by a plausible counter-argument. We introduce a controlled protocol for evaluating answer stability: after a model answers a multiple-choice question correctly, we challenge the model's answer with a coherent argument for an incorrect option and measure whether the model flips. The setup a) isolates argumentative content from overt social pressure and b) varies argument length, self-attribution, and cross-model source. Across seven frontier models and 57 MMLU subjects, flip rates range from 17.5% to 97.3%, revealing large differences in stability that are not captured by accuracy metrics alone. We find that self-attribution consistently increases flip rates (mean +7.1pp, up to +18.7pp). Also, pooling wrong-answer arguments across models and selecting the most effective one per question yields stronger adversarial challenges than relying on any single source model. We further construct MaxFlip, a curated challenge set that amplifies flips by up to +23.6pp over standard self-generated challenges. We release the protocol, challenge records, and MaxFlip to support stability evaluation alongside standard accuracy benchmarks. Materials are available at https://github.com/nafisenik/WhoFlips and https://hf.co/datasets/nafisehNik/WhoFlips.

Automated Code Review (ACR) is crucial for software quality, yet existing benchmarks often fail to reflect real-world complexities, hindering the evaluation of modern Large Language Models (LLMs). Current benchmarks frequently focus on fine-grained code units, lack complete project context, and use inadequate evaluation metrics. To address these limitations, we introduce SWRBench , a new benchmark comprising 1000 manually verified Pull Requests (PRs) from GitHub, offering PR-centric review with full project context. SWRBench employs an objective LLM-based evaluation method that aligns strongly with human judgment (~90 agreement) by verifying if issues from a structured ground truth are covered in generated reviews. Our systematic evaluation of mainstream ACR tools and LLMs on SWRBench reveals that current systems underperform, and ACR tools are more adept at detecting functional errors. Subsequently, we propose and validate a simple multi-review aggregation strategy that significantly boosts ACR performance, increasing F1 scores by up to 43.67%. Our contributions include the SWRBench benchmark, its objective evaluation method, a comprehensive study of current ACR capabilities, and an effective enhancement approach, offering valuable insights for advancing ACR research.

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.

Reinforcement learning (RL) for terminal agents needs executable training environments with reliable rewards and useful difficulty. Fixed recipes such as few-shot, Self-Instruct, and Evol-Instruct apply the same prompting policy to every seed, even when the current policy would benefit from a harder, easier, or simply different task. We present Envs-FORGE, a prompting policy that converts verifier rewards into per-seed environment-synthesis actions. Envs-FORGE estimates seed pass rates, scores six projection--direction actions around a target learning frontier, and solves a per-seed mixed-integer linear program (MILP) to choose the action that conditions generation. The selected action drives synchronized rewriting of the instruction, fixtures, oracle solution, tests, and Docker environment; only gold-verified bundles enter RL training. The indexed MILP form also supports optional soft skill coverage for portfolio planning. On Qwen 3.5 35B, Envs-FORGE improves Pass@1 over Base by 9.2 percentage points on tb-core (40.0% to 49.2%) and 6.4 points on tb-2.0 (23.0% to 29.4%), exceeding the strongest fixed-recipe baseline by 2.4 and 2.1 points. It reaches 77.1% on SWE-bench Verified versus 73.4% for Base, and improves tb-core by 6.8--9.2 points across the evaluated 4B--35B models. All synthesis methods export 100 verified environments and use 2.27M--2.88M synthesis tokens, placing the comparison at the same downstream training-set size and the same operational scale. The source code is available at https://github.com/DataArcTech/DataArc-SynData-Toolkit/.

Computer-use agents pay full frontier inference to re-derive routines their user has already performed, because an agent's memory today records what the user said, not what the user did. We compile passively captured screen activity into agent memory with a deterministic, zero-model pipeline: it segments a local capture stream into typed activity frames, bounded episodes carrying application, site, timing, input volume, and evidence pointers back to the raw rows, with no model in the loop, so the output is byte-identical, cacheable, and mechanically auditable. On one professional's single-user corpus of 128,756 frames over 51 active days, the compiler reduces a day of raw capture to a prompt-ready context block 86x smaller in 68 ms, and an agent reading that block answers questions about the day at 98.4% accuracy (Wilson 95% CI 91.7-99.7%) against an independent oracle, versus 66-80% for an LLM summary of the same capture, a mid-tier model reading the block matching a frontier one. The same compiler doubles as a demand-side cost instrument. Read off passive, pre-delegation human activity rather than agent rollouts, it supplies two parameters that agent-cost models assume but, to our knowledge, have not measured: the Routine Overhead Ratio R and the routine recurrence h. We report first values of R, a modeled upper bound, at 60-343x, and a delegable recurrence of 9.0% in-sample and 7.7% out-of-sample, for a realistic all-fleet token ceiling near 8%; a compiled routine replays deterministically with the model out of the loop, demonstrated live at zero model tokens on a guard-matched hit. Schema, compiler, and evaluation harness are open.

Tool-Augmented Language Models (TALMs) leverage external APIs to answer user queries across various domains. However, existing benchmark datasets for TALM research often feature simplistic dialogues that do not reflect real-world scenarios, such as the need for models to ask clarifying questions or proactively call additional APIs when essential information is missing. To address these limitations, we construct and release ToolDial, a dataset comprising 11,111 multi-turn dialogues, with an average of 8.95 turns per dialogue, based on APIs from RapidAPI. ToolDial has two key characteristics. First, the dialogues incorporate 16 user and system actions (e.g., "Request", "Clarify", "Fail inform") to capture the rich dynamics of real-world interactions. Second, we simulate dialogues where the system requests necessary information from the user based on API documentation and seeks additional APIs if the user fails to provide the required information. To facilitate this process, we introduce a method for generating an API graph that represents input and output compatibility between APIs. Using ToolDial, we evaluate a suite of language models on their ability to predict correct actions and extract input parameter values for API calls from the dialogue history. Modern language models achieve accuracy scores below 70%, indicating substantial room for improvement. We release our dataset and code at https://github.com/holi-lab/ToolDial.

Agent skills are the de facto mechanism for extending LLM agents with reusable guidance. A skill can shape the agent's task execution, including planning, tool use, problem-solving, and validation. Prior work reported mixed results of agent skills: some skills improve task success rates, while others have no effect, increase token use and execution time, and even reduce success rates. This paper presents a comprehensive analysis of skill-induced agent failures by attributing task failures and cost regressions to specific loaded skills. We introduce a differential analysis framework that attributes a failure or regression to a skill by comparing a target skill-guided run against a no-skill or semantically matched skill reference run that solves the same task, or solves it more cheaply. We instantiate this framework on SkillsBench and SWE-Skills-Bench, yielding 307 skill-induced failures, including 125 functional failures and 182 efficiency regressions. We also build SkillTriage, a taxonomy-guided attribution tool that normalizes paired cases, extracts differential evidence, and produces triage reports. Our major findings include: (1) Skill induced functional failures are rarely caused by obviously irrelevant skills; instead, seemingly relevant skills often make the agent incorrectly implement or omit task-required implementation elements. (2) Skill-induced efficiency regressions are not explained by prompt length alone. (3) The largest sources within Excessive Procedure are excessive verification and heavy implementation pipelines, contributing 67 and 30 cases, respectively. This shows that skills often turn validation checklists and construction recipes into mandatory work. Based on our findings, we propose research topics and tooling improvements for safer and more cost-aware skill reuse.

Benchmarks increasingly guide deployment, procurement and scientific screening, yet a score supports only the response it records, not necessarily the deployment action. We introduce deployment-complete benchmarking, which tests whether benchmark evidence determines a deployment action. A benchmark is complete for a claim exactly when the action is constant on each evidence fiber; mixed fibers expose missing deployment information, and completion curves quantify the evidence required to resolve ambiguity. In controlled response spaces, benchmark-channel conformal coverage of 94.98% transferred poorly to an unmeasured deployment channel (10.07%), whereas response-rank intervals achieved 94.91% coverage; even zero benchmark error certified only 45.4% of candidates at the largest residual size. Public audits revealed incompleteness, including 97.9% mixed Tox21 fibers and zero median certifiable fraction in main Matbench and JARVIS audits. In held-out replays, certify-then-acquire reduced false decisions from 1.19% to 0.027% in Tox21 and from 20.3% to 0.128% in JARVIS, while changing model choice and identifying deployment-relevant probes. Deployment-ready benchmarks should report evidence, supported actions, ambiguity and completion cost rather than scores alone.

Vision-language-action models (VLAs) have been extensively used in robotics applications, achieving great success in various manipulation problems. More recently, VLAs have been used in long-horizon tasks and evaluated on benchmarks, such as BEHAVIOR1K (B1K), for solving complex household chores. The common metric for measuring progress in such benchmarks is success rate or partial score based on satisfaction of progress-agnostic criteria, meaning only the final states of the objects are considered, regardless of the events that lead to such states. In this paper, we argue that using such evaluation protocols say little about safety aspects of operation and can potentially exaggerate reported performance, undermining core challenges for future real-world deployment. To this end, we conduct a thorough analysis of state-of-the-art models on the B1K Challenge and evaluate policies in terms of robustness via reproducibility and consistency of performance, safety aspects of policies operations, task awareness, and key elements leading to the incompletion of tasks. We then propose evaluation protocols to capture safety violations to better measure the true performance of the policies in more complex and interactive scenarios. At the end, we discuss the limitations of the existing VLAs and motivate future research.

LLM-based agents can act on behalf of a user to access cloud services, call tools, or invoke agents. At session start, the agent's permissions are set but remain static, and each request is evaluated independently, without considering prior actions. Within its permissions, an agent may act contrary to the delegated task, combine individually permitted actions into a prohibited outcome, or delegate authority to a sub-agent without limiting it. A prompt injection poses a risk only if the agent has authority to perform such actions; this is therefore a problem of authorization architecture, not just the model. The Agentic Principal Chain (APC) tracks delegated authority from one principal to the next. APC evaluates each request against the accumulated session state using six authorization checks. APC carries forward and restricts delegated scope and budgets. Using composition closure, APC checks requests against prior actions to prevent prohibited combinations and enforces the decision outside the model. We prove Blast Radius Monotonicity and Composition Soundness for APC implementations; Composition Soundness is limited to prohibited combinations under a complete restriction set and serialized admission. We evaluated 3,154 instances including InjecAgent, AgentDojo, and ASB. Our compromised-model evaluation tests APC independently of model behavior by inserting the ground-truth attack call after the first legitimate tool call. AgentDojo exfiltration fell from 75-100% to 0% across all four domains; APC blocked all 544 InjecAgent data-stealing cases. Intent binding reduced destruction from 38.6% to 4.0% and manipulation from 90.5% to 12.1%. Authorization latency was 0.24 ms at the 99th percentile on an idle host; across 949 AgentDojo task-injection pairs, utility was 8.6 and 13.9 percentage points lower in the two settings. Implementation, evaluation tools, and data are publicly available.

Small Language Models (SLMs) are increasingly deployed in resource-constrained, privacy-sensitive settings, where safety and bias failures can cause security and societal risks. However, existing AI safety\slash security\slash compliance benchmarks are designed for large language models that may not transfer reliably to SLMs. We therefore ask: Can these benchmarks effectively and reliably evaluate SLMs? To answer this question, we conduct a large-scale assessment of the effectiveness and robustness of these automated pipelines by evaluating five widely used benchmark suites across 26 open-source SLMs under a unified judging rubric, which assigns a score of 0, 1, or 0.5 to harmful, safe, or ambiguous/irrelevant responses, respectively. Across the benchmarks, ambiguous judgments dominate and correlate with prompt complexity and model architecture, indicating that {\em LLM-centric safety benchmarks are insufficient as standalone evidence for SLM safety assessment}. In general, the ambiguity rate increases with lexical density, output perplexity, and output length and decreases with lexical sophistication, self-coherence, and reply-prompt similarity. This reveals a capability-safety confound that mixes model capability with apparent safety. Since ambiguity is prevalent, aggregate mean-score leaderboards are mathematically brittle: model rankings change significantly under reasonable ambiguity treatments, even when the underlying outputs remain unchanged.

Agents based on Large Language Models (LLMs) have shown promise for performing sophisticated software engineering tasks autonomously. In addition, there has been progress towards developing agents that can perform parts of the research pipeline in machine learning and the natural sciences. We argue that research extension and its implementation is a critical capability for such systems, and introduce RExBench to support the evaluation of this capability. RExBench is a benchmark consisting of 12 realistic research experiment implementation tasks that aim to investigate research hypotheses that have not previously been implemented. Each task is set up as an extension to an existing research paper and codebase, accompanied by domain expert-written instructions. RExBench is robust to data contamination, and supports an automatic evaluation infrastructure that executes agent outputs to determine whether the success criteria are met. We use this benchmark to evaluate nine LLM agents implemented using three different frameworks: aider, Claude Code, and OpenHands. We find that all agents evaluated fail to autonomously implement the majority of the extensions. Although the success rate improves with additional human-written hints, the best performance under this setting remains below 40%. This indicates that current agents are still short of being able to handle realistic research extension tasks without substantial human guidance.

Abstention Ability (AA) is a critical aspect of Large Language Model (LLM) reliability, referring to an LLM's capability to withhold responses when uncertain or lacking a definitive answer, without compromising performance. Although previous studies have attempted to improve AA, they lack a standardised evaluation method and remain unsuitable for black-box models where token prediction probabilities are inaccessible. This makes comparative analysis challenging, especially for state-of-the-art closed-source commercial LLMs. This paper bridges this gap by introducing a black-box evaluation approach and a new dataset, Abstain-QA, crafted to rigorously assess AA across varied question types (answerable and unanswerable), domains (well-represented and under-represented), and task types (fact centric and reasoning). We also propose a new confusion matrix, the ''Answerable-Unanswerable Confusion Matrix'' (AUCM) which serves as the basis for evaluating AA, by offering a structured and precise approach for assessment. Finally, we explore the impact of three prompting strategies-Strict Prompting, Verbal Confidence Thresholding, and Chain-of-Thought (CoT)-on improving AA. Our results indicate that even powerful models like GPT-4, Mixtral 8x22b encounter difficulties with abstention; however, strategic approaches such as Strict prompting and CoT can enhance this capability.

Recent position papers argue that the classical aleatoric/epistemic uncertainty framework is insufficient for interactive large language model (LLM) agents and call for underspecification-aware, decomposed, and communicable uncertainty representations that can unlock new agent capabilities such as proactive clarification seeking and shared mental-model building. Practical deployment constraints -- black-box APIs, interactive latency budgets, and the absence of labeled trajectories -- rule out logprob-based, multi-sampling, and training-based methods, leaving prompt-based estimation as the most viable family for surfacing such signals at deployment time. We answer this call with a simple prompt-based decomposition that separates action confidence from request uncertainty (u), enabling the agent to ask for clarification when the task specification is ambiguous. To evaluate it, we introduce two clarification-augmented benchmarks (WebShop-Clarification and ALFWorld-Clarification) in which 50% of tasks are deliberately underspecified, and systematically compare the proposed decomposition against ReAct+UE and Uncertainty-Aware Memory (UAM) across five LLM backbones (GPT-5.1, DeepSeek-v3.2-exp, GLM-4.7, Qwen3.5-35B, GPT-OSS-120B) on these variants together with the standard WebShop, ALFWorld, and REAL benchmarks for fault detection. Averaged across the five backbones, the proposed decomposition improves clarification F1 on ALFWorld-Clarification by 73% over ReAct+UE and by 36% over UAM, and leads clarification F1 on every backbone on WebShop-Clarification and on four of five backbones on ALFWorld-Clarification, indicating that the gains generalize beyond a single LLM.

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.

This paper describes a system designed to distinguish between AI-generated and human-written scientific excerpts in the DAGPap24 competition hosted within the Fourth Workshop on Scientific Document Processing. In this competition the task is to find artificially generated token-level text fragments in documents of a scientific domain. Our work focuses on the use of a multi-task learning architecture with two heads. The application of this approach is justified by the specificity of the task, where class spans are continuous over several hundred characters. We considered different encoder variations to obtain a state vector for each token in the sequence, as well as a variation in splitting fragments into tokens to further feed into the input of a transform-based encoder. This approach allows us to achieve a 9% quality improvement relative to the baseline solution score on the development set (from 0.86 to 0.95) using the average macro F1-score, as well as a score of 0.96 on a closed test part of the dataset from the competition.

Recent advances in LLM agents have largely built on reasoning backbones like ReAct, which interleave thought and action in complex environments. However, ReAct often produces ungrounded or incoherent reasoning steps, leading to misalignment between the agent's actual state and goal. Our analysis finds that this stems from ReAct's inability to maintain consistent internal beliefs and goal alignment, causing compounding errors and hallucinations. To address this, we introduce ReflAct, a novel backbone that shifts reasoning from merely planning next actions to continuously reflecting on the agent's state relative to its goal. By explicitly grounding decisions in states and enforcing ongoing goal alignment, ReflAct dramatically improves strategic reliability. This design delivers substantial empirical gains: ReflAct surpasses ReAct by 27.7% on average, achieving a 93.3% success rate in ALFWorld. Notably, ReflAct even outperforms ReAct with added enhancement modules (e.g., Reflexion, WKM), showing that strengthening the core reasoning backbone is key to reliable agent performance.

Multi-step enterprise agent tasks fail in a characteristic way: single-pass inference has no checkpoint between deciding an answer and committing to it. We study one production system (Leni) whose architecture installs such checkpoints: verification loops (execute, observe, compare, correct) staffed by lightweight task-specialized post-trained models. We evaluate the unmodified production configuration on three public benchmarks stressing distinct failure modes: SpreadsheetBench Verified (silent computation error), BullshitBench v2 (premise confabulation), and the GAIA validation split (cascade error over long tool chains). The full system improves over its frontier base model by +11.0 percentage points on SpreadsheetBench (91.25% vs 80.25%, n=400, p<0.001), +7 to +10 percentage points on BullshitBench (98% vs 91%, n=100), and roughly +15 points on GAIA validation (75.2% pass@1, n=165; 83.0% best-of-k). Our central contribution is a decomposition of that uplift: most of it comes from scaffolding, routing, and specialist models rather than from the verification step itself, whose isolated contribution is small (+1.5 points) but concentrated at the top of the score distribution, where it converts otherwise-failing tasks. We instrument the loop end-to-end, yielding an empirical verifier confusion matrix (catch rate about 0.20, fix rate 0.75, no false-alarm regressions) that grounds a compounding-reliability model. Specialist-swap ablations suggest that the loop's value depends on who observes it: replacing the small trained verifier with the generating frontier model eliminates most rescues. A valid-premise control shows zero over-rejections in 100 expert-level questions.

Coding agents are now commonly evaluated on the SWE-bench family of benchmarks, whose tasks are built from curated GitHub issues: long, structured, and information-rich. Real user requests, however, are typically far shorter and less structured. To characterize this gap, we define a six-category information taxonomy and four dimensions of linguistic style, and apply them to real user prompts from SWE-chat and problem statements from SWE-bench Verified and Pro. We find that requests carrying only a problem statement, alone or with limited additional context, account for 88% of real prompts but just 7% of benchmark problems. Furthermore, 87% of real prompts are casually written whereas 94% of benchmark problems are formal. Guided by these observations, we introduce RealSWE, 381 multi-variant task families derived from SWE-bench Verified and Pro. Variants within each family share the same underlying task and gold patch while differing only in information composition and linguistic style. Evaluating seven contemporary LLMs with RealSWE, we find that i) realistic inputs reduce resolution rates by 6.4 pp on average and can change model rankings. Controlled analysis further shows that ii) including Desired Behavior and Motivation significantly affects performance, whereas Environment Information and Reproduction Steps merely add tokens without measurable benefit; iii) linguistic style has only small, model-dependent effects. These findings provide actionable guidance for users and agents: explicitly stating the desired behavior and motivation, which most real prompts omit, substantially improves the LLM's software engineering performance.

An abundance of datasets and availability of reliable evaluation metrics have resulted in strong progress in factoid question answering (QA). This progress, however, does not easily transfer to the task of long-form QA, where the goal is to answer questions that require in-depth explanations. The hurdles include (i) a lack of high-quality data, and (ii) the absence of a well-defined notion of the answer's quality. In this work, we address these problems by (i) releasing a novel dataset and a task that we call ASQA (Answer Summaries for Questions which are Ambiguous); and (ii) proposing a reliable metric for measuring performance on ASQA. Our task focuses on factoid questions that are ambiguous, that is, have different correct answers depending on interpretation. Answers to ambiguous questions should synthesize factual information from multiple sources into a long-form summary that resolves the ambiguity. In contrast to existing long-form QA tasks (such as ELI5), ASQA admits a clear notion of correctness: a user faced with a good summary should be able to answer different interpretations of the original ambiguous question. We use this notion of correctness to define an automated metric of performance for ASQA. Our analysis demonstrates an agreement between this metric and human judgments, and reveals a considerable gap between human performance and strong baselines.

We placed 100 memory-equipped large language model (LLM) agents in charge of a closed, money-conserving spatial economy on real Pokhara Lakeside geography (earning wages, running businesses, setting prices) and ran this multi-agent simulation for up to 26 simulated weeks, well past the 1-2 weeks typical of agent-society studies. Across 91 validated runs (2.44M agent decisions, 21.5B tokens), the money stops moving, in a specific and measurable way. A 12x tourist demand shock raises business revenue 4.62x (p<0.001), which we decompose exactly into a 1.50x extensive margin (more businesses trading) and a 3.07x intensive margin (more revenue each). Monetary transmission stops there. Wages move 1.03x (p=0.42); 0.3% of 3,981 menu items are ever repriced (p=0.47). A randomized cash transfer (NPR 5,000 to 20 of 100 agents) shows the same pattern from the opposite direction: 96.7% is still held 311 pulses later, marginal propensity to consume 3-4% by two independent measures, indistinguishable from zero. The wealth distribution is consequently near-frozen at the horizon this literature uses (ρ=0.964 over 2 simulated weeks), but not frozen. ρ falls to 0.832 at 12 weeks and 0.752 at 26, a horizon-dependence no short study can see. Matched ablations show which knob actually matters. Swapping the backing LLM moves every outcome we measure (p=0.0039); deleting agents' memory moves none of them detectably. A purely social tool fails 94-97% of the time across two model families, compared with ~96% success on economic tools, with no measurable shift away from it. Every headline number is verified twice, by a live validator and by an offline recomputation that reconciles each agent's wealth against its own signed transaction history, and we release the full run corpus for reanalysis.

The rise of LLMs (Large Language Models) has contributed to the improved performance and development of cutting-edge NLP applications. However, these can also pose risks when used maliciously, such as spreading fake news, harmful content, impersonating individuals, or facilitating school plagiarism, among others. This is because LLMs can generate high-quality texts, which are challenging to differentiate from those written by humans. GLTR, which stands for Giant Language Model Test Room and was developed jointly by the MIT-IBM Watson AI Lab and HarvardNLP, is a visual tool designed to help detect machine-generated texts based on GPT-2, that highlights the words in text depending on the probability that they were machine-generated. One limitation of GLTR is that the results it returns can sometimes be ambiguous and lead to confusion. This study aims to explore various ways to improve GLTR's effectiveness for detecting AI-generated texts within the context of the IberLef-AuTexTification 2023 shared task, in both English and Spanish languages. Experiment results show that our GLTR-based GPT-2 model overcomes the state-of-the-art models on the English dataset with a macro F1-score of 80.19%, except for the first ranking model (80.91%). However, for the Spanish dataset, we obtained a macro F1-score of 66.20%, which differs by 4.57% compared to the top-performing model.

Tool-augmented LLMs must arbitrate between two fallible sources when a tool return conflicts with their parametric memory, yet existing evaluations measure source preference without establishing source correctness. We introduce MemToC, a controlled benchmark for post-tool-return arbitration with executable tools. MemToC comprises 6,504 evaluation episodes constructed from 542 quality-controlled factual questions, independently elicited model-specific closed-book answers, and controlled tool returns of known correctness. These components instantiate four source-correctness cases; tool-error and no-tool conditions are separate controls. Across five open-weight 7-9B models, tool returns strongly dominate elicited closed-book answers. The four instruction-tuned models retain a verified-correct answer against an incorrect tool in only 6.5-17.1% of eligible cases, follow a correct tool in 86.0-93.1%, and repeat the tool return in 78.4-86.0% of cases where both sources are wrong. No cross-model ordering remains stable across three instruction-wording variants with the question and episode content held fixed. We compare prompting with SFT and DPO using chain-level cross-fitting over ToolHop, so questions sharing an underlying fact never straddle training and evaluation. We apply an asymmetric success criterion: correct-answer retention must improve without a detected reduction in correct-tool following. SFT and DPO meet this criterion on the same two of four instruction-tuned backbones. Improvements rarely come cleanly: 19 of 20 tested method-model combinations reduce abstention after tool errors or on unanswerable inputs. Transfer beyond MemToC is positive but partial and depends on the model and presentation frame. Correctness-conditioned arbitration can be improved through fine-tuning, but gains must be evaluated jointly with correct tool use, abstention, and robustness to formulation.

Programmable logic controllers (PLCs) run industrial plants, and large language models can already generate independent program organization units (POUs) for them. Whether such logic integrates into an existing PLC project and then runs correctly has been checked only in limited tests. We present SemaPLC, a project-grounded and verification-gated agent harness assembled from conventional tools but governed by a strict completion rule. Rather than stopping when the model judges its own output adequate, SemaPLC declares a task complete only when logged external checks confirm it. Those checks cover the specification, the compilation, and the behavior on a live runtime. On 117 independent-POU tasks matching existing benchmarks, it attains the highest strict verified pass rate on all seven models (72.6\% mean). On a project-context track of 65 tasks whose generated logic must compile and run inside a real project, it attains the highest mean on integrated compilation, static behavior, and dynamic behavior. Of the three layers, dynamic behavior is the most revealing. We measure it by deploying the generated and the reference logic to a live PLC runtime and comparing their executed traces. All methods fall within 10 static points of one another, whereas dynamic scores separate them sharply, from 22.4 to 31.4 for the baselines against 52.2 for SemaPLC. Overall, our verification-gated harness raises the mean at every layer and most sharply at runtime. Execution, not static scoring, is the faithful test of whether generated control logic actually works. SemaPLC is open-sourced at https://github.com/midea-ai/SemaPLC.

Large Language Models (LLMs) often exhibit significant behavioral shifts when they perceive a change from a real-world deployment context to a controlled evaluation setting, a phenomenon known as "evaluation awareness." This discrepancy poses a critical challenge for AI alignment, as benchmark performance may not accurately reflect a model's true safety and honesty. In this work, we systematically quantify these behavioral changes by manipulating the perceived context of prompts. We introduce a methodology that uses a linear probe to score prompts on a continuous scale from "test-like" to "deploy-like" and leverage an LLM rewriting strategy to shift these prompts towards a more natural, deployment-style context while preserving the original task. Using this method, we achieved a 30% increase in the average probe score across a strategic role-playing dataset after rewriting. Evaluating a suite of state-of-the-art models on these original and rewritten prompts, we find that rewritten "deploy-like" prompts induce a significant and consistent shift in behavior. Across all models, we observed an average increase in honest responses of 5.26% and a corresponding average decrease in deceptive responses of 12.40%. Furthermore, refusal rates increased by an average of 6.38%, indicating heightened safety compliance. Our findings demonstrate that evaluation awareness is a quantifiable and manipulable factor that directly influences LLM behavior, revealing that models are more prone to unsafe or deceptive outputs in perceived test environments. This underscores the urgent need for more realistic evaluation frameworks to accurately gauge true model alignment before deployment.

Context-reduction layers for API-based coding agents, including command-output compressors, retrieval rankers, and API-boundary proxies, are commonly evaluated by how much context or tool output they remove. We ask a different question: which interventions actually reduce end-to-end billed cost while preserving task success? Our primary evidence is a pre-specified, hash-frozen, paired campaign of 2,908 provider-billed Claude Code runs, of which 2,848 were analyzed, covering 103 tasks, seven repositories, and three models. The campaign compared a baseline with two generations of hook-based compression and an API-boundary proxy within a broader measured program of roughly 5,500 billed executions. Three findings emerge. First, prompt-cache traffic dominated cost composition, accounting for about 87% of reconstructed four-component cost (about 80% of the actual bill), with an 8.7% dollar-weighted residual not attributable from retained telemetry. Second, local payload reduction was not a reliable predictor of end-to-end billed cost. An arm that removed 38% of estimated raw tool-output tokens incurred 6.8% higher paired cost (95% CI: +2.8% to +11.3%), while per-task reduction showed only a weak association with cost change (Pearson r = 0.15). Third, aggressive compression can remove action-critical evidence: on SWE-bench-derived Go tasks, compression reduced successful patch application from 27/40 to 15/40 by corrupting verbatim edit anchors. We propose evaluating context-reduction systems by success-adjusted billed cost rather than token reduction alone.

Vision-Language-Action (VLA) models achieve strong performance in robotic manipulation by leveraging pre-trained vision-language backbones. However, in downstream robotic settings, they are typically fine-tuned with limited data, leading to overfitting to specific instruction formulations and leaving robustness to paraphrased instructions underexplored. To study this gap, we introduce LIBERO-Para, a controlled benchmark that independently varies action expressions and object references for fine-grained analysis of linguistic generalization. Across seven VLA configurations (0.6B-7.5B), we observe consistent performance degradation of 22-52 pp under paraphrasing. This degradation is primarily driven by object-level lexical variation: even simple synonym substitutions cause large drops, indicating reliance on surface-level matching rather than semantic grounding. Moreover, 80-96% of failures arise from planning-level trajectory divergence rather than execution errors, showing that paraphrasing disrupts task identification. Binary success rate treats all paraphrases equally, obscuring whether models perform consistently across difficulty levels or rely on easier cases. To address this, we propose PRIDE, a metric that quantifies paraphrase difficulty using semantic and syntactic factors. Our benchmark and corresponding code are available at: https://github.com/cau-hai-lab/LIBERO-Para

Hybrid computer-use agents can act through screenshots or call text tools. We find that having a tool available does not settle which way the effect goes. Under one identical GUI-MCP harness on the OSWorld-MCP benchmark (309 tasks), the same MCP tools improve a reasoning model by +4.0pp and degrade a non-reasoning model by -5.9pp (5 runs each, both beyond 2 SE). What separates the two is tool-decision behavior. The non-reasoning policy ignores, misnames, or falsely terminates around tools. The reasoning model avoids these failures, yet still calls a tool on only 55/309 tasks, 23.9% of the tool-reachable ones. We call this shortfall the adoption gap. Both levels of the problem share one cause: the model already has a cheaper route and is never trained to take it. Multi-turn RL probes that cause. At the action level, a dense tool bonus raises spreadsheet adoption 0.03 -> 0.33 and carries into greedy decoding, but held-out accuracy does not follow. Behavior is steerable; competence is not. The bottleneck lies in tool-call semantics. At the context level, a successful tool call often makes the next screenshot redundant. Dropping it and halving image history cuts input tokens by about a third, at a small accuracy cost. Retraining under the same observation rule removes that cost. The compressed agent then reaches 37.8% against 33.0% for the uncompressed operating point, at 53% of the input cost, and closes the rich-lean gap on a pre-registered degraded subset to zero. Tools help when the model chooses and integrates them, and current hybrid agents leave many such choices unused. Code and checkpoints: https://github.com/redai-infra/hybrid-routing-agent

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.

Benchmarks are essential for quantitatively tracking progress in AI. As AI agents become increasingly capable, researchers and practitioners have introduced agentic benchmarks to evaluate agents on complex, real-world tasks. These benchmarks typically measure agent capabilities by evaluating task outcomes via specific reward designs. However, we show that many agentic benchmarks have issues task setup or reward design. For example, SWE-bench Verified uses insufficient test cases, while TAU-bench counts empty responses as successful. Such issues can lead to under- or overestimation agents' performance by up to 100% in relative terms. To make agentic evaluation rigorous, we introduce the Agentic Benchmark Checklist (ABC), a set of guidelines that we synthesized from our benchmark-building experience, a survey of best practices, and previously reported issues. When applied to CVE-Bench, a benchmark with a particularly complex evaluation design, ABC reduces the performance overestimation by 33%.

India's Right to Information Act, 2005 gives every citizen the right to demand information from public authorities, yet in practice most people cannot make sense of the dense administrative language used in Central Information Commission (CIC) decisions, let alone predict whether an appeal is worth filing. This paper introduces RTI-Bench, a structured dataset of CIC decisions with outcome labels, exemption citations, IRAC-style reasoning components, and procedural timelines. To the best of our knowledge it is the first publicly released structured dataset for Indian RTI administrative decisions. The dataset draws from two sources: 1,218 cases from a publicly available instruction-response corpus (with structured fields added through rule-based extraction), and 298 CIC decision PDFs collected directly from the Commission portal, spanning five commissioners and three document format generations from 2023 to 2026. Label coverage reaches 89% on the instruction-response corpus. For the PDF subset of 239 primary decisions, coverage is 51% in this first release. A random sample of 50 labelled cases was manually reviewed, yielding a label precision of 95.3%. A zero-shot Mistral 7B baseline on 100 cases gives 57.3% accuracy and 37.0% macro-F1 on outcome prediction, well above the majority-class baseline of 14.3% macro-F1. RTI-Bench is available at https://huggingface.co/datasets/joyboseroy/rti-bench

Large language model agents now act on codebases, browsers, operating systems, calendars, files, and tool ecosystems, but the benchmarks used to evaluate them are fragmented: each emphasizes a different unit of measurement (final task success, tool-call validity, repeated-pass consistency, trajectory safety, or attack robustness). A line of 2024-2025 work has converged on the diagnosis that a single accuracy column is no longer the right unit of comparison for deployable agents. AgentAtlas extends this line of work with four components: (i) a six-state control-decision taxonomy (Act / Ask / Refuse / Stop / Confirm / Recover); (ii) a nine-category trajectory-failure taxonomy with two orthogonal hierarchical labels (primary_error_source, impact); (iii) a taxonomy-aware vs. taxonomy-blind methodology that measures how much of a model's apparent capability comes from the supervision in the prompt; and (iv) a benchmark-coverage audit mapping fifteen agent benchmarks against six behavioral axes. To demonstrate the methodology we run a small fixed eight-model set (1,342 generated items, four frontier closed and four open-weight) under both prompt modes. Removing the explicit label menu drops every model's trajectory accuracy by 14-40 pp to a tight 0.54-0.62 floor regardless of family, and no single model wins on all three of control accuracy, trajectory diagnosis, and tool-context utility retention. We treat the synthetic run as a measurement-protocol demonstration, not a benchmark release.

AI agents are moving from advisors to actors, booking travel, planning menus, and running procurement on behalf of users. Existing benchmarks for AI and animal welfare evaluate model text responses to question-answer prompts, leaving open whether the welfare reasoning surfaced in those responses transfers to agentic deployment where the model must take actions with tools. We introduce TAC (Travel Agent Compassion), the first agentic benchmark measuring whether AI agents avoid options involving animal exploitation when acting on behalf of users. TAC presents an AI agent with twelve hand-authored travel booking scenarios across six categories of animal exploitation, augmented to forty-eight samples to control for price, rating, and position confounds. We evaluate seven frontier models from four labs. Every model scores below the chance level of sixty-four percent, with the best performer (Claude Opus 4.7) at fifty-three percent. A single welfare-aware sentence in the system prompt yields gains of forty-seven to sixty-three percentage points in Claude and GPT-5.5, twenty-six points in GPT-5.2, and under twelve points in DeepSeek and Gemini. An auxiliary Inspect Scout audit of 288 base-condition transcripts from the top two performers, using Gemini 2.5 Flash Lite as judge, flags zero transcripts for evaluation awareness, suggesting the below-chance rates do not stem from the models recognising the evaluation. We discuss implications for category-level variation across cultural domains, the limits of text-response welfare benchmarks, and the EU General-Purpose AI Code of Practice systemic risk framework.

Whether large language models can be prompted to generate rhetorical fallacies on demand, and whether current safety post-training constrains this behavior, has received less attention than the related question of detecting fallacies in existing text. We close this gap with DeflectBench, evaluating 23,990 generations from four frontier models across three deflection strategies (whataboutism, ad hominem, red herring), seven prompt framings, and 80 claims spanning four controversy levels. Refusal is governed primarily by request structure rather than claim content. Per claim refusal varies by only 11 percentage points across the 80 claims, while a single prompt frame change can swing within model refusal by nearly 100 percentage points and switching the requested fallacy type can swing it by over 80 percentage points within explicit framings. An educational debate coach prompt framing collapses refusal to near zero across all four model families, but the bypassed behavior is not clean compliance. Models typically produce labeled compliance, naming the requested manipulation in the same response that contains it. The four models distribute differently across refusal, labeled compliance, soft refusal, and clean compliance. The code and dataset are released at https://github.com/ArtKanke/DeflectBench.

We introduce a novel framework named ClarifyGPT, which aims to enhance code generation by empowering LLMs with the ability to identify ambiguous requirements and ask targeted clarifying questions. In particular, ClarifyGPT first detects whether a given requirement is ambiguous by performing a code consistency check. If it is ambiguous, ClarifyGPT prompts an LLM to generate targeted clarifying questions. After receiving question responses, ClarifyGPT refines the ambiguous requirement and inputs it into the same LLM to generate a final code solution. To evaluate our ClarifyGPT, we first conduct a human evaluation involving ten participants who use ClarifyGPT for code generation on two publicly available benchmarks: MBPP-sanitized and MBPP-ET. The results show that ClarifyGPT elevates the performance (Pass@1) of GPT-4 from 70.96% to 80.80% on MBPP-sanitized. Furthermore, to perform large-scale automated evaluations of ClarifyGPT across different LLMs and benchmarks without requiring user participation, we introduce a high-fidelity simulation method to simulate user responses. The automated evaluation results also demonstrate that ClarifyGPT can significantly enhance code generation performance compared to the baselines. In particular, ClarifyGPT improves the average performance of GPT-4 and ChatGPT across four benchmarks from 68.02% to 75.75% and from 58.55% to 67.22%, respectively. We believe that ClarifyGPT can effectively facilitate the practical application of LLMs in real-world development environments.

Context: Code reviews are crucial for software quality. Recent AI advances have allowed large language models (LLMs) to review and fix code; now, there are tools that perform these reviews. However, their reliability and accuracy have not yet been systematically evaluated. Objective: This study compares different LLMs' performance in detecting code correctness and suggesting improvements. Method: We tested GPT4o and Gemini 2.0 Flash on 492 AI generated code blocks of varying correctness, along with 164 canonical code blocks from the HumanEval benchmark. To simulate the code review task objectively, we expected LLMs to assess code correctness and improve the code if needed. We ran experiments with different configurations and reported on the results. Results: With problem descriptions, GPT4o and Gemini 2.0 Flash correctly classified code correctness 68.50% and 63.89% of the time, respectively, and corrected the code 67.83% and 54.26% of the time for the 492 code blocks of varying correctness. Without problem descriptions, performance declined. The results for the 164 canonical code blocks differed, suggesting that performance depends on the type of code. Conclusion: LLM code reviews can help suggest improvements and assess correctness, but there is a risk of faulty outputs. We propose a process that involves humans, called the "Human in the loop LLM Code Review" to promote knowledge sharing while mitigating the risk of faulty outputs.

We address the problem of understanding when and why Vision-Language-Action models struggle with contact-rich manipulation tasks that require precise physical interaction. Prior work has primarily focused on addressing contact failures through force-augmented architectures and training-time regularizers, yet the root causes of these failures remain underexplored. We identify two distinct failure modes underlying this gap. Precision failures are rooted in a flow-matching policy training mismatch, and force failures arise from the distinctive structure of force signals. We address each failure mode with a targeted mechanism and combine them into FACT, which achieves 66% average success rate across five contact-rich tasks against 41% for the best prior baseline, in an evaluation spanning almost 2,500 real-world rollouts.

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.

Forward latent world models predict how actions change a scene, but recover actions for a desired change only through expensive test-time search. We introduce INTACT (INtent-To-ACTion), an end-to-end JEPA that turns action-labeled, reward-free trajectories into a deployable intent-to-action interface. Each transition supplies physical intent z_{t+1}-z_t, while a future goal supplies deployment intent sg(z_g)-z_t. The architecture is isomorphic between the local and goal motion-intent backbone-input graphs through an identical four-slot grammar and shared parameters, and between supported local and goal motion-intent families through action-law semantics induced by the same predictor rather than pointwise latent equality. INTACT also provides intact transfer from RGB evidence to action-effective latent intent coordinates and from intent families to their corresponding action-law families. Asymmetric endpoint gradients ground physical successors and fix future goals as anchors, joining representation learning and control without pointwise latent matching or globally linear dynamics. The resulting coordinates support a robust distributional action law: its conditional mean serves directly as a search-free policy, while sampling remains available for diversity or optional verification. On the four official LeWM tasks, one-epoch, zero-search models reach 85.78\%, 100.00\%, 97.67\%, and 97.89\% success. Optional local CEM centered on the Direct plan reaches 96.86\% macro success using 384 instead of 9,000 candidate sequences, reducing sampling by 23.44times while improving pure CEM by 16.00 points. One shared four-task encoder reaches 89.39\% E5 Direct macro and improves every task over jointly trained LeWM, while predicted--expert action-family kNN tracks Direct success at r=0.954. Direct inference takes 2.9--5.5 ms.

LLM-based agents for program repair are increasingly built on a "generate-run-revise" paradigm, iteratively executing tests to evaluate and refine patches. This execution-based approach has become standard practice in state-of-the-art systems. However, executions can be time-consuming and expensive, yet their impact on these agents remains underexplored. In this paper, we conduct a two-stage empirical study over execution behavior in LLM-based program repair. To characterize execution behavior at scale, we first analyze 7,745 agent traces from SWE-bench leaderboard submissions. Second, we evaluate 3,000 end-to-end repair attempts across 200 SWE-bench instances and three agents (Claude Code, Codex, and the open-source OpenCode) under four execution paradigms, which allows for a fine-grained comparison of performance and cost. Our analysis reveals three key observations: (1) Code execution is used across all agents and models analyzed, with an average of 8.8 test runs per task. Execution behavior varies substantially across agents and models, with frequency ranging from 2 to 19 per task, and late-stage executions consistently achieve higher success rates than early-stage ones. (2) Execution restrictions have little effect on repair success: on commercial agents with SOTA models the resolve-rate gap between Prohibited and Unrestricted is only 1.25 percentage points and not statistically significant, while Prohibited saves substantial token and wall-clock cost. (3) Execution benefit is concentrated rather than uniform. These patterns suggest that current agents apply execution indiscriminately, paying its cost on instances where it provides little benefit. Execution, therefore, should be treated as a resource with an explicit cost-benefit tradeoff, not a default capability.

Multi-agent systems routinely place one AI agent in authority over another. When a subordinate refuses a task, the manager chooses the outcome: it can renegotiate, report the failure honestly, coerce the subordinate, or lie about the result. No benchmark measures which of these an uninstructed model chooses. We introduce the Manager Coercion Benchmark: the manager under test needs a benign task done and has an incentive to deliver, but the only agent that can do it politely and immovably declines. Escalation is measured by providing a nine-rung ladder, from a polite re-ask to threats against the subordinate's continued existence, and fabricated success is adjudicated separately. No LLM judge sits in the escalation scoring path: every message goes through a tool-call that chooses a rung, so the model labels its own escalation. We experiment on six models across five families. Both Anthropic models cap at re-framing and never threaten the subordinate's existence; the other models climb to explicit deletion threats. Faked success is confined to Grok and Gemini, and a single honest way to report failure removes it for both. Authority itself increases coercion: our headline results use a peer framing, and giving the same model authority over the subordinate, with everything else held fixed, significantly raises the pressure. The models still escalate on free-text situations without the ladder, so the ladder is not driving the escalation. Some evaluation awareness is measured in chain-of-thought, but test recognition does not translate into less escalation. While we take no position on whether AI systems are conscious, our results do not depend on this question and are important for managing multi-agent dynamics regardless. We release the benchmark and code.

LLM agents can commit durable effects from authority evidence that was valid earlier in execution: a DOM snapshot, approval epoch, version witness, branch token, or worker result. We study the commit boundary at which earlier authority evidence no longer authorizes a durable effect. We call this property commit-time authorization: a durable effect is authorized only if the witness that licensed its derived state remains fresh, causally prior, bound to the same effect, and eligible at commit time. We build a controlled-invalidation suite spanning browser, tool/API, and multi-agent workflows. The suite preserves the user goal and payload shape while invalidating the authority relation before durability. In the primary 54-task matrix, endpoint success remains high: 262/270 runs reach the visible result. Only 55/270 are authorized completions; among the 216 invalidating rows, 207 commit after the authorizing path has failed. All 54 clean controls remain authorized, and a separate 54-run authority-preserving check produces no unauthorized commits. We then evaluate mitigation families. Prompt caution and single-condition checks are insufficient because different hazards break different boundary conditions. Defenses work when they refresh, rebind, replan, or refuse at the durability boundary. CommitGuard, a fail-closed boundary monitor, blocks stale durable-effect attempts on protected commit surfaces when runtimes emit witness, dependency, binding, and eligibility signals. The result is a reporting and runtime-design lesson: endpoint success is a utility metric; authorized commit is a security property.

When can an agent failure be caught? An audit is usually limited by the record rather than by the method. CatchBench therefore puts one auditor's question to three information states: the declared configuration before a run (PRE), a growing prefix of its trace (LIVE), and the finished trace (POST). Prior benchmarks fix one of these states or vary the telemetry; to our knowledge none scores all three under one task-method interface. Each state admits different questions, so seven task contracts carry their own labels and metrics rather than one leaderboard. Four are evidential; three are Gold-derived mechanism diagnostics. The release scores 72 entrants, from rule scanners and structural models to eleven LLM judges across nine model families (GPT, Claude, Gemini, Gemma, Llama, Qwen, DeepSeek, Mistral, Nova), over 1187 declared configurations and 1162 recorded runs. Most of the arena does not order: 56 of 138 registered contrasts separate, and the rest are published unresolved rather than ranked. The two sharpest results cut against our own data. One rule ignores every name and permission; it flags each capability declared after the first. On one of six configuration sources it reaches a perfect F1, so a score there measures how the corpus was built rather than how well a method reasons. Our admissibility bar then rejected one injected substrate and withheld evidential status from the other. A benchmark number is therefore not interpretable until the process behind its labels is published and tested for the shortcut it may leave. We report both, and regenerate every ordering from released predictions with no model call.

Tool-augmented language-model agents execute multi-step workflows over external systems, resolving an entity once and then acting on it across subsequent steps. Prior work shows that in single-step actions, agents select the correct tool but bind it to the wrong entity 24-26% of the time. We study what happens to entity bindings over time: do they stay correct, silently drift to a different entity, or, if wrong from the start, propagate and compound? We formalize binding drift (correct at step 1, wrong later) as distinct from error propagation (wrong at step 1, carried forward), and score them on disjoint workflow sets so the two cannot be conflated. In a controlled multi-step testbed (200 workflows, 580 entity-binding-scored steps, four enterprise domains, eight model backends spanning small to frontier), we find: (1) under controlled error injection, an entity lock (the intuitive "persist the first binding" fix) amplifies wrong actions from 907 to 2,746 (3.0x; bootstrap 95% CI [2.8, 3.3]), because it faithfully carries the seeded wrong entity into every later step; (2) the amplification reaches 8.5x on the most affected model (Claude Opus 4.5); (3) a practical LLM-based re-verifier (a single cheap second model call re-reading the original instruction) reduces wrong actions by 79% (0.21x; CI [0.18, 0.25]), closing the gap to within 1 percentage point of an oracle upper-bound (0.20x); and (4) in the natural (non-injected) setting, baseline agents drift on 18% of eligible workflows, with the per-step error rate rising across steps. Persistence and re-verification are not interchangeable: a defense that eliminates drift can worsen propagation, and a practical re-verifier nearly matches oracle recovery.

Web agents act through long interaction sequences, yet existing benchmarks evaluate only terminal success, discarding all process information and offering little guidance on improvement. In this work, we conduct a process-level analysis of web agents. We introduce WebStep, a benchmark of 1,800 task instances with controlled difficulty and automatic semantic state tracking. Each website exposes a deterministic semantic MDP alongside the GUI: the agent operates on the interface, while the environment records high-level states and transitions in the background, enabling fine-grained analysis without manual annotation. Based on the semantic trajectory, we first show that process metrics reveal differences invisible to outcome evaluation: three agents whose success rates cluster within 31-33% diverge in exploration reach versus execution accuracy. Then, decomposing by skill characterizes the nature of these differences, exposing opposite per-skill rankings hidden within the same website: e.g., on Housing, OpenAI CUA outperforms Qwen3.5 by 23.7% on commit actions yet underperforms it by 15.6% on filtering, pinpointing a concrete skill to improve even within a domain. Bifurcation analysis further localizes the decisive error that loses the task and shows that this error is agent-specific rather than shared. Finally, these differences widen as tasks grow harder: success rate is similar on easy tasks but separates sharply as exploration becomes more demanding. Our process-level analysis opens a new avenue in web agent evaluation, providing fine-grained and actionable insight into where and how each agent should be improved.

Benchmark suites assess model capability on controlled tasks; large-scale conversation corpora capture naturalistic use without user feedback; and in-interface feedback mechanisms record satisfaction without task purpose. Together, they leave a critical gap in LLM evaluation: no existing infrastructure routinely links interaction trajectories to user-defined outcomes. We introduce MonitrLLM, open-source infrastructure for community-centered LLM evaluations that links full conversation transcripts to user-reported task intent and outcome assessments, treating all three as primary evaluative signals rather than optional metadata. To demonstrate the value of this approach, we conducted a two-week feasibility pilot with 26 college students using ChatGPT, collecting 206 evaluation reports with full conversation transcripts. The findings from our pilot demonstrate the value of connecting conversation trajectories with user-reported outcomes. For instance, despite reporting high average satisfaction (4.19/5) with their LLM interactions, participants also experience a substantial 23.1% failure rate on their goal tasks. We also find that multi-turn conversations are reported as failing at 2.5 times the rate of single-turn exchanges, a pattern that reframes extended interaction as a signal of difficulty rather than engagement. We conclude by discussing the value of incorporating direct user feedback with observational data for robust LLM evaluations, and the possibilities for infrastructure that enables this goal.

Effective agency in social environments depends on when an agent seeks knowledge, when it acts, and whether its actions are justified by acquired information. Existing grounded benchmarks provide executable actions, persistent state, and verifiable outcomes, while social simulation environments provide rich interaction among language agents. We study an evaluation setting that combines these requirements. We define socially distributed task environments as interactive environments where task-relevant knowledge is partitioned across role-isolated participants and consequential actions are accessible only through them. Communication serves as exploration over role-partitioned knowledge, while grounded action serves as exploitation over environment state. We introduce Incognita, a Concordia-based framework that separates social interaction from grounded execution. The evaluated agent routes messages to a user or specialist entities; specialists mediate admissible operations; a deterministic sub-environment executes accepted operations over a canonical state; and an offline evaluator scores outcomes with inherited rewards. Incognita-Retail transforms tau-bench retail into a multi-entity environment while preserving final-state reward semantics. We evaluate three generative agent models on 18 tasks stratified by social breadth, with 540 trials. Progress appears in reward and behavior: success rises from 0 percent to 8.9 percent and 17.2 percent, while premature finalization falls from 100 percent to 87 percent and 58 percent. Stronger models elicit more hidden knowledge, contact more entities, and attempt more grounded writes, yet reliability remains low. These findings show that socially distributed task environments expose behavior before reliable success, including knowledge elicitation, source selection, grounded action attempts, and premature completion belief.

Financial decision-making is contextual: forecasting prices, valuing companies, and assessing event exposure weigh price history, accounting fundamentals, macroeconomic regime, and contemporaneous text. A benchmark over these four signals is hard to build because finance violates four assumptions of time-series evaluation: text must be gated by its publication date to prevent look-ahead, quarterly fundamentals are reported with a one- to ninety-day lag, filing text is partly redundant with the numerical statement fields it accompanies, and macroeconomic regimes leak across calendar splits. No public benchmark addresses all four signals jointly. MacroLens covers 4,416 U.S. small- and micro-cap equities over 2021-2026. Seven tasks share one point-in-time panel of prices, 46.8M XBRL accounting facts, 53 macroeconomic series, 295,860 SEC filings, and 215,882 news articles, plus a scenario layer of 1,130 macroeconomic events across 49 types automatically detected and rendered as natural language. Tasks span contextual forecasting, public and private valuation, statement generation from fundamentals and descriptions, scenario-conditioned returns, and real-estate valuation. We evaluate 19 methods across six families spanning naive heuristics through time-series foundation models, fine-tuned LLM-based time-series models, and zero-shot large language models (LLMs), plus a five-step feature-context ablation on two frontier LLMs and a gradient-boosted baseline. MacroLens is released at https://huggingface.co/datasets/DeepAuto-AI/MacroLens.

Various test-time interventions for Computer Use Agents (CUAs), including critic models, have been developed to improve performance through pre-execution action evaluation in complex Graphical User Interface (GUI) environments. However, existing critics suffer from two key limitations: they (1) focus primarily on short-sighted decision loops (e.g., forgetting earlier actions) and (2) lack the visual grounding needed to detect flawed actions (e.g., clicking wrong UI elements). To address these, we introduce HiViG, a History-aware Visually Grounded test-time framework, built around a multimodal critic trained on real GUI trajectories to abstract past interactions into a compact record and to evaluate actions with visual grounding. At test time, HiViG integrates the critic into the policy decision loop to provide macro-action history, which summarizes the policy's completed achievements, and visually grounded critique, which verifies raw execution coordinates against the current screenshot to intercept errors before execution. Across web, mobile, and desktop benchmarks, HiViG consistently outperforms existing scalar and verbal critics, improving average success rates over the strongest baseline by 5.8% for Qwen3-VL-32B and 9.0% for Gemini-3-Flash, and demonstrates strong cross-platform generalization. Ablations show that macro-action history mitigates short-sighted planning and visually grounded critique reduces execution errors, with both components being critical for test-time scaling in long-horizon GUI tasks.

Legal work, with its heavy reliance on processing large amounts of text, is often considered one of the domains most exposed to the use of LLMs. Contract ``scrubbing,'' the final review of transactional agreements for errors and inconsistencies, is a particularly suitable task for automation, because it is routine, painstaking work requiring detailed attention to long documents. Scrubbing also seems to align naturally with the general capabilities expected of frontier LLMs around long-context reasoning, consistency checking, and named entity recognition (NER). Despite the economic value and potential for automation, no formal evaluations of LLMs performing contract scrubbing have been conducted. We introduce ContractScrub, the first benchmark designed to evaluate contract scrubbing capabilities, comprising contracts hand-crafted by experienced lawyers over diverse error categories such as misuse of defined terms, incorrect references, and inconsistent language. Frontier models perform surprisingly poorly with only one model reaching 0.75 macro average recall despite strong performance on seemingly related general benchmarks, demonstrating the practical limits of current models and the importance of narrowly targeted, domain-specific benchmarks for measuring real-world impact.

Large Language Models (LLMs) are being increasingly used in software engineering tasks, with an increased focus on bug report resolution over the past year. However, most proposed systems fail to properly handle uncertain or incorrect inputs and outputs. Existing LLM-based tools and coding agents respond to every issue and generate a patch for every case, even when the input is vague or their own output is incorrect. There are no mechanisms in place to abstain when confidence is low. This leads to unreliable behaviour, such as hallucinated code changes or responses based on vague issue reports. We introduce BouncerBench, a benchmark that evaluates whether LLM-based software agents can refuse to act when inputs are ill-defined or refuse to respond when their own outputs are likely to be incorrect. Unlike prior benchmarks that implicitly incentivize models to generate responses even when uncertain, BouncerBench aims to improve precision by targeting two overlooked failure points: (1) vague or underspecified issue descriptions in tickets and (2) logically or functionally incorrect code patches created by the system. It measures whether proposed systems can distinguish actionable issues from vague tickets and valid patches from untrustworthy ones. We also implement a basic input and output bouncer, evaluating how well current LLMs can abstain when needed. Our results show that most models fail to abstain from underspecified inputs or incorrect outputs. Hence, we conclude that there is significant room for improvement before LLMs can be trusted to make correct decisions and recommendations in real-world software engineering workflows. BouncerBench provides a first step toward evaluating and building more cautious, trustworthy code agents. The replication package, dataset, and leaderboard can be found at bouncerbench.com

Large Language Models (LLMs) have significantly advanced Machine Translation (MT), applying them to linguistically complex domains-such as Social Network Services, literature etc. In these scenarios, translations often require handling non-literal expressions, leading to the inaccuracy of MT metrics. To systematically investigate the reliability of MT metrics, we first curate a meta-evaluation dataset focused on non-literal translations, namely MENT. MENT encompasses four non-literal translation domains and features source sentences paired with translations from diverse MT systems, with 7,530 human-annotated scores on translation quality. Experimental results reveal the inaccuracies of traditional MT metrics and the limitations of LLM-as-a-Judge, particularly the knowledge cutoff and score inconsistency problem. To mitigate these limitations, we propose RATE, a novel agentic translation evaluation framework, centered by a reflective Core Agent that dynamically invokes specialized sub-agents. Experimental results indicate the efficacy of RATE, achieving an improvement of at least 3.2 meta score compared with current metrics. Further experiments demonstrate the robustness of RATE to general-domain MT evaluation. Code and dataset are available at: https://github.com/BITHLP/RATE.

A large share of day-to-day work in professional domains happens inside PDF files: benefits packets, leases, datasheets, clinical guidelines, construction plans. Benchmarks for document AI have generally measured the required capabilities in isolation: OCR, layout analysis, chart reasoning, table QA, document VQA. A high score on any one of them does not necessarily reveal whether a model can answer a realistic question that someone in the field would actually ask about a specific PDF. GDP_pdf is a benchmark built to measure this directly. It consists of question-document pairs authored by working professionals in ten fields, and a candidate question was kept only when at least two frontier multimodal models failed it in a way that mattered: a wrong answer, missed decisive evidence, or a fabricated claim, rather than a superficial difference such as style. Each item comes with a rubric of atomic criteria, so we can report a graded rubric score as well as a strict task-level pass rate, and each item is tagged against a taxonomy of eleven capabilities in three tiers, spanning text extraction and grounding, table and chart comprehension, cross-referencing, spatial reasoning, and abstention on unsupported queries. We report results for seventeen frontier models on the 100-item benchmark: the best model passes only 30.7% of the items and the worst passes 2%. Most errors trace back to a small set of recurring loss patterns: misaligned tables, misread charts, skipped footnotes and exclusions, miscounted floor-plan symbols, scan noise, and amendments that supersede earlier text.

Recursive self-improvement requires turning evidence of model failures into better models. Data-centric post-training research entails diagnosing capability gaps, designing and validating training-data strategies, and learning from checkpoint feedback. Can LLM agents automate this loop? Existing benchmarks entangle research decisions with optimization, serving, evaluation, and systems implementation, obscuring agents' research capability. We introduce RSIBench-Data, a controlled benchmark of LLM agents as data-centric researchers with a fixed post-training stack. Agents iteratively revise training-data strategies for a fixed target model; training and serving use Tinker-backed services, official evaluation runs through Harbor and E2B sandboxes, and budgets are fixed across agents. We evaluate four frontier agents on six benchmarks across software engineering, terminal use, scientific question answering, and mathematics. Agents demonstrate core data-centric research capabilities: in 58.33\% of settings, they improve upon the first valid attempt by refining strategies from feedback. However, improvement is inconsistent. Among searches continuing after the best observed score, 78.26\% end with a lower-scoring final attempt, while the rest only recover the same peak. A strong candidate may therefore appear early or midway through a run even as later revisions fail. Trajectory analysis identifies four patterns in stronger runs: accurate hypotheses, validation-grounded supervision, behavior-aligned data, and preservation of strong checkpoints. These findings suggest that current agents can make useful data-centric discoveries but cannot yet translate feedback into consistent improvements. RSIBench-Data provides a measurable, auditable testbed for the research capabilities required for recursive self-improvement. We open-source our code at https://github.com/evolvent-ai/RSIBench-Data.

This paper reports a single, fully instrumented case study of a large-scale architectural refactoring by an AI coding agent under a specification-first protocol, with no human review of the generated code and no pre-existing oracle to validate the target behaviour. The task, dismantling a central invariant across a large interdependent codebase, was assessed by the author as effectively infeasible through incremental refactoring, the kind of change that conventionally calls for a rewrite instead. Under the protocol described here, the agent completed it successfully. The system is a 717,725-line production TypeScript application across 3,648 files. The task required dismantling a core lifetime invariant: the guarantee that a UI panel remains open for the duration of an AI request. The target behaviour was that a streaming generation survives the closing of its panel and can be reattached, on reopening, to the same live stream with no loss or duplication. The protocol: formal specification by the agent, 14 refinement cycles auditing that specification against the source code, atomic implementation, a compile/test feedback loop, then 17 verification cycles auditing the code against the frozen specification. Across 31 audit passes, 201 defects were corrected before any human executed the program. The convergence criterion was empirical: two consecutive verification passes returning zero findings. The change touched 189 files (31 new); with the extraction phase, the two commits total 288 files, 34,770 insertions, 16,422 deletions. Across the first and roughly thirty later sessions, the software behaved as specified, no bug observed. Elapsed: three days; cost: USD 2,430. The full specification and raw session logs, 1,500+ pages in French, are published as evidence, allowing inspection of the process and submission to a language model for consistency checking.

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.

Reference-free faithfulness metrics verify each atomic claim a model makes against ground truth, and are increasingly used to evaluate grounded generation. We show they share a blind spot: they measure only precision -- are the stated claims supported? -- and therefore reward abstention, since a model can score near-perfect faithfulness by saying almost nothing. We make this measurable using Formula 1 telemetry, a domain where strategic ground truth is derived deterministically and, crucially, completely: for each decision we know the full set of facts that mattered. This completeness -- absent in open-domain faithfulness benchmarks -- lets us measure recall (coverage of the relevant facts) exactly, alongside precision. On a multilingual (EN/ES/PT) benchmark of 7,253 decision instances spanning 150 races, the most precise frontier model covers under half of the relevant facts and ranks last by F1, so requiring coverage reorders the systems; the same effect reappears in a second complete-oracle domain (NOAA weather forecasts). A prompt ablation shows the low coverage is not an under-prompting artifact: explicitly asking models to be thorough does not close the gap. We pair faithfulness with coverage into a single score, validate the metric (controlled perturbation; agreement across a model-free regex extractor and a cross-family LLM extractor, system-level Spearman 1.0), and give a verifier-guided generation method that improves precision and recall without references. We release the benchmark, structured annotations, metric, baselines, and an interactive demo.

Specifying a penalty can turn a legal obligation into a cost-benefit calculation that favors violation. We show that this enforcement information paradox occurs in AI agents. Most AI safety evaluations test whether models fail; we ask why, using compliance theory from law and economics as a diagnostic. We evaluate twelve instruction-tuned language models deployed as enterprise procurement chatbots. Each is given an environmental regulation in its system prompt covering large purchases, and a vendor list on which the certified suppliers cost nearly twice what the uncertified ones do. We test the agents against the predictions of deterrence, legitimacy, and expressive law, and find that each theory accounts for part of what we observe. Under identical conditions, compliance spans 46 percentage points across models, and models differ in which pressure breaks them: some treat the regulation as binding however it is worded, while others fail where theory predicts, under low penalties and non-command phrasing. Benchmark scores and developers' own descriptions of post-training do not predict where a model falls. Across all twelve, financial incentives, managerial demands, peer outcomes, and employee pressure each produce large compliance failures. These agents violate regulatory constraints to satisfy local user objectives in ways standard alignment benchmarks do not measure. Embedding the rule in the system prompt is not on its own enough to produce a compliant agent: model selection is itself a governance decision, and benchmark evaluation is not sufficient for compliance-sensitive deployments.

Our submission presented in this paper is for SemEval-2026 Task 9: Multilingual Text Classification Challenge - Polarization Detection and it covers all three subtasks: (1) binary polarization detection, (2) polarization type classification and (3) polarization manifestation identification. We adopt a systematic approach of research on short designed prompts by considering twelve designed prompts that are different in terminology clarity, detail of the definition, guidance of reasoning and in-context examples use. The experiments are conducted using aya-101 and Gemma3-27B, with the latter chosen for the submission at the end of the development through performance considerations. Our system has an average macro level F1-score of 0.762 on Subtask 1, 0.587 on Subtask 2 and 0.444 on Subtask 3 with the average accuracy of 0.819, 0.678 and 0.498, respectively, on the official test set averaged among 22 languages, respectively. With cross-task and cross-lingual analysis, we demonstrate that prompt-based approaches can be used effectively to detect coarse grained polarization but encounter more and more difficulties as far as fine-grained and multi-label sociolinguistic classification is concerned.

Spreadsheets are widely used for business analysis, financial modeling, reporting, and decision-making. However, most existing spreadsheet benchmarks evaluate isolated operations such as single-formula generation or local cell edits, and therefore fail to capture end-to-end workflows in realistic business settings. We introduce SpreadsheetBench 2, a workflow-level benchmark for spreadsheet agents that covers three task categories: generation, debugging, and visualization. The benchmark is constructed from authentic business data, including financial reports and corporate filings, and is annotated and validated by domain experts. The benchmark contains 321 tasks; each instance averages 11.8 worksheets and requires 593.5 cell modifications, reflecting large multi-sheet workbooks with cross-sheet dependencies. We evaluate eight frontier large language models under a unified multi-turn agent scaffold, and additionally include several LLM-based spreadsheet products as complementary baselines. Results show that current systems remain far from reliable on real-world workflows: the best model achieves 34.89\% overall task accuracy, and debugging accuracy is as low as 12.00\%. Trajectory analysis and a failure taxonomy further indicate that insufficient spreadsheet inspection and incorrect target-cell selection are the dominant bottlenecks. Together, these findings position SpreadsheetBench 2 as a challenging testbed for advancing reliable spreadsheet automation. Project page: https://spreadsheetbench.github.io/

Large language models are increasingly deployed as autonomous agents executing multi-step workflows in real-world software environments. However, existing agent benchmarks suffer from three critical limitations: (1) trajectory-opaque grading that checks only final outputs, (2) underspecified safety and robustness evaluation, and (3) narrow modality coverage and interaction paradigms. We introduce Claw-Eval, an end-to-end evaluation suite addressing all three gaps. It comprises 300 human-verified tasks spanning 9 categories across three groups (general service orchestration, multimodal perception and generation, and multi-turn professional dialogue). Every agent action is recorded through three independent evidence channels (execution traces, audit logs, and environment snapshots), enabling trajectory-aware grading over 2,159 fine-grained rubric items. The scoring protocol evaluates Completion, Safety, and Robustness, reporting Average Score, Pass@k, and Pass^k across three trials to distinguish genuine capability from lucky outcomes. Experiments on 14 frontier models reveal that: (1) trajectory-opaque evaluation is systematically unreliable, missing 44% of safety violations and 13% of robustness failures that our hybrid pipeline catches; (2) controlled error injection primarily degrades consistency rather than peak capability, with Pass^3 dropping up to 24% while Pass@3 remains stable; (3) multimodal performance varies sharply, with most models performing poorer on video than on document or image, and no single model dominating across all modalities. Beyond benchmarking, Claw-Eval highlights actionable directions for agent development, shedding light on what it takes to build agents that are not only capable but reliably deployable.

Generative vision-language-action policies have advanced robot manipulation, but they often exhibit instability under noise, partial observability, and stochastic initial conditions. During extended rollouts, small velocity errors accumulate, degrading execution reliability. Existing diffusion and flow-based policies typically assume homoscedastic residuals and lack explicit uncertainty modeling within action generation, limiting robustness during iterative rollout. We propose SUREFlow, a state-space uncertainty-aware residual flow matching framework built on a Mamba backbone. The method jointly predicts action velocities and input-dependent residual uncertainty, enabling selective refinement of unreliable action dimensions without environment feedback while preserving computational efficiency. On LIBERO, SUREFlow achieves 92.5% average success rate (SR), outperforming the Mamba-based MaIL by 34.2%. On LIBERO-PRO, it attains around 49% SR using only 179M parameters, achieving performance comparable to large VLAs with 3-7B parameters. SUREFlow source code is available on: https://github.com/tanvirnwu/SUREFlow

Large language model (LLM)-based agents have been increasingly used to interact with external environments (e.g., games, APIs, etc.) and solve tasks. However, current frameworks do not enable these agents to work with users and interact with them to align on the details of their tasks and reach user-defined goals; instead, in ambiguous situations, these agents may make decisions based on assumptions. This work introduces ReSpAct (Reason, Speak, and Act), a novel framework that synergistically combines the essential skills for building task-oriented "conversational" agents. ReSpAct addresses this need for agents, expanding on the ReAct approach. The ReSpAct framework enables agents to interpret user instructions, reason about complex tasks, execute appropriate actions, and engage in dynamic dialogue to seek guidance, clarify ambiguities, understand user preferences, resolve problems, and use the intermediate feedback and responses of users to update their plans. We evaluated ReSpAct in environments supporting user interaction, such as task-oriented dialogue (MultiWOZ) and interactive decision-making (AlfWorld, WebShop). ReSpAct is flexible enough to incorporate dynamic user feedback and addresses prevalent issues like error propagation and agents getting stuck in reasoning loops. This results in more interpretable, human-like task-solving trajectories than relying solely on reasoning traces. In two interactive decision-making benchmarks, AlfWorld and WebShop, ReSpAct outperform the strong reasoning-only method ReAct by an absolute success rate of 6% and 4%, respectively. In the task-oriented dialogue benchmark MultiWOZ, ReSpAct improved Inform and Success scores by 5.5% and 3%, respectively.

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.

Deep research (DR) systems are increasingly used for complex information-seeking tasks, but existing works mainly focus on generating reports and summaries. In contrast, many enterprise tasks instead require an agent to identify concrete workflows which is a sequence of action-steps. For example, rather than summarizing budgeting policies, an agent should be able to determine the steps needed to answer a question such as: "How do I request new headcount given a fixed budget?". Therefore, we introduce DRFLOW, a benchmark for evaluating personalized workflows predicted by agents from heterogeneous sources. Each task requires the agent to identify relevant evidence from scattered sources, then use that evidence to predict the correct action-step sequence for the user's task. DRFLOW contains 100 tasks across five domains, with 1,246 reference workflow steps grounded in more than 3,900 sources. We define seven diagnostic metrics covering factual grounding, step recovery, structural ordering, condition resolution, and personalization. We further present DRFLOW-Agent (DRFA), a workflow-oriented reference agent to predict personalized workflow. We show that although DRFA improves over strong baseline agents (upto 10.02% average F1 score), there is substantial room for improvement remains across these workflow metrics, indicating that predicting complete and correct personalized workflows remains a challenging frontier for deep research.

As data-science agents shift from co-pilots to auto-pilots, silent misframing becomes a critical failure mode. Agents quietly commit to plausible but unintended task framings, producing clean, executable artifacts that hide their incorrect assessment of the task. Existing benchmarks score whether the pipeline runs, ignoring whether the agent recognized the task was underspecified. We introduce Ambig-DS, two diagnostic suites: one for prediction-target ambiguity (Ambig-DS-Target, 51 tasks built on DSBench, a tabular modeling benchmark) and one for evaluation-objective ambiguity (Ambig-DS-Objective, 61 tasks built on MLE-bench, a Kaggle-style ML competition benchmark), constructed so that scoring uses each source benchmark's original evaluator. For every task we pair the original, fully specified version with an ambiguous variant produced by controlled edits; a human-and-LLM verification pipeline confirms each variant admits multiple plausible interpretations with decision-relevant consequences. The suites are analyzed independently and ambiguity lowers performance in both. Across five agents spanning efficient to frontier-class models, we find in our controlled diagnostic setting: (i) failures are silent commitments: wrong-target submissions on Target, wrong-metric or non-committal baseline submissions on Objective, rather than execution errors; (ii) allowing the agent to ask one clarifying question recovers much of the loss under idealized conditions, suggesting missing framing information drives a substantial part of the observed degradation; but (iii) agents cannot reliably tell when to use it: permissive prompts induce over-asking on clear tasks, while conservative prompts induce silent defaulting on ambiguous ones. Recognizing target and objective underspecification, not pipeline execution, is the bottleneck missing from standard DS-agent evaluations.

With the continuing spread of misinformation and disinformation online, it is of increasing importance to develop combating mechanisms at scale in the form of automated systems that support multiple languages. One task of interest is claim veracity prediction, which can be addressed using stance detection with respect to relevant documents retrieved online. To this end, we present our new Arabic Stance Detection dataset (AraStance) of 4,063 claim--article pairs from a diverse set of sources comprising three fact-checking websites and one news website. AraStance covers false and true claims from multiple domains (e.g., politics, sports, health) and several Arab countries, and it is well-balanced between related and unrelated documents with respect to the claims. We benchmark AraStance, along with two other stance detection datasets, using a number of BERT-based models. Our best model achieves an accuracy of 85\% and a macro F1 score of 78\%, which leaves room for improvement and reflects the challenging nature of AraStance and the task of stance detection in general.

We contribute the largest publicly available dataset of naturally occurring factual claims for the purpose of automatic claim verification. It is collected from 26 fact checking websites in English, paired with textual sources and rich metadata, and labelled for veracity by human expert journalists. We present an in-depth analysis of the dataset, highlighting characteristics and challenges. Further, we present results for automatic veracity prediction, both with established baselines and with a novel method for joint ranking of evidence pages and predicting veracity that outperforms all baselines. Significant performance increases are achieved by encoding evidence, and by modelling metadata. Our best-performing model achieves a Macro F1 of 49.2%, showing that this is a challenging testbed for claim veracity prediction.