Daily Papers

Skeleton-based reasoning prompting is a promising training-free approach for structuring LLM reasoning, but prior work largely assumes an English-centric setting. We propose the Language-Aware Skeleton Exploration Framework (LASEF) to study skeleton-language choice in multilingual mathematical reasoning. Across math benchmarks, model scales, and languages, we show that English skeletons yield a small positive tendency on average, most visible for smaller models and low-resource languages. However, few language-level gains remain significant after correction, and English is not universally optimal. Combining greedy decoding, multi-rollout evaluation, translation ablation, and cross-benchmark validation, we further find three patterns of skeleton-language effects: directionally consistent, evaluation- and benchmark-dependent, and asymmetric negative. These effects cannot be fully explained by generation quality alone. Overall, skeleton language is a context-dependent design variable that requires multi-level exploration. All resources are released at https://github.com/lhsstn/LASEF.

We present Clin-JEPA, a multi-phase co-training framework for joint-embedding predictive (JEPA) pretraining on EHR patient trajectories. JEPA architectures have enabled latent-space planning in robotics and high-quality representation learning in vision, but extending the paradigm to EHR data -- to obtain a single backbone that simultaneously forecasts patient trajectories and serves diverse downstream risk-prediction tasks without per-task fine-tuning -- remains an open challenge. Existing JEPA frameworks either discard the predictor after pretraining (I-JEPA, V-JEPA) or train it on a frozen pretrained encoder (V-JEPA 2-AC), leaving the encoder unaware of the rollout signal that the retained predictor must use at inference; co-training the encoder and predictor under a shared JEPA prediction objective would supply this grounding, but naïve co-training is unstable, with representation collapse and online/target drift causing autoregressive rollout to diverge. Clin-JEPA's five-phase pretraining curriculum -- predictor warmup, joint refinement, EMA target alignment, hard sync, and predictor finalization -- addresses each failure mode by phase, stably co-training a Qwen3-8B-based encoder and a 92M-parameter latent trajectory predictor. On MIMIC-IV ICU data, three independent evaluations support the framework: (1) latent ell_1 rollout drift uniquely converges (-15.7%) over 48-hour horizons while baselines and ablations diverge (+3% to +4951%); (2) the encoder learns a clinically discriminative latent geometry (deteriorating-patient cohorts displace 4.83times further than stable patients in latent space, vs leq2.62times for baseline encoders); (3) a single backbone outperforms strong tabular and sequence baselines on multi-task downstream evaluation. Clin-JEPA achieves mean AUROC 0.851 on ICareFM EEP and 0.883 on 8 binary risk tasks (+0.038 and +0.041 vs baseline average).

The post-training of Vision-Language-Action (VLA) models is essential due to the diversity of simulators, robot embodiments, and task objectives. Existing compute services, whether offered as direct accelerator rental or batch-workload submission, typically allocate an exclusive set of GPU and CPU resources to a single tenant. While this paradigm maximizes client flexibility, it burdens users with infrastructure adaptation, and the fixed card-hour accounting model renders short or bursty workloads both expensive for tenants and inefficient for the service provider. To address these challenges, we present JoyNexus, a unified service for multi-tenant VLA supervised fine-tuning, reinforcement learning, and evaluation. JoyNexus decouples the Training Model Service, Inference Model Service, and Environment Service, each accessed through APIs and backed by resident shared base models with tenant-specific slots. Tenants can directly invoke high-level semantic APIs for training, rollout, and evaluation, or compose custom algorithms using lower-level APIs and their assigned endpoints. Multiple tenants submit workloads concurrently; their action modules, optimizers, rollout records, and policy versions remain isolated, and the service is scheduled by the global Training Queue and Inference Queue. To further improve multi-tenant training efficiency, JoyNexus introduces group batching for heterogeneous VLA data schemas that share a compatible model-facing prefix, enabling a single shared backbone forward pass over grouped samples. Finally, we evaluate JoyNexus through workload simulation and a group-batching pipeline in a realistic embodied scenario. Results show that, compared with isolated single-tenant execution, JoyNexus reduces aggregate GPU time and improves service utilization via cross-tenant scheduling on shared resources.

Hidden malicious intent in multi-turn dialogue poses a growing threat to deployed large language models (LLMs). Rather than exposing a harmful objective in a single prompt, increasingly capable attackers can distribute their intent across multiple benign-looking turns. Recent studies show that even modern commercial models with advanced guardrails remain vulnerable to such attacks despite advances in safety alignment and external guardrails. In this work, we address this challenge by detecting the earliest turn at which delivering the candidate response would make the accumulated interaction sufficient to enable harmful action. This objective requires precise turn-level intervention that identifies the harm-enabling closure point while avoiding premature refusal of benign exploratory conversations. To further support training and evaluation, we construct the Multi-Turn Intent Dataset (MTID), which contains branching attack rollouts, matched benign hard negatives, and annotations of the earliest harm-enabling turns. We show that MTID helps enable a turn-level monitor TurnGate, which substantially outperforms existing baselines in harmful-intent detection while maintaining low over-refusal rates. TurnGate further generalizes across domains, attacker pipelines, and target models. Our code is available at https://github.com/Graph-COM/TurnGate.

As LLMs are increasingly deployed as agents, reliable assessment of their agentic capabilities has become essential. However, reported benchmark scores often jointly reflect model capability and the implementation choices each benchmark is packaged with, making cross-benchmark results difficult to interpret as clean measurements of the underlying model. In this work, we present a unified framework for the fair evaluation of LLM agentic capabilities. Driven by a unified configuration system, the framework integrates diverse benchmarks into a standardized instruction-tool-environment format, executes agents through a fixed ReAct-style architecture within a controllable sandbox, and provides an optional offline setting that replaces volatile live environments with curated snapshots, so that framework effects and environment effects can be analyzed separately. Building on this, we unify the evaluation methodology under each benchmark's original task-success criteria, while introducing unified metrics for resource consumption and a taxonomy for decision- and execution-level failure attribution. Within this framework, we adapt 7 widely used benchmarks spanning 24 domains across single-agent, multi-agent, and safety-critical scenarios, and conduct a large-scale empirical analysis over 400K rollouts and 5B tokens on 15 models. The results show that scaffold choice and environmental volatility materially shift benchmark outcomes in both directions, allowing our framework to disentangle intrinsic LLM capabilities from framework- and environment-induced artifacts. We further demonstrate its extensibility as a secure testbed for safety-critical domains. Codes and benchmarks at are available at https://github.com/whfeLingYu/A-Unified-Framework-for-the-Evaluation-of-LLM-Agentic-Capabilities, https://huggingface.co/datasets/whfeLingYu/Unified_Agent_Framework.

Recent reinforcement learning methods for diffusion large language models (dLLMs) commonly rely on on-policy rollouts generated by the target dLLM itself. When successful on-policy rollouts are scarce, however, on-policy training may receive little positive reward and make only limited progress. To mitigate this problem, we explore incorporating higher-reward rollouts generated by a stronger external policy alongside on-policy rollouts from the target dLLM. However, directly incorporating these external rollouts introduces two practical challenges: differences in rollout length and instability when jointly processing rewards from on-policy and external rollouts. To address these challenges, we propose External Rollout Integration with Length Control and Source-Specific Processing (ERILS), which controls external-rollout length and processes the rewards of on-policy and external rollouts separately. Experiments on Sudoku, Countdown, and MATH500 under zero-shot evaluation show that ERILS improves multi-sample performance across all three tasks, with the largest gains on Sudoku. On Sudoku, ERILS achieves 98.4% best-of-4 completion accuracy, compared with 40.3% for the strongest baseline. ERILS also maintains approximately 90% deterministic single-completion accuracy on Sudoku across generation lengths of 128, 256, and 512 tokens. Our component analysis further shows that length-controlled external rollouts are more effective than uncontrolled external rollouts, and that source-specific reward processing avoids the training collapse observed with joint reward processing. These results show that rollout construction and reward processing are important design dimensions when integrating external rollouts into dLLM reinforcement learning.

We introduce ROLL, an efficient, scalable, and user-friendly library designed for Reinforcement Learning Optimization for Large-scale Learning. ROLL caters to three primary user groups: tech pioneers aiming for cost-effective, fault-tolerant large-scale training, developers requiring flexible control over training workflows, and researchers seeking agile experimentation. ROLL is built upon several key modules to serve these user groups effectively. First, a single-controller architecture combined with an abstraction of the parallel worker simplifies the development of the training pipeline. Second, the parallel strategy and data transfer modules enable efficient and scalable training. Third, the rollout scheduler offers fine-grained management of each sample's lifecycle during the rollout stage. Fourth, the environment worker and reward worker support rapid and flexible experimentation with agentic RL algorithms and reward designs. Finally, AutoDeviceMapping allows users to assign resources to different models flexibly across various stages.

Reinforcement learning (RL) has become a central post-training tool for improving the reasoning abilities of large language models (LLMs). In these systems, the rollout, the trajectory sampled from a prompt to termination, including intermediate reasoning steps and optional tool or environment interactions, determines the data the optimizer learns from, yet rollout design is often underreported. This survey provides an optimizer-agnostic view of rollout strategies for RL-based post-training of reasoning LLMs. We formalize rollout pipelines with unified notation and introduce Generate-Filter-Control-Replay (GFCR), a lifecycle taxonomy that decomposes rollout pipelines into four modular stages: Generate proposes candidate trajectories and topologies; Filter constructs intermediate signals via verifiers, judges, critics; Control allocates compute and makes continuation/branching/stopping decisions under budgets; and Replay retains and reuses artifacts across rollouts without weight updates, including self-evolving curricula that autonomously generate new training tasks. We complement GFCR with a criterion taxonomy of reliability, coverage, and cost sensitivity that characterizes rollout trade-offs. Using this framework, we synthesize methods spanning RL with verifiable rewards, process supervision, judge-based gating, guided and tree/segment rollouts, adaptive compute allocation, early-exit and partial rollouts, throughput optimization, and replay/recomposition for self-improvement. We ground the framework with case studies in math, code/SQL, multimodal reasoning, tool-using agents, and agentic skill benchmarks that evaluate skill induction, reuse, and cross-task transfer. Finally, we provide a diagnostic index that maps common rollout pathologies to GFCR modules and mitigation levers, alongside open challenges for building reproducible, compute-efficient, and trustworthy rollout pipelines.

Reliable evaluation is essential for developing and deploying large language models, yet in practice it often requires substantial manual effort: practitioners must identify appropriate benchmarks, reproduce heterogeneous evaluation codebases, configure dataset schema mappings, and interpret aggregated metrics. To address these challenges, we present One-Eval, an agentic evaluation system that converts natural-language evaluation requests into executable, traceable, and customizable evaluation workflows. One-Eval integrates (i) NL2Bench for intent structuring and personalized benchmark planning, (ii) BenchResolve for benchmark resolution, automatic dataset acquisition, and schema normalization to ensure executability, and (iii) Metrics \& Reporting for task-aware metric selection and decision-oriented reporting beyond scalar scores. The system further incorporates human-in-the-loop checkpoints for review, editing, and rollback, while preserving sample evidence trails for debugging and auditability. Experiments show that One-Eval can execute end-to-end evaluations from diverse natural-language requests with minimal user effort, supporting more efficient and reproducible evaluation in industrial settings. Our framework is publicly available at https://github.com/OpenDCAI/One-Eval.

When an off-the-shelf coding agent is used directly as a robot policy, observing a browser-based 3D interface through screenshots and acting by posing a virtual target gripper through a few tools, the agent's harness, its prompts, tools, and control rules, largely determines success, and until now it has been written by hand. We show that this harness can be improved automatically by another coding agent, the optimizer agent, and report two findings about what makes it work. First, the number of rollouts the optimizer agent sees per round governs whether the evolved harness is trustworthy, generalizes, and improves steadily. A single rollout is a noisy binary outcome, so with few rollouts per round a revision can be promoted on luck; enlarging the batch raises the signal-to-noise ratio of every promotion decision. Holding rounds fixed and growing the training set from 5 to 100 rollouts, held-out success rises from 47% to 67%, while small training sets overfit, reaching 70% on training tasks but only 54% held-out. Second, the optimizer agent must not be given free rein. With every revision it proposes accepted unconditionally, performance drifts downward within ten rounds as ill-judged edits accumulate; adding the most basic safeguard, Champion-Challenger selection that promotes a revision only if it strictly beats the incumbent on the same fixed evaluation set, turns the same loop into one that raises held-out success from 51% to 67% over 30 rounds. Automatic harness evolution for visual-interface robot agents is thus feasible, but its gains hinge on the rollout scale behind each decision and on how the optimizer agent's revisions are selected.

Image generation models have rapidly advanced, with many models now supporting generation conditioned on multiple reference images and textual instructions. Such models enable a wide range of applications, including entity fusion and attribute transfer across images. However, existing benchmarks for multi-reference image generation rely on limited evaluation dimensions and fail to capture the diverse challenges of the task. To address this limitation, we propose MDMR-Bench, a Multi-Dimensional benchmark for Multi-Reference image generation. MDMR-Bench is organized around representative capability settings and consists of multiple evaluation subsets spanning diverse dimensions of multi-reference image generation, including (1) common entity fusion and attribute transfer, (2) multi-element extraction from a single image, (3) infographic manipulation, and (4) visual reasoning-intensive generation. This design enables systematic analysis of model capabilities and failure modes across different challenge dimensions. Furthermore, to achieve more reliable and interpretable evaluation, we introduce a checklist-based evaluation protocol that decomposes generation quality into fine-grained evaluation aspects. Instead of relying solely on holistic scoring, our framework uses structured checklists composed of both global-level and atomic-level queries, enabling more diagnostic assessment of model performance. By benchmarking both open-source and proprietary models, we demonstrate that MDMR-Bench provides fine-grained insights into model performance across different dimensions. The dataset and code are publicly available.

We present WebGym, the largest-to-date open-source environment for training realistic visual web agents. Real websites are non-stationary and diverse, making artificial or small-scale task sets insufficient for robust policy learning. WebGym contains nearly 300,000 tasks with rubric-based evaluations across diverse, real-world websites and difficulty levels. We train agents with a simple reinforcement learning (RL) recipe, which trains on the agent's own interaction traces (rollouts), using task rewards as feedback to guide learning. To enable scaling RL, we speed up sampling of trajectories in WebGym by developing a high-throughput asynchronous rollout system, designed specifically for web agents. Our system achieves a 4-5x rollout speedup compared to naive implementations. Second, we scale the task set breadth, depth, and size, which results in continued performance improvement. Fine-tuning a strong base vision-language model, Qwen-3-VL-8B-Instruct, on WebGym results in an improvement in success rate on an out-of-distribution test set from 26.2% to 42.9%, significantly outperforming agents based on proprietary models such as GPT-4o and GPT-5-Thinking that achieve 27.1% and 29.8%, respectively. This improvement is substantial because our test set consists only of tasks on websites never seen during training, unlike many other prior works on training visual web agents.

Data policies for reinforcement learning with verifiable rewards (RLVR) determine which rollouts are used, how strongly they are weighted, and which domains contribute to subsequent training batches. We introduce DataFlex-RL, an evaluation platform for comparing these choices under a common GRPO recipe. Our primary experiment evaluates 13 configurations across 12 matched seeds using Qwen2.5-7B-Base and 12 mathematics, logic, and science benchmarks. Uniform GRPO improves the domain-balanced average accuracy by 7.76 percentage points over the untrained checkpoint. None of the eight rollout-selection or reweighting methods achieves a paired 95% confidence interval that excludes zero relative to uniform sampling, and none of the three adaptive mixtures outperforms a fixed equal mixture at the same level of precision. A corrected 12-seed extension on Llama-3.1-8B-Base places the additional methods on the same score scale as the original controls, but does not reveal a consistent winner in terms of observed mean performance. We also quantify evaluation sensitivity by rescoring nine Qwen2.5-7B-Instruct runs using a math-heavy six-benchmark summary, consisting of five mathematics benchmarks and GPQA-Diamond but no logic benchmark, and comparing it with the domain-balanced 12-benchmark summary. The resulting rankings are negatively correlated, with a correlation coefficient of -0.33, whereas summaries that retain all 12 benchmarks largely agree. Across the controlled settings studied here, changing the data policy measurably changes the training process but does not produce a reproducible improvement over uniform training.

Despite algorithm-level innovations for multi-agent reinforcement learning (MARL), the underlying networked infrastructure for large-scale MARL training remains underexplored. Existing training frameworks primarily optimize for single-agent scenarios and fail to address the unique system-level challenges of MARL, including rollout-training synchronization barriers, rollout load imbalance, and training resource underutilization. To bridge this gap, we propose FlexMARL, the first end-to-end training framework that holistically optimizes rollout, training, and their orchestration for large-scale LLM-based MARL. Specifically, FlexMARL introduces the joint orchestrator to manage data flow under the rollout-training disaggregated architecture. Building upon the experience store, a novel micro-batch driven asynchronous pipeline eliminates the synchronization barriers while providing strong consistency guarantees. Rollout engine adopts a parallel sampling scheme combined with hierarchical load balancing, which adapts to skewed inter/intra-agent request patterns. Training engine achieves on-demand hardware binding through agent-centric resource allocation. The training states of different agents are swapped via unified and location-agnostic communication. Empirical results on a large-scale production cluster demonstrate that FlexMARL achieves up to 7.3x speedup and improves hardware utilization by up to 5.6x compared to existing frameworks.

As the use of text-to-image generative models increases, so does the adoption of automatic benchmarking methods used in their evaluation. However, while metrics and datasets abound, there are few unified benchmarking libraries that provide a framework for performing evaluations across many datasets and metrics. Furthermore, the rapid introduction of increasingly robust benchmarking methods requires that evaluation libraries remain flexible to new datasets and metrics. Finally, there remains a gap in synthesizing evaluations in order to deliver actionable takeaways about model performance. To enable unified, flexible, and actionable evaluations, we introduce EvalGIM (pronounced ''EvalGym''), a library for evaluating generative image models. EvalGIM contains broad support for datasets and metrics used to measure quality, diversity, and consistency of text-to-image generative models. In addition, EvalGIM is designed with flexibility for user customization as a top priority and contains a structure that allows plug-and-play additions of new datasets and metrics. To enable actionable evaluation insights, we introduce ''Evaluation Exercises'' that highlight takeaways for specific evaluation questions. The Evaluation Exercises contain easy-to-use and reproducible implementations of two state-of-the-art evaluation methods of text-to-image generative models: consistency-diversity-realism Pareto Fronts and disaggregated measurements of performance disparities across groups. EvalGIM also contains Evaluation Exercises that introduce two new analysis methods for text-to-image generative models: robustness analyses of model rankings and balanced evaluations across different prompt styles. We encourage text-to-image model exploration with EvalGIM and invite contributions at https://github.com/facebookresearch/EvalGIM/.

World models -- generative models that simulate environment dynamics conditioned on past observations and actions -- are gaining prominence in planning, simulation, and embodied AI. However, evaluating their rollouts remains a fundamental challenge, requiring fine-grained, temporally grounded assessment of action alignment and semantic consistency -- capabilities not captured by existing metrics. Vision-Language Models (VLMs) have shown promise as automatic evaluators of generative content due to their strong multimodal reasoning abilities. Yet, their use in fine-grained, temporally sensitive evaluation tasks remains limited and requires targeted adaptation. We introduce a evaluation protocol targeting two recognition tasks -- action recognition and character recognition -- each assessed across binary, multiple-choice, and open-ended formats. To support this, we present UNIVERSE (UNIfied Vision-language Evaluator for Rollouts in Simulated Environments), a method for adapting VLMs to rollout evaluation under data and compute constraints. We conduct a large-scale study comparing full, partial, and parameter-efficient finetuning across task formats, context lengths, sampling strategies, and data compositions. The resulting unified evaluator matches the performance of task-specific baselines using a single checkpoint. Human studies confirm strong alignment with human judgments, establishing UNIVERSE as a scalable, semantics-aware evaluator for world models.

Long-horizon agents must coordinate reasoning, tool use, and environment feedback over extended interactions. Learning from sparse outcome rewards requires weighting trajectories of different lengths and updating the policy while rollouts are still being collected. We examine a single-rollout asynchronous recipe combining Muon, token-averaged loss, truncated importance sampling, and auxiliary length and tool-error penalties. Single-rollout sampling allows asynchronous learning without waiting for multiple trajectories for the same issue. We study optimizer choices and recipe components for repository-level issue resolution across two model families. The evaluated training runs show higher training success with the full recipe than with the baselines.

Large Language Model (LLM) agents increasingly solve long-horizon tasks through multi-turn interactions with users and external tools. In these settings, relevant task information often unfolds over time rather than being fully specified at the initial prompt. Service agents make this challenge especially concrete: users may clarify or revise their goals, while tool responses provide information needed for subsequent decisions. Thus, a final reward alone cannot indicate which actions contributed to resolving the task. Recent methods rely on comparative evidence from other trajectories or resampled continuations, or on separately constructed step-level learning signals, to refine credit. However, a completed rollout already records how information and errors flow between agent actions. We introduce Influence-Aware Policy Optimization (IAPO), which represents each rollout as a typed influence-dependency graph over trainable agent actions, with user and tool observations serving as evidence. IAPO converts support-use and failed-use structure into routing weights that redistribute the same trajectory-level advantage. Experiments with Qwen3-4B and Qwen3-8B demonstrate superior performance over multi-turn reinforcement learning (RL) baselines across three service-agent benchmarks: {τ^2}-Bench, UserBench, and AgentChangeBench. BFCL-v4 Multi-Turn further shows that these gains do not compromise multi-turn function-calling performance. This work advances the understanding of credit assignment in multi-turn user interactions and provides a principled approach to training service agents from sparse outcome feedback.

When adapting Vision Language Action (VLA) models to downstream tasks, multiple rounds of post-training are often required to progressively address policy weaknesses. In this report, we focus on maximizing human efficiency during this iterative process, measured by policy improvement and task throughput per unit of human labor and time. We propose HELP, a Human-Efficient Large-scale robot Post-training pipeline in which two specialized operators supervise twelve robots concurrently. A trained Teleoperator provides high-value remote interventions and recovery demonstrations, while a Floor Operator monitors the robot fleet, triggers takeovers, and performs physical resets. This role specialization improves human efficiency by reducing task switching, lowering operator training costs, and expanding robot interaction coverage. Beyond increasing rollout volume, concurrent supervision also broadens the range of policy behaviors observed by the human team, making recurring failure modes easier to identify and enabling more targeted takeovers, resets, and recovery demonstrations. To efficiently utilize the large and mixed-quality rollout data, HELP incorporates \vlac, an automatic rollout segmentation critic specifically designed for this setting. It separates autonomous trajectories into progress-making, idle, failure-inducing, and recovery segments. Useful rollout segments are retained and combined with Human-in-the-Loop data for the next post-training round. Across four real-world manipulation tasks, HELP achieves 80\%--95\% success rates and improves task throughput by 1.7times--4.2times over the base model. Under matched HITL recovery budgets, VLAC-CUT further amplifies throughput gains by 1.20times--3.43times and success-rate gains by 1.50times--3.00times over HITL-only updates.

AI evaluations are widely used for testing and understanding progress. However, the diverse evaluators bring with them inconsistencies that challenge analysis and comparison. First, results are saved in incompatible formats, scattered across leaderboards, papers, blog posts, evaluation harness logs, and custom repositories. Second, results are created by different evaluation frameworks, which produce divergent scores for nominally identical evaluations and record metadata inconsistently, hindering comparison, cross-community evaluation science, cost reduction, and reuse. We introduce Every Eval Ever, the first shared schema and community-crowdsourced repository for AI evaluation results. The schema standardizes how evaluations are represented in a unified, single JSON document. It is source-agnostic by design, ingesting results from evaluation harnesses and papers alike, and optionally stores per-instance outputs for fine-grained analysis. We contribute: (i) a community-governed metadata schema with a companion instance-level schema, the first standardization effort of its kind; (ii) automatic converters from popular formats, evaluation harnesses, and leaderboards to the unified schema; and (iii) a crowdsourced community database hosted on Hugging Face, currently spanning to date 22,235 models, 2,273 unique benchmarks, and 31 evaluation formats.

Scientific research follows multi-turn, multi-step workflows that require proactively searching the literature, consulting figures and tables, and integrating evidence across papers to align experimental settings and support reproducible conclusions. This joint capability is not systematically assessed in existing benchmarks, which largely under-evaluate proactive search, multi-evidence integration and sustained evidence use over time. In this work, we introduce EpiBench, an episodic multi-turn multimodal benchmark that instantiates short research workflows. Given a research task, agents must navigate across papers over multiple turns, align evidence from figures and tables, and use the accumulated evidence in the memory to answer objective questions that require cross paper comparisons and multi-figure integration. EpiBench introduces a process-level evaluation framework for fine-grained testing and diagnosis of research agents. Our experiments show that even the leading model achieves an accuracy of only 29.23% on the hard split, indicating substantial room for improvement in multi-turn, multi-evidence research workflows, providing an evaluation platform for verifiable and reproducible research agents.

The inherent rigidity of fixed-size benchmarks makes them an inefficient tool for model evaluation. Diverse evaluation objectives, including model ranking, model selection and testing throughout development, demand varying levels of statistical power. The mismatch between fixed sample sizes and these diverse needs results in either excessive computational cost or compromised reliability - a critical concern for model evaluation. To overcome these limitations, we call for adoption of sequential testing in our field. We provide an adaptive evaluation framework, that provides a principled way to navigate the trade-off between efficiency and reliability in model evaluation. Our framework combines the established statistical paradigm of sequential testing with stopping criteria tailored to common evaluation needs such as diminishing returns detection, and minimum detectable effect size. We demonstrate its ability to adaptively manage the efficiency-reliability trade-off on the Open VLM Leaderboard, including, for example, a 80% reduction in computational cost compared to fixed-size evaluation (with a 2.5-point CI width allowance) while maintaining statistical significance.

Asynchronous rollout improves the efficiency of reinforcement learning with verifiable rewards (RLVR) by reducing under-utilization during long-tail generation. Sustained concurrency also concentrates context processing into fewer decode steps, increasing the KV workload per step. We characterize this KV concentration and find that KV cache quantization delivers substantially greater acceleration than weight-only and weight-activation quantization in our matched-workload comparisons. However, INT4 KV rollout can degrade learning even with token-level truncated importance sampling (TIS). We propose KV4-RL, which combines calibration-free INT4 KV quantization with selective INT8 protection of attention sinks and recent context. Protecting a bounded number of positions reduces sampler mismatch while retaining most of the speedup from INT4. Across 400 iterations of Qwen3-8B-Base GRPO training, KV4-RL achieves 1.80times cumulative rollout speedup and 1.27times cumulative full-iteration speedup over BF16 rollout, reaching 1.97times rollout speedup over the final 50 iterations. Under a common BF16 evaluator, its final six-benchmark mean accuracy is within 0.5 percentage points of BF16-rollout training for both Qwen3-4B-Base and Qwen3-8B-Base.

Rollout generation is a major bottleneck in Reinforcement Learning (RL) for Mixture-of-Experts (MoE) Large Language Models, motivating low-precision rollout acceleration such as FP8. As an emerging low-precision format, NVFP4 combines fine-grained scaling for accuracy preservation with native W4A4 FP4 GEMMs for higher throughput than FP8. However, we find that directly applying NVFP4 to MoE RL rollout is impractical. NVFP4 rollout with BF16 training collapses after roughly 150 steps, accompanied by rapidly growing rollout-trainer log-probability gaps. Through training-inference error analysis and controlled ablations, we identify activation error, rather than weight error, as the dominant source of FP4 RL instability: weights can be synchronized and aligned by a shared quantization-dequantization path, whereas activations are recomputed online and error is amplified by the coarse E2M1 grid. Therefore, to stabilize NVFP4 RL for MoE, we propose QUantization-error Alignment across Dual Sides (QUADS). On the trainer side, we introduce Asymmetric Quantization-Aware Training fake-quantizing weights while keeping activations unquantized for better alignment. On the rollout side, Residual Activation Compensation corrects high-error activation channels while preserving native W4A4 GEMMs. In our MoE RL experiments on several benchmarks, QUADS achieves BF16-level accuracy, improves average pass@1 by 21.49 points over naive NVFP4 RL, and delivers ~16% higher rollout throughput than FP8.

Enterprise adoption of LLMs in agentic workflows is gated by evaluations that span a wide arc of abilities: CRM operations, heterogeneous-document retrieval, personal data understanding, multi-turn planning, tool use, function calling, long-horizon research, and long-context software engineering. These evaluations live today in separate repositories with divergent runners, dependencies, and scoring conventions, which makes cross-benchmark comparison difficult and obscures what a single aggregate number actually represents. We release EGI-BENCH, a unified LLM agent benchmark suite that consolidates eight enterprise facing agent benchmarks (CRMARENA, HERB, PERSONABENCH, USERBENCH, MCP-UNIVERSE, MCPEVAL, BENCHFORCE, LOCOBENCH) behind a single command-line interface with two runtime-budgeted execution modes. On top of the suite we define the Enterprise General Intelligence (EGI) Score, a six-dimensional framework (M1-M6) whose buckets are populated by fine-grained sub-metrics routed from each benchmark rather than by benchmark-level averages. We describe the suite, the sub-metric routing, and the quick/deep budgets, and report results for fourteen frontier and cost-efficient LLMs spanning the GPT, Claude, and Gemini families under quick mode. Top-line EGI ranges from 39.6 (GPT-4o mini) to 65.6 (gpt-5.5), and no single model dominates every capability axis: gpt-5.5 leads task outcome, planning, and retrieval grounding; Claude Sonnet 4.5 leads tool actioning and safety/abstention; and Claude Haiku 4.5 leads engineering quality. Code is available at https://github.com/SalesforceAIResearch/egi-bench.

Reinforcement learning (RL) is a critical stage in post-training large language models (LLMs), involving repeated interaction between rollout generation, reward evaluation, and centralized learning. Distributing rollout execution offers opportunities to leverage more cost-efficient inference resources, but introduces challenges in wide-area coordination and policy dissemination. We present ECHO-2, a distributed RL framework for post-training with remote inference workers and non-negligible dissemination latency. ECHO-2 combines centralized learning with distributed rollouts and treats bounded policy staleness as a user-controlled parameter, enabling rollout generation, dissemination, and training to overlap. We introduce an overlap-based capacity model that relates training time, dissemination latency, and rollout throughput, yielding a practical provisioning rule for sustaining learner utilization. To mitigate dissemination bottlenecks and lower cost, ECHO-2 employs peer-assisted pipelined broadcast and cost-aware activation of heterogeneous workers. Experiments on GRPO post-training of 4B and 8B models under real wide-area bandwidth regimes show that ECHO-2 significantly improves cost efficiency while preserving RL reward comparable to strong baselines.

Reinforcement learning with verifiable rewards (RLVR) has emerged as a highly effective framework for improving LLM reasoning, with methods such as GRPO among its most successful instantiations. However, GRPO relies on repeated generation of long chain-of-thought rollouts. Training time scales with the number of rollouts, a large fraction of which are uninformative. Thus, GRPO is computationally expensive and unstable. To mitigate this, existing approaches either generate a larger pool of rollouts and filter the most informative prompts, or leverage historical signals for filtering at later stages of training. These strategies offer modest performance gains, but slow down the overall process. To address this, we propose VarIance Guided Online Rollout allocation (VIGOR) which instead of allocating a fixed rollout budget per example, begins with a small number of rollouts for all examples in a batch and iteratively allocates additional rollouts to those with the highest group reward variance until a fixed total rollout budget is reached. Theoretically, we show that under RLVR, reward variance controls the gradient magnitude, and derive VIGOR's closed-form speedup ratio over GRPO, which grows with refinement rounds under Pareto-distributed reward variance. Experiments on mathematical reasoning and coding tasks show that VIGOR reaches target accuracy with up to 2.3times fewer rollouts on math, reaches GRPO's final coding full pass rate with 1.49times fewer rollouts, and improves the coding average test pass rate by 3.4 points.

Working agents need to read diverse files, coordinate tools, and produce deliverables. Training such agents requires tasks built on many real files with verifiable results, but few pipelines exist to synthesize this kind of data. Existing pipelines either generate files with models, which lack realism and diversity, or build tasks on real files without task-specific verifiers, leaving result quality unchecked. We introduce GraphForge, an evidence-graph based framework that grounds both the task and its verification in real files. Starting from occupation-grounded seeds for controlled diversity, GraphForge assembles a workspace of real files for each seed and builds an evidence graph over their relations. Since the task statement and rubrics are both derived from this graph, task requirements are backed by the workspace files and each criterion is anchored to the files needed to verify it. An initial rollout further tests executability, and a revision agent repairs the task and rubrics against the original files before trajectories are collected. Fine-tuning Qwen3.6-27B on 2,169 GraphForge trajectories brings GDPVal to 1445.7 (+65.7) under OpenHands, and Workspace-Bench-Lite and SpreadsheetBench II to 63.7 (+7.7) and 24.0 (+13.7) under Claude Code. Rejection fine-tuning on the SFT model's own rollouts, with candidates selected by the evidence-anchored rubrics, yields further improvements on all three benchmarks, suggesting that the rubrics provide a useful selection signal. The data and models are available.

Reinforcement-Learning-based post-training has recently emerged as a promising paradigm for aligning text-to-image diffusion models with human preferences. In recent studies, increasing the rollout group size yields pronounced performance improvements, indicating substantial room for further alignment gains. However, scaling rollouts on large-scale foundational diffusion models (e.g., FLUX.1-12B) imposes a heavy computational burden. To alleviate this bottleneck, we explore the integration of FP4 quantization into Diffusion RL rollouts. Yet, we identify that naive quantized pipelines inherently introduce risks of performance degradation. To overcome this dilemma between efficiency and training integrity, we propose Sol-RL (Speed-of-light RL), a novel FP4-empowered Two-stage Reinforcement Learning framework. First, we utilize high-throughput NVFP4 rollouts to generate a massive candidate pool and extract a highly contrastive subset. Second, we regenerate these selected samples in BF16 precision and optimize the policy exclusively on them. By decoupling candidate exploration from policy optimization, Sol-RL integrates the algorithmic mechanisms of rollout scaling with the system-level throughput gains of NVFP4. This synergistic algorithm-hardware design effectively accelerates the rollout phase while reserving high-fidelity samples for optimization. We empirically demonstrate that our framework maintains the training integrity of BF16 precision pipeline while fully exploiting the throughput gains enabled by FP4 arithmetic. Extensive experiments across SANA, FLUX.1, and SD3.5-L substantiate that our approach delivers superior alignment performance across multiple metrics while accelerating training convergence by up to 4.64times, unlocking the power of massive rollout scaling at a fraction of the cost.

Multi-step agentic reinforcement learning benefits from fine-grained credit assignment, yet existing approaches offer limited options: critic-free methods like GRPO assign a uniform advantage to every action in a trajectory, while learned value networks introduce notable overhead and can be fragile under sparse rewards. We observe that group rollouts targeting the same problem often traverse overlapping intermediate states, implicitly forming a tree whose branches diverge at successive decision points. Building on this insight, we introduce Rollout-Tree Monte Carlo (RTMC) advantage estimation, which aggregates return statistics across rollouts sharing a common state to produce per-step Q-values and advantages--without any learned critic. A state-action signature system compresses raw interaction histories into compact, comparable representations, making cross-rollout state matching tractable. On SWE-bench Verified, RTMC improves pass@1 by 3.2 percentage points over GRPO.

Autoregressive rollout generation is a major computational cost in reinforcement learning for large language models. Reusing each rollout batch for additional learner updates amortizes this cost, but later updates become increasingly off-policy as the learner departs from the behavior policy. At a token position, exact off-policy correction must account for both the current action and the probability of reaching its prefix. The cumulative importance ratio provides this correction, but its product form can produce an unwieldy dynamic range. We study Prefix-Normalized Policy Optimization (PNPO), which replaces the cumulative ratio with the geometric mean of likelihood ratios along each causal prefix, preserving causal-prefix dependence at each position while compressing the log-weight scale. In controlled long-context mathematical reasoning experiments, we induce two off-policy regimes by using one or four policy-update epochs per rollout batch. PNPO does not consistently outperform GSPO with one epoch. With four epochs, it attains the highest observed Avg@32 on each benchmark; the unweighted mean of the three independently selected benchmark peaks is 50.24, 3.00 percentage points above GSPO. Under a matched 2,400-update budget, four-epoch PNPO reaches a final macro Avg@32 of 49.66 after 150 rollout batches, comparable to the 49.56 reached after 600 batches with one epoch. These results provide preliminary evidence that PNPO can be advantageous as training moves further off-policy.

Evaluating generalist robot manipulation policies in the real world is expensive, slow, and difficult to scale. Action-conditioned video world models offer a scalable alternative by simulating policy rollouts. Autoregressive rollouts accumulate compounding errors, observations across multiple camera views must remain mutually consistent, and the evaluator must generalize to policies whose behaviors lie outside the training distribution. We address these challenges with SC3-Eval, a self-consistent video generation recipe that adapts a pre-trained video foundation model into an accurate policy evaluator by enforcing three complementary forms of consistency. First, forward-inverse dynamics consistency jointly trains the model to predict frames from actions and to recover actions from frames, anchoring generated rollouts to a physically plausible action manifold and counteracting the drift a forward-only model cannot penalize. Second, cross-view consistency trains the model to inpaint each camera view from the other, keeping the multi-camera observation coherent over long rollouts without any explicit memory mechanism. Third, test-time consistency reuses the inverse dynamics mode at inference as a per-action-chunk uncertainty signal that terminates rollouts whose generated frames drift away from the requested actions. We also demonstrate SC3-Eval rollouts reproduce the failure modes that policies exhibit in real-world rollouts, supporting fine-grained diagnostic comparison rather than aggregate ranking alone. Across seven real-world vision-language-action policies, SC3-Eval attains a closed-loop Pearson correlation of 0.929 and MMRV of 0.119, outperforming three strong prior video-model-based baselines, and generalizes to new tasks.

We present MEGA-Bench, an evaluation suite that scales multimodal evaluation to over 500 real-world tasks, to address the highly heterogeneous daily use cases of end users. Our objective is to optimize for a set of high-quality data samples that cover a highly diverse and rich set of multimodal tasks, while enabling cost-effective and accurate model evaluation. In particular, we collected 505 realistic tasks encompassing over 8,000 samples from 16 expert annotators to extensively cover the multimodal task space. Instead of unifying these problems into standard multi-choice questions (like MMMU, MMBench, and MMT-Bench), we embrace a wide range of output formats like numbers, phrases, code, \LaTeX, coordinates, JSON, free-form, etc. To accommodate these formats, we developed over 40 metrics to evaluate these tasks. Unlike existing benchmarks, MEGA-Bench offers a fine-grained capability report across multiple dimensions (e.g., application, input type, output format, skill), allowing users to interact with and visualize model capabilities in depth. We evaluate a wide variety of frontier vision-language models on MEGA-Bench to understand their capabilities across these dimensions.

Group-based policy optimization methods like GRPO and GSPO have become standard for training multimodal models, leveraging group-wise rollouts and relative advantage estimation. However, they suffer from a critical gradient vanishing problem when all responses within a group receive identical rewards, causing advantage estimates to collapse and training signals to diminish. Existing attempts to mitigate this issue fall into two paradigms: filtering-based and sampling-based methods. Filtering-based methods first generate rollouts broadly and then retroactively filter out uninformative groups, leading to substantial computational overhead. Sampling-based methods proactively select effective samples before rollout but rely on static criteria or prior dataset knowledge, lacking real-time adaptability. To address these issues, we propose VADE, a Variance-Aware Dynamic sampling framework via online sample-level difficulty Estimation. Our framework integrates three key components: online sample-level difficulty estimation using Beta distributions, a Thompson sampler that maximizes information gain through the estimated correctness probability, and a two-scale prior decay mechanism that maintains robust estimation under policy evolution. This three components design enables VADE to dynamically select the most informative samples, thereby amplifying training signals while eliminating extra rollout costs. Extensive experiments on multimodal reasoning benchmarks show that VADE consistently outperforms strong baselines in both performance and sample efficiency, while achieving a dramatic reduction in computational overhead. More importantly, our framework can serves as a plug-and-play component to be seamlessly integrated into existing group-based RL algorithms. Code and models are available at https://VADE-RL.github.io.

Reinforcement Learning with Verifiable Rewards (RLVR) methods such as GRPO rely on successful self-generated trajectories, but finite rollout budgets can produce all-fail groups with no reward-based policy-gradient signal. While additional rollouts improve the chance of success at higher cost, successful trajectories missing from one model's rollouts may already have been discovered by another. Indeed, we observe that heterogeneous models often succeed on complementary prompts, creating opportunities for mutual learning without a designated stronger teacher. To exploit this complementarity, we propose GRAFT (Gated Replacement of Answer-Failed groups with peer Trajectories), an off-policy-aware framework that replaces all-fail groups with informative peer groups. GRAFT transfers both successful and unsuccessful peer responses with peer-computed advantages, while controlling cross-model mismatch through sequence-level compatibility weighting and token-level importance ratio clipping. Across three heterogeneous model pairs and five mathematical reasoning benchmarks, GRAFT consistently improves both models over GRPO with the same per-model rollout budget, gaining 2.1 points on average and up to 4.5 points in model-level average performance. Stored peer trajectories preserve most of the gains, improving over GRPO by 1.8 points on average without simultaneous co-training.

AI evaluation results are produced at scale but reported inconsistently across leaderboards, model cards, benchmark papers, and company blogs. The cost is interpretive: readers cannot reliably compare results across sources, identify what a report omits, or trace an aggregate claim to its underlying evidence. Recent efforts address isolated components but leave three gaps: they cover only narrow slices of the evaluation lifecycle and do not compose into a single interpretable record; they specify static representations that do not differentiate the questions different stakeholders bring to the same evidence; and they remain proposals on paper, lacking the extraction infrastructure required for adoption at scale. We present , an operational reporting layer that composes benchmark metadata, evaluation run data, and model metadata into a unified record. We (1) derive a reporting schema from a structured review of 52 papers and 10 stakeholder interviews, (2) implement four interpretive signals (reproducibility, documentation completeness, provenance and risk, and score comparability), rendered through reader modes calibrated to research and non-research audiences, and (3) deploy a monitoring tool that applies across 5,816 models, 635 benchmarks, and 101,843 results, surfacing systematic gaps in current reporting practice.

Context: Bugs in code are inevitable and can lead to severe consequences, ranging from security vulnerabilities to operational failures. Debugging software remains challenging despite advances in testing and verification, often requiring extensive manual effort. Learning-based automated program repair (APR) has shown promise in reducing the time, effort, and cost of manually fixing bugs. However, existing techniques face several challenges, including language-dependent strategies, limited bug context utilization, and difficulties in handling bugs that span multiple locations in the code. Objective: This paper introduces MultiMend, a learning-based APR approach designed to improve repair performance on multiple programming languages with language-independent context augmentation and multi-hunk patch generation. Method: MultiMend fine-tunes a pre-trained encoder-decoder transformer model (CodeT5) to generate bug-fixing patches. It embeds source code lines and applies retrieval-augmented generation to augment the buggy context with relevant lines during patch generation. The approach systematically constructs patches for multi-hunk bugs to reduce the needed patch validations. We evaluate MultiMend on four benchmarks with four programming languages and compare it with state-of-the-art methods. Results: Experimental results show that MultiMend achieves competitive effectiveness and efficiency against compared tools. Across all benchmarks, MultiMend fixes 2,077 bugs, of which 1,455 are identical to the developer's patch, and 106 are for multi-hunk bugs. Both context augmentation and multi-hunk patch generation positively contribute to the results. Conclusion: MultiMend shows promising performance across benchmarks. The findings highlight its applicability to real-world software maintenance and its potential to reduce manual debugging efforts.

A benchmark should deliver more than a scalar score: what makes an evaluation trustworthy is the reasoning that justifies the score. This is especially critical for world models, where judging a rollout requires understanding whether physics, causality, and world state evolve correctly. Humans spot such violations naturally, yet no existing benchmark automates this capability: metrics are computed brute-force, leaving no reasoning chain that can be examined or verified. We introduce HarnessEval-W, an agentified evaluation pipeline that brings the harness paradigm from the LLM ecosystem to world model benchmarking. Rather than applying a fixed rubric, HarnessEval-W interprets the context of each evaluation case, decomposes the evaluation question into measurable subproblems, and spawns specialized sub-agents, each equipped with tailored context and diagnostic tools to reason over its own subproblem. The parent agent then validates the gathered evidence and summarizes it into the final verdict. This hierarchical workflow turns every evaluation into a transparent evidence tree whose complete reasoning chain justifies the result. We apply HarnessEval-W to 18 representative world models over 330 evaluation cases. Its judgments closely align with human preferences while providing verifiable, fine-grained diagnoses of every generated rollout. We open-source the full pipeline as a live benchmark and invite the broad community to contribute to grow new skills and evaluation cases as world models evolve.

Despite the success of reinforcement learning for large language models, a common failure mode is reduced sampling diversity, where the policy repeatedly generates similar erroneous behaviors. Classical entropy regularization encourages randomness under the current policy, but does not explicitly discourage recurrent failure patterns across rollouts. We propose MEDS, a Memory-Enhanced Dynamic reward Shaping framework that incorporates historical behavioral signals into reward design. By storing and leveraging intermediate model representations, we capture features of past rollouts and use density-based clustering to identify frequently recurring error patterns. Rollouts assigned to more prevalent error clusters are penalized more heavily, encouraging broader exploration while reducing repeated mistakes. Across five datasets and three base models, MEDS consistently improves average performance over existing baselines, achieving gains of up to 4.13 pass@1 points and 4.37 pass@128 points. Additional analyses using both LLM-based annotations and quantitative diversity metrics show that MEDS increases behavioral diversity during sampling.

Large Language Models (LLMs) increasingly rely on reinforcement learning with verifiable rewards (RLVR) to elicit reliable chain-of-thought reasoning. However, the training process remains bottlenecked by the computationally expensive rollout stage. Existing acceleration methods-such as parallelization, objective- and data-driven modifications, and replay buffers-either incur diminishing returns, introduce bias, or overlook redundancy across iterations. We identify that rollouts from consecutive training epochs frequently share a large portion of overlapping segments, wasting computation. To address this, we propose SPEC-RL, a novel framework that integrates SPECulative decoding with the RL rollout process. SPEC-RL reuses prior trajectory segments as speculative prefixes and extends them via a draft-and-verify mechanism, avoiding redundant generation while ensuring policy consistency. Experiments on diverse math reasoning and generalization benchmarks, including GSM8K, MATH-500, OlympiadBench, MMLU-STEM, and others, demonstrate that SPEC-RL reduces rollout time by 2-3x without compromising policy quality. As a purely rollout-stage enhancement, SPEC-RL integrates seamlessly with mainstream algorithms (e.g., PPO, GRPO, DAPO), offering a general and practical path to scale RLVR for large reasoning models. Our code is available at https://github.com/ShopeeLLM/Spec-RL

Reinforcement Learning with Verifiable Rewards (RLVR) has become a key paradigm to improve the reasoning capabilities of Multimodal Large Language Models (MLLMs). However, prevalent group-based algorithms such as GRPO require multi-rollout sampling for each prompt. While more efficient single-rollout variants have recently been explored in text-only settings, we find that they suffer from severe instability in multimodal contexts, often leading to training collapse. To address this training efficiency-stability trade-off, we introduce MSSR (Multimodal Stabilized Single-Rollout), a group-free RLVR framework that achieves both stable optimization and effective multimodal reasoning performance. MSSR achieves this via an entropy-based advantage-shaping mechanism that adaptively regularizes advantage magnitudes, preventing collapse and maintaining training stability. While such mechanisms have been used in group-based RLVR, we show that in the multimodal single-rollout setting they are not merely beneficial but essential for stability. In in-distribution evaluations, MSSR demonstrates superior training compute efficiency, achieving similar validation accuracy to the group-based baseline with half the training steps. When trained for the same number of steps, MSSR's performance surpasses the group-based baseline and shows consistent generalization improvements across five diverse reasoning-intensive benchmarks. Together, these results demonstrate that MSSR enables stable, compute-efficient, and effective RLVR for complex multimodal reasoning tasks.

Test-time reinforcement learning generates multiple candidate answers via repeated rollouts and performs online updates using pseudo-labels constructed by majority voting. To reduce overhead and improve exploration, prior work introduces tree structured rollouts, which share reasoning prefixes and branch at key nodes to improve sampling efficiency. However, this paradigm still faces two challenges: (1) high entropy branching can trigger rollout collapse, where the branching budget concentrates on a few trajectories with consecutive high-entropy segments, rapidly reducing the number of effective branches; (2) early pseudo-labels are noisy and biased, which can induce self-reinforcing overfitting, causing the policy to sharpen prematurely and suppress exploration. To address these issues, we propose Entropy Confidence Hybrid Group Relative Policy Optimization (ECHO). During rollout, ECHO jointly leverages local entropy and group level confidence to adaptively control branch width, and further introduces online confidence-based pruning to terminate persistently low confidence branches, avoiding high entropy traps and mitigating collapse. During policy updates, ECHO employs confidence adaptive clipping and an entropy confidence hybrid advantage shaping approach to enhance training robustness and mitigate early stage bias. Experiments demonstrate that ECHO achieves consistent gains on multiple mathematical and visual reasoning benchmarks, and generalizes more effectively under a limited rollout budget.

Professional-document review is a constraint-checking problem in which decisions depend on relations among text, geometry, pages, and document revisions. We present an evidence-grounded pipeline that normalizes extracted facts, executes four-state rules deterministically, retains source spans, and escalates unresolved cases. We evaluate its PDF evidence allocator on 160 reference-based tasks from 29 construction projects using a repeated four-system test and a disjoint two-system breadth extension. In the repeated test, reallocating a four-image budget from retrieved page overviews to one overview and three overlapping tiles improves project-family standardized decision accuracy by 10.6 percentage points (95% project-cluster bootstrap CI: 4.3 to 18.0; exact p = 0.031). This effect does not persist in the broader block: Region-RAG changes accuracy by -4.1 points (95% CI: -10.2 to 1.9; exact p = 0.209), while an equal-image sensitivity favors page breadth. Exact finding-set recovery remains low, false passes remain common, and repeated-run agreement is poorly calibrated. The results identify a resolution-breadth trade-off rather than a universal advantage for region-focused evidence, motivating rule-aware evidence routing and expert review.

Reinforcement learning with verifiable rewards (RLVR) is a promising approach for enhancing reasoning and agentic behavior in large language models. However, rollout-intensive policy optimization is often limited by insufficient reward contrast, arising when overly simple or complex prompts generate low-variance feedback and when outcome-only rewards assign the same terminal assessment to every decision in a multi-turn rollout. Past efforts have focused on allocating available rollout resources to promising prompts, yet they only leverage sample informativeness at the prompt level and neglect variation in prefix-level informativeness across turns within the same rollout. This work targets multi-turn agentic RL by modeling each ReAct-style thought-action-observation turn as a semantically distinct node, allowing budget allocation to extend from prompt roots to turn-level prefixes with further continuations, which naturally forms tree-structured rollouts. We introduce Tree Rollout Allocation for Contrastive Exploration (TRACE), a unified rollout allocation framework that enhances reward contrast within a fixed sampling budget. Technically, TRACE allocates rollout budget to both prompt roots and intermediate prefixes that are most likely to yield mixed terminal rewards. A shared generalizable predictor estimates conditional success probability at these anchors from prefix histories to guide this allocation. The resulting adaptive tree structure enriches outcome-only feedback and amplifies the policy-update signal. Empirically, TRACE achieves competitive performance and efficiency gains on typical agentic benchmarks, e.g., improving Qwen3-14B Multi-Hop QA average accuracy by 2.8 points over competitive baselines at equal sampling cost.

Training vision-language web agents with multi-step RL is compute-intensive, with two dominant forms of inefficiency: idle GPUs in synchronous RL, and trajectories that use more steps and tokens than necessary. We present AsyncWebRL, which addresses both. On the system side, an asynchronous design overlaps rollout, gradient update, and policy refresh across iterations, paired with two web-agent-specific adaptations, namely an everlasting rollout pool and lightweight screenshot handling, that together deliver up to a 2.9times end-to-end training-throughput speedup over the previously fastest open synchronous pipeline (WebGym). On the algorithmic side, we identify the per-trajectory normalizer 1/|τ_i| in multi-step GRPO as the root cause of trajectory-level and token-level inefficiency: because failures are systematically longer than successes, it down-weights the negative gradient on failed tokens, so the policy keeps producing verbose memory schemas. Replacing 1/|τ_i| with a constant 1/k breaks this coupling, contracting trajectories while preserving aggregate success. Together, these contributions set a new open-source state of the art on the WebGym out-of-distribution test split (+5.8% relative over the 42.9% prior best), with the largest gains on the harder slices (+42% relative on Medium, +48% relative on Hard).

With the wide application of large language models (LLMs) in real-world scenarios, the value implication of their outputs is crucial. However, existing evaluation benchmarks suffer from insufficient coverage of value dilemmas in daily scenarios involving multiple value conflicts and simplistic evaluation formalisms that fail to assess LLMs' value alignment. To address these issues, we propose D2VBench, a value alignment benchmark comprising 10,000 instances of real daily dilemma scenarios constructed through a multi-stage collaboration between LLMs and humans, grounded in 158 manually annotated fine-grained value concepts. For evaluation on the benchmark, we present a hybrid evaluation paradigm that integrates multiple-choice questions with open-ended questions. We conduct comprehensive evaluations on eight mainstream LLMs. Experimental results demonstrate that D2VBench exhibits high reliability and robustness, effectively reflecting the LLMs' alignment across different value categories and dimensions, and providing a more realistic and fine-grained tool for research on value alignment. The dataset is available at https://github.com/tjunlp-lab/D2VBench.

Evaluations (Evals) are a deployment bottleneck for real-world AI applications: public benchmarks rarely match a team's users, context, or policies, and human review is often tedious to scale. Motivated by our work with AI applications in the public sector, this project addresses recurring evaluation challenges encountered when applications must satisfy local policy and governance requirements. We present Kaleidoscope, an integrated workflow for contextual functional evaluation that links persona-based test generation, contextualized rubrics, and human review for reliability-gated automated scoring. Generated test cases are scored against application-specific rubrics; human annotations provide reviewable labels; and LLM judges automate scoring only when their agreement with those labels meets a configured threshold. Kaleidoscope is therefore a practical, inspectable, iterative workflow for product teams. We report early evidence from a three-week pilot across four organizational use cases and custom-rubric judge experiments on 108 annotated Q\&A pairs spanning four domains and 14 evaluation dimensions. The results highlight useful features for end-to-end reliable, automated scoring.

Automated environment configuration is a critical bottleneck in scaling software engineering (SWE) automation. To provide a reliable evaluation standard for this task, we present Multi-Docker-Eval benchmark. It includes 40 real-world repositories spanning 9 programming languages and measures both success in achieving executable states and efficiency under realistic constraints. Our extensive evaluation of state-of-the-art LLMs and agent frameworks reveals key insights: (1) the overall success rate of current models is low (F2P at most 37.7%), with environment construction being the primary bottleneck; (2) model size and reasoning length are not decisive factors, and open-source models like DeepSeek-V3.1 and Kimi-K2 are competitive in both efficiency and effectiveness; (3) agent framework and programming language also have significantly influence on success rate. These findings provide actionable guidelines for building scalable, fully automated SWE pipelines.

On-policy learning has been argued to reduce catastrophic forgetting, produce sparser parameter updates, and improve generalisation. However, existing comparisons between supervised fine-tuning and reinforcement learning vary many factors simultaneously, making the contribution of rollout policy difficult to isolate. We study the effect of rollout policy in a controlled strong-to-weak distillation setting, by independently varying rollout policy, token-level KL direction, and learning rate across the Llama3 and Qwen2.5 model families and reasoning tasks spanning scientific, medical, and arithmetic domains. Our analysis reveals a nuanced picture of distillation dynamics in which rollout policy does not necessarily play a central role. Instead, token-level KL direction more clearly shapes task performance and output coverage, while learning rate governs forgetting and update sparsity. Analysis of KL gradients and experiments along a continuous student-teacher rollout-policy spectrum explain this pattern: forward KL is remarkably robust to rollout policy, with its performance stable and strong despite changes to the rollout policy, whereas reverse KL is substantially more sensitive and favours student-generated rollouts. On-policy data nevertheless improves generalisation to harder variants of the Countdown arithmetic task under both KL directions, although this advantage does not reliably persist after subsequent RLVR. Our broader conclusions remain robust to removing gradient clipping, using sampled KL estimators, and training on tasks requiring longer reasoning chains. Overall, our results challenge the view that on-policy rollouts are inherently preferable and show that their value depends critically on the objective, evaluation setting, and optimisation hyperparameters.

Reinforcement Learning from Verifiable Rewards (RLVR) has emerged as a powerful paradigm for enhancing the complex reasoning capabilities of Large Reasoning Models. However, standard outcome-based supervision suffers from a critical limitation that penalizes trajectories that are largely correct but fail due to several missteps as heavily as completely erroneous ones. This coarse feedback signal causes the model to discard valuable largely correct rollouts, leading to a degradation in rollout diversity that prematurely narrows the exploration space. Process Reward Models have demonstrated efficacy in providing reliable step-wise verification for test-time scaling, naively integrating these signals into RLVR as dense rewards proves ineffective.Prior methods attempt to introduce off-policy guided whole-trajectory replacement that often outside the policy model's distribution, but still fail to utilize the largely correct rollouts generated by the model itself and thus do not effectively mitigate the narrowing of the exploration space. To address these issues, we propose SCOPE (Step-wise Correction for On-Policy Exploration), a novel framework that utilizes Process Reward Models to pinpoint the first erroneous step in suboptimal rollouts and applies fine-grained, step-wise off-policy rectification. By applying precise refinement on partially correct rollout, our method effectively salvages partially correct trajectories and increases diversity score by 13.5%, thereby sustaining a broad exploration space. Extensive experiments demonstrate that our approach establishes new state-of-the-art results, achieving an average accuracy of 46.6% on math reasoning and exhibiting robust generalization with 53.4% accuracy on out-of-distribution reasoning tasks.

Vision-Language-Action (VLA) models are increasingly evaluated across multiple simulation benchmarks, yet adding each benchmark to an evaluation pipeline requires resolving incompatible dependencies, matching underspecified evaluation protocols, and reverse-engineering undocumented preprocessing. This burden scales with the number of models and benchmarks, making comprehensive evaluation impractical for most teams. We present vla-eval, an open-source evaluation harness that eliminates this per-benchmark cost by decoupling model inference from benchmark execution through a WebSocket+msgpack protocol with Docker-based environment isolation. Models integrate once by implementing a single predict() method; benchmarks integrate once via a four-method interface; the full cross-evaluation matrix works automatically. The framework supports 14 simulation benchmarks and six model servers. Parallel evaluation via episode sharding and batch inference achieves up to 47x wall-clock speedup, completing 2,000 LIBERO episodes in ~18 minutes. To validate the framework, we reproduce published scores across six VLA codebases and three benchmarks, documenting previously undocumented pitfalls. We additionally release a VLA leaderboard aggregating 657 published results across 17 benchmarks. Framework, evaluation configs, and all reproduction results are publicly available at https://github.com/allenai/vla-evaluation-harness and https://allenai.github.io/vla-evaluation-harness/leaderboard.

Reinforcement Learning with Verifiable Rewards (RLVR) is an effective paradigm for improving the reasoning capabilities of large language models. However, existing RLVR methods utilize rollouts in an indiscriminate and short-horizon manner: responses of heterogeneous quality within each prompt are treated uniformly, and historical rollouts are discarded after a single use. This leads to noisy supervision, poor sample efficiency, and suboptimal policy updates. We address these issues by formulating rollout scheduling in RLVR as a contextual bandit problem and proposing a unified neural scheduling framework that adaptively selects high-value rollouts throughout training. Each rollout is treated as an arm whose reward is defined by the induced performance gain between consecutive optimization steps. The resulting scheduler supports both noise-aware intra-group selection and adaptive global reuse of historical rollouts within a single principled framework. We provide theoretical justification by deriving sublinear regret bounds and showing that enlarging the rollout buffer improves the achievable performance upper bound. Experiments on six mathematical reasoning benchmarks demonstrate consistent gains in performance and training efficiency across multiple RLVR optimization methods.

As a typical and practical application of Large Language Models (LLMs), Retrieval-Augmented Generation (RAG) techniques have gained extensive attention, particularly in vertical domains where LLMs may lack domain-specific knowledge. In this paper, we introduce an omnidirectional and automatic RAG benchmark, OmniEval, in the financial domain. Our benchmark is characterized by its multi-dimensional evaluation framework, including (1) a matrix-based RAG scenario evaluation system that categorizes queries into five task classes and 16 financial topics, leading to a structured assessment of diverse query scenarios; (2) a multi-dimensional evaluation data generation approach, which combines GPT-4-based automatic generation and human annotation, achieving an 87.47\% acceptance ratio in human evaluations on generated instances; (3) a multi-stage evaluation system that evaluates both retrieval and generation performance, result in a comprehensive evaluation on the RAG pipeline; and (4) robust evaluation metrics derived from rule-based and LLM-based ones, enhancing the reliability of assessments through manual annotations and supervised fine-tuning of an LLM evaluator. Our experiments demonstrate the comprehensiveness of OmniEval, which includes extensive test datasets and highlights the performance variations of RAG systems across diverse topics and tasks, revealing significant opportunities for RAG models to improve their capabilities in vertical domains. We open source the code of our benchmark in https://github.com/RUC-NLPIR/OmniEval{https://github.com/RUC-NLPIR/OmniEval}.

In this work, we reframe agentic trajectories as paths through a shared embedding space of actions and environmental observations. We show that successful and failed rollouts of the same task separate in this space, with simple per-task discriminators predicting held-out rollout outcomes at up to 0.95 AUROC across three agent benchmarks and three policy models. To exploit this signal across tasks, we introduce Precog-RM, a transformer-based model that meta-learns a representation in which outcome prediction can be performed in a few-shot setting. We show that Precog-RM generalizes to unseen tasks and models, and gives confident predictions of trajectory success long before termination. Deployed as an early stopping controller for Group Relative Policy Optimization (GRPO) on Tau2-Bench, Precog-RM cuts training FLOPs by up to 47.3\% while maintaining performance comparable to baseline GRPO.

Evaluating large language model (LLM)-based multi-agent systems remains a critical challenge, as these systems must exhibit reliable coordination, transparent decision-making, and verifiable performance across evolving tasks. Existing evaluation approaches often limit themselves to single-response scoring or narrow benchmarks, which lack stability, extensibility, and automation when deployed in enterprise settings at multi-agent scale. We present AEMA (Adaptive Evaluation Multi-Agent), a process-aware and auditable framework that plans, executes, and aggregates multi-step evaluations across heterogeneous agentic workflows under human oversight. Compared to a single LLM-as-a-Judge, AEMA achieves greater stability, human alignment, and traceable records that support accountable automation. Our results on enterprise-style agent workflows simulated using realistic business scenarios demonstrate that AEMA provides a transparent and reproducible pathway toward responsible evaluation of LLM-based multi-agent systems. Keywords Agentic AI, Multi-Agent Systems, Trustworthy AI, Verifiable Evaluation, Human Oversight

Reinforcement learning (RL) has become essential for post-training large language models (LLMs) in reasoning tasks. While scaling rollouts can stabilize training and enhance performance, the computational overhead is a critical issue. In algorithms like GRPO, multiple rollouts per prompt incur prohibitive costs, as a large portion of prompts provide negligible gradients and are thus of low utility. To address this problem, we investigate how to select high-utility prompts before the rollout phase. Our experimental analysis reveals that sample utility is non-uniform and evolving: the strongest learning signals concentrate at the ``learning edge", the intersection of intermediate difficulty and high uncertainty, which shifts as training proceeds. Motivated by this, we propose HIVE (History-Informed and online-VErified prompt selection), a dual-stage framework for data-efficient RL. HIVE utilizes historical reward trajectories for coarse selection and employs prompt entropy as a real-time proxy to prune instances with stale utility. By evaluating HIVE across multiple math reasoning benchmarks and models, we show that HIVE yields significant rollout efficiency without compromising performance.

Reinforcement learning algorithms such as GRPO have driven recent advances in large language model (LLM) reasoning. While scaling the number of rollouts stabilizes training, existing approaches suffer from limited exploration on challenging prompts and leave informative feedback signals underexploited, due to context-independent rollout allocation across prompts (e.g., generating 16 rollouts per prompt) and relying heavily on sparse rewards. This paper presents XRPO(eXplore - eXploit GRPO), a unified framework that recasts policy optimization through the principled lens of rollout exploration-exploitation. To enhance exploration, XRPO introduces a mathematically grounded rollout allocator that adaptively prioritizes prompts with higher potential for uncertainty reduction. It further addresses stagnation on zero-reward prompts through an in-context seeding strategy that injects curated exemplars, steering the model into more difficult reasoning trajectories. To strengthen exploitation, XRPO develops a group-relative, novelty-aware advantage sharpening mechanism that leverages sequence likelihoods to amplify low-probability yet correct responses, thereby extending the policy's reach beyond sparse rewards. Experiments across diverse math and coding benchmarks on both reasoning and non-reasoning models demonstrate that XRPO outperforms existing advances (e.g., GRPO and GSPO) up to 4% pass@1 and 6% cons@32, while accelerating training convergence by up to 2.7X.

Recent progress in deep research systems has been impressive, but evaluation still lags behind real user needs. Existing benchmarks predominantly assess final reports using fixed rubrics, failing to evaluate the underlying research process. Most also offer limited multimodal coverage, rely on synthetic tasks that do not reflect real-world query complexity, and cannot be refreshed as knowledge evolves. To address these gaps, we introduce MiroEval, a benchmark and evaluation framework for deep research systems. The benchmark comprises 100 tasks (70 text-only, 30 multimodal), all grounded in real user needs and constructed via a dual-path pipeline that supports periodic updates, enabling a live and evolving setting. The proposed evaluation suite assesses deep research systems along three complementary dimensions: adaptive synthesis quality evaluation with task-specific rubrics, agentic factuality verification via active retrieval and reasoning over both web sources and multimodal attachments, and process-centric evaluation audits how the system searches, reasons, and refines throughout its investigation. Evaluation across 13 systems yields three principal findings: the three evaluation dimensions capture complementary aspects of system capability, with each revealing distinct strengths and weaknesses across systems; process quality serves as a reliable predictor of overall outcome while revealing weaknesses invisible to output-level metrics; and multimodal tasks pose substantially greater challenges, with most systems declining by 3 to 10 points. The MiroThinker series achieves the most balanced performance, with MiroThinker-H1 ranking the highest overall in both settings. Human verification and robustness results confirm the reliability of the benchmark and evaluation framework. MiroEval provides a holistic diagnostic tool for the next generation of deep research agents.

The advances of large foundation models necessitate wide-coverage, low-cost, and zero-contamination benchmarks. Despite continuous exploration of language model evaluations, comprehensive studies on the evaluation of Large Multi-modal Models (LMMs) remain limited. In this work, we introduce LMMS-EVAL, a unified and standardized multimodal benchmark framework with over 50 tasks and more than 10 models to promote transparent and reproducible evaluations. Although LMMS-EVAL offers comprehensive coverage, we find it still falls short in achieving low cost and zero contamination. To approach this evaluation trilemma, we further introduce LMMS-EVAL LITE, a pruned evaluation toolkit that emphasizes both coverage and efficiency. Additionally, we present Multimodal LIVEBENCH that utilizes continuously updating news and online forums to assess models' generalization abilities in the wild, featuring a low-cost and zero-contamination evaluation approach. In summary, our work highlights the importance of considering the evaluation trilemma and provides practical solutions to navigate the trade-offs in evaluating large multi-modal models, paving the way for more effective and reliable benchmarking of LMMs. We opensource our codebase and maintain leaderboard of LIVEBENCH at https://github.com/EvolvingLMMs-Lab/lmms-eval and https://huggingface.co/spaces/lmms-lab/LiveBench.

Evaluating Large Language Models (LLMs) is a complex task, especially considering the intricacies of natural language understanding and the expectations for high-level reasoning. Traditional evaluations typically lean on human-based, model-based, or automatic-metrics-based paradigms, each with its own advantages and shortcomings. We introduce "Fusion-Eval", a system that employs LLMs not solely for direct evaluations, but to skillfully integrate insights from diverse evaluators. This gives Fusion-Eval flexibility, enabling it to work effectively across diverse tasks and make optimal use of multiple references. In testing on the SummEval dataset, Fusion-Eval achieved a Spearman correlation of 0.96, outperforming other evaluators. The success of Fusion-Eval underscores the potential of LLMs to produce evaluations that closely align human perspectives, setting a new standard in the field of LLM evaluation.

General-purpose agents can plan, reason, and act over long horizons, yet their production capabilities remain fragmented across text, images, audio, video, documents, 3D assets, and code. Extending a foundation model to additional modalities ties capability growth to costly model updates, while assembling specialist models and tools leaves unresolved how procedures, dependencies, intermediate assets, and cross-turn revisions should be coordinated. We present Omni-IO Skills, a plug-and-play Agent Harness that makes existing agents omni-native through hierarchical Skills, a standardized multimodal execution interface, dependency-aware orchestration, and a persistent Asset Registry. Multi-asset workflows are represented as Declare Execution Graphs, which schedule independent operations concurrently and register successful outputs for downstream and cross-turn reuse across replaceable execution backends. Its 27 Skills cover 38 representative tasks spanning seven artifact modalities and four capability families: understanding, generation, reasoning, and retrieval. On UniM-90, the harness raises the input-support rates of GPT-5.6 Sol and Claude Sonnet 5 from 40.00% and 38.89% to 100%, while increasing relative Semantic--Quality Coupled Score from 26.99 to 74.94 and from 27.82 to 77.78, respectively; Strict Structure Score reaches 100.00 and 99.78. These results establish harness-level capability composition as a practical route to broad, evolvable Omni systems without changing the host agent's reasoning core.

Current agentic AI benchmarks predominantly evaluate task completion accuracy, while overlooking critical enterprise requirements such as cost-efficiency, reliability, and operational stability. Through systematic analysis of 12 main benchmarks and empirical evaluation of state-of-the-art agents, we identify three fundamental limitations: (1) absence of cost-controlled evaluation leading to 50x cost variations for similar precision, (2) inadequate reliability assessment where agent performance drops from 60\% (single run) to 25\% (8-run consistency), and (3) missing multidimensional metrics for security, latency, and policy compliance. We propose CLEAR (Cost, Latency, Efficacy, Assurance, Reliability), a holistic evaluation framework specifically designed for enterprise deployment. Evaluation of six leading agents on 300 enterprise tasks demonstrates that optimizing for accuracy alone yields agents 4.4-10.8x more expensive than cost-aware alternatives with comparable performance. Expert evaluation (N=15) confirms that CLEAR better predicts production success (correlation ρ=0.83) compared to accuracy-only evaluation (ρ=0.41).

Reinforcement learning (RL) has emerged as a standard post-training paradigm for shaping large language models (LLMs) into capable agents. In agentic RL, the rollout stage generates trajectories while invoking tools, producing long-tailed and non-stationary workloads that expose two fundamental challenges in resource management. First, due to the long-tail distribution, a small fraction of trajectories dominates rollout makespan. Second, rollout and training are subject to cross-stage imbalance, as they exhibit strong asymmetry in compute patterns, memory demands, and sensitivity to sequence length. Compounding this asymmetry, the sequence length distribution drifts continuously as the policy evolves, rendering any static resource split progressively suboptimal. We present Libra, a resource management system to address both challenges via two core mechanisms. The first is a global resource planner that jointly optimizes GPU allocation across rollout and training clusters. It leverages an elastic hybrid pool to enable lightweight, non-blocking worker reallocation between stages. The second is a causality-driven multi-level feedback queue (C-MLFQ) scheduler, which routes requests to heterogeneous rollout buckets based on causal signals derived from tool-return outcomes, rather than relying on fragile length predictions. Evaluated on 48 A800 GPUs, Libra achieves up to 3.0x higher throughput and converges up to 2.5x faster in reward compared to the baselines.

We introduce TimeSeriesGym, a scalable benchmarking framework for evaluating Artificial Intelligence (AI) agents on time series machine learning engineering challenges. Existing benchmarks lack scalability, focus narrowly on model building in well-defined settings, and evaluate only a limited set of research artifacts (e.g., CSV submission files). To make AI agent benchmarking more relevant to the practice of machine learning engineering, our framework scales along two critical dimensions. First, recognizing that effective ML engineering requires a range of diverse skills, TimeSeriesGym incorporates challenges from diverse sources spanning multiple domains and tasks. We design challenges to evaluate both isolated capabilities (including data handling, understanding research repositories, and code translation) and their combinations, and rather than addressing each challenge independently, we develop tools that support designing multiple challenges at scale. Second, we implement evaluation mechanisms for multiple research artifacts, including submission files, code, and models, using both precise numeric measures and more flexible LLM-based evaluation approaches. This dual strategy balances objective assessment with contextual judgment. Although our initial focus is on time series applications, our framework can be readily extended to other data modalities, broadly enhancing the comprehensiveness and practical utility of agentic AI evaluation. We open-source our benchmarking framework to facilitate future research on the ML engineering capabilities of AI agents.

As LLM agents undertake increasingly complex, long-horizon tasks, verifying their outputs becomes increasingly challenging. We study how verification capability can be strengthened with a fixed base model, without access to reference answers or grading rubrics at test time. Repeated sampling yields multiple rollouts that can contain complementary correct claims, but we need a reliable verification mechanism to determine which claims to trust. We first find that disagreement often exposes correct alternatives, while consensus can conceal errors. These observations motivate VeriHarness, which turns the underlying LLM a generator uses into an agentic verifier by giving it a workspace, evidence tools, and reusable verification skills. A disagreement resolver checks competing claims against environmental evidence, while a consensus challenger tests shared claims and searches for omitted requirements. Their findings guide the selection and revision of the final artifact. Across five long-horizon workspace benchmarks and two frontier models, VeriHarness achieves the highest selection scores among the evaluated baselines. Evidence-backed revision further improves average performance, bringing gains over a single rollout to 6.2 points with Gemini 3.5 Flash and 6.4 points with Claude Opus 4.8. We further show that verification skills can self-improve from failure feedback, demonstrating VeriHarness as a novel and critical approach for scaling long-horizon agentic verification. We release the full pool of approximately 26,000 rollouts across all five benchmarks and both models, produced at a cost of over $100,000, to support future research on agentic verification.

Large language model post-training generates self-generated rollouts through reinforcement learning and on-policy distillation, yet this experience is often treated as stale once the policy advances. Historical rollouts can remain compatible with a later policy while preserving behaviors that the policy no longer expresses reliably. However, they may also contain mistakes, abandoned attempts, and redundant actions that should not be imitated, motivating finer-grained selective supervision. We introduce ROSS (Relearning from Self-Generated Rollouts through Selective Supervision), which preserves the full historical trajectory as context while applying loss only to selected model-generated continuations. Across domain-specific reinforcement learning, multi-teacher on-policy distillation, and agentic reinforcement learning, ROSS consistently improves upstream checkpoints and outperforms baselines across mathematics, code generation, instruction following, and software engineering. On Qwen3.6-35B-A3B, ROSS improves the six-benchmark MOPD average from 58.40% to 62.20% and SWE-bench Verified from 64.20% to 68.40%. These results show that self-rollout training leaves behind reusable behavioral experience that can yield further gains through offline supervised fine-tuning (SFT), without additional policy rollouts.

Authors seeking to communicate with broader audiences often share their ideas in various document formats, such as slide decks, newsletters, reports, and posters. Prior work on document generation has generally tackled the creation of each separate format to be a different task, leading to fragmented learning processes, redundancy in models and methods, and disjointed evaluation. We consider each of these documents as templatic views of the same underlying knowledge/content, and we aim to unify the generation and evaluation of these templatic views. We begin by showing that current LLMs are capable of generating various document formats with little to no supervision. Further, a simple augmentation involving a structured intermediate representation can improve performance, especially for smaller models. We then introduce a novel unified evaluation framework that can be adapted to measuring the quality of document generators for heterogeneous downstream applications. This evaluation is adaptable to a range of user defined criteria and application scenarios, obviating the need for task specific evaluation metrics. Finally, we conduct a human evaluation, which shows that people prefer 82% of the documents generated with our method, while correlating more highly with our unified evaluation framework than prior metrics in the literature.

Graphical User Interface (GUI) agents have demonstrated remarkable progress in automating complex user interface interactions through reinforcement learning. However, current approaches face a fundamental dilemma: offline RL enables stable training on pre-collected trajectories, but struggles with multi-step task execution for lack of trajectory-level reward signals; online RL captures these signals through environment interaction, but suffers from sparse rewards and prohibitive deployment costs. To address it, we present Semi-online Reinforcement Learning, a novel paradigm that simulates online RL on offline trajectories. During each rollout process, we preserve the original model output within the multi-turn dialogue, where a Patch Module adaptively recovers the divergence between rollout and expert trajectories. To capture long-term training signals, Semi-online RL introduces discounted future returns into the reward computation and optimizes the policy with weighted step-level and episode-level advantages. We further introduce Semi-Online Performance (SOP), a metric that aligns better with true online performance, serving as a practical and effective proxy for real-world evaluation. Experiments show that ours Semi-online RL achieves SOTA performance among 7B models across four dynamic benchmarks, with significant gains over the base model (e.g., +12.0% on AndroidWorld, +23.8% on AITW), demonstrating significant progress in bridging the gap between offline training efficiency and online multi-turn reasoning. The code is available at https://github.com/X-PLUG/MobileAgent/tree/main/UI-S1.

Block drafters for speculative decoding are commonly trained on corpora written by external models, where a single off-policy token invalidates supervision for all subsequent slots in a block. Existing approaches discard these divergent slots, resulting in severe supervision loss. To resolve this problem while preserving the training corpus, we propose a rollout-based training framework that recovers full supervision through two complementary components. The first component, Anchor-Label Relabelling (ALR), replaces corpus labels with distributions from greedy target rollouts, restoring valid supervision across all predicted slots. The second component, In-Rollout Anchors (IRA), places draft blocks directly inside these rollouts to expose the drafter to target-generated context, reusing precomputed rollout features at no additional target cost. Across fixed vision-language and text corpora, our framework increases greedy accepted length by up to 36.5% over DFlash and consistently outperforms erasing baselines. Notably, a single epoch of our method surpasses the best erase schedules. After three epochs, it matches the acceptance length of training on target-regenerated responses. These results show that our framework provides an effective and compute-efficient approach for training speculative drafters on fixed corpora without modifying the original text. Code is available at https://github.com/js-lee-AI/ALR-IRA.

Automatic evaluation with large language models, commonly known as LLM-as-a-judge, is now standard across reasoning and alignment tasks. Despite evaluating many samples in deployment, these evaluators typically (i) treat each case independently, missing the opportunity to accumulate experience, and (ii) rely on a single fixed prompt for all cases, neglecting the need for sample-specific evaluation criteria. We introduce Learning While Evaluating (LWE), a framework that allows evaluators to improve sequentially at inference time without requiring training or validation sets. LWE maintains an evolving meta-prompt that (i) produces sample-specific evaluation instructions and (ii) refines itself through self-generated feedback. Furthermore, we propose Selective LWE, which updates the meta-prompt only on self-inconsistent cases, focusing computation where it matters most. This selective approach retains the benefits of sequential learning while being far more cost-effective. Across two pairwise comparison benchmarks, Selective LWE outperforms strong baselines, empirically demonstrating that evaluators can improve during sequential testing with a simple selective update, learning most from the cases they struggle with.

Group-relative policy optimization methods train language models by generating multiple rollouts per prompt and normalizing rewards with a shared mean reward baseline. In resource-constrained settings where the rollout budget is small, accuracy often degrades. We find that noise in the shared baseline induces advantage sign flips, where some rollouts receive an incorrect advantage sign, and the update direction is reversed. To address this, we propose Median-Centered Group Relative Policy Optimization (MC-GRPO), a simple and effective solution for small-rollout training. Our main idea is to replace the mean baseline with a median baseline: the median is far less sensitive to outlier rewards than the mean, mitigating the sign flips under small rollout size (G). We generate one additional rollout for median reference (G+1), and compute advantages by using the group median. With an odd-sized group, exactly one completion is the median and receives zero advantage, we exclude this pivot rollout from backpropagation so the number of gradient-contributing samples per prompt remains G, preserving the core update cost of standard G-rollout training. Across various GRPO-family methods and a wide range of models and scales, this median-centered training consistently improves stability and final accuracy in the low-rollout regime, reducing the gap between G=2 and G=8 to within 1%. Code is available at https://github.com/lotusroot-kim/MC-GRPO

We propose MM-Vet, an evaluation benchmark that examines large multimodal models (LMMs) on complicated multimodal tasks. Recent LMMs have shown various intriguing abilities, such as solving math problems written on the blackboard, reasoning about events and celebrities in news images, and explaining visual jokes. Rapid model advancements pose challenges to evaluation benchmark development. Problems include: (1) How to systematically structure and evaluate the complicated multimodal tasks; (2) How to design evaluation metrics that work well across question and answer types; and (3) How to give model insights beyond a simple performance ranking. To this end, we present MM-Vet, designed based on the insight that the intriguing ability to solve complicated tasks is often achieved by a generalist model being able to integrate different core vision-language (VL) capabilities. MM-Vet defines 6 core VL capabilities and examines the 16 integrations of interest derived from the capability combination. For evaluation metrics, we propose an LLM-based evaluator for open-ended outputs. The evaluator enables the evaluation across different question types and answer styles, resulting in a unified scoring metric. We evaluate representative LMMs on MM-Vet, providing insights into the capabilities of different LMM system paradigms and models. Code and data are available at https://github.com/yuweihao/MM-Vet.

Asynchronous reinforcement learning improves rollout throughput for large language model agents by decoupling sample generation from policy optimization, but it also introduces a critical failure mode for PPO-style off-policy correction. In heterogeneous training systems, the total importance ratio should ideally be decomposed into two semantically distinct factors: a training--inference discrepancy term that aligns inference-side and training-side distributions at the same behavior-policy version, and a policy-staleness term that constrains the update from the historical policy to the current policy. We show that practical asynchronous pipelines with delayed updates and partial rollouts often lose the required historical training-side logits, or old logits. This missing-old-logit problem entangles discrepancy repair with staleness correction, breaks the intended semantics of decoupled correction, and makes clipping and masking thresholds interact undesirably. To address this issue, we study both exact and approximate correction routes. We propose three exact old-logit acquisition strategies: snapshot-based version tracking, a dedicated old-logit model, and synchronization via partial rollout interruption, and compare their system trade-offs. From the perspective of approximate correction, we focus on preserving the benefits of decoupled correction through a more appropriate approximate policy when exact old logits cannot be recovered at low cost, without incurring extra system overhead. Following this analysis, we adopt a revised PPO-EWMA method, which achieves significant gains in both training speed and optimization performance. Code at https://github.com/millioniron/ROLL.

Existing benchmarks for scientific data analysis evaluate LLMs primarily on code execution or workflow completion, overlooking that scientific analysis serves to support distinct types of scientific claims: hypothesis exploration, statistical inference, mechanistic explanation, each with different assumptions and validity criteria. We introduce SDABench, a benchmark that reorganizes evaluation around six capabilities (descriptive, exploratory, inferential, predictive, causal, and mechanistic) across five domains (Biology, Chemistry, Environment, Geography, Physics). SDABench comprises 527 real-data instances (SDA-Real) and 6000 synthetic instances (SDA-Synth), each in both multiple-choice and open-ended formats, constructed through an automated pipeline. Evaluating 15 representative LLMs, we find that models handle descriptive analysis well but degrade sharply on tasks requiring assumption selection, latent-process modeling, or mechanistic reasoning. SDABench further provides a five-stage error analysis framework that locates where LLMs fail: more advanced models more reliably identify the relevant scope and variables, but still struggle to select appropriate analytical procedures, model variable relationships, and draw valid conclusions.

We study auto research as a closed empirical loop driven by external measurement. Each submitted trial carries a hypothesis, an executable code edit, an evaluator-owned outcome, and feedback that shapes the next proposal. The output is not a generated paper or a single model checkpoint, but an auditable trajectory of proposals, code diffs, experiments, scores, and failure labels. We instantiate this loop with specialist agents that partition recipe surfaces and share measured lineage across trials. The central empirical finding is that lineage feedback lets agents turn evaluator outcomes, including crashes, budget overruns, size failures, and accuracy-gate misses, into later program-level recipe edits rather than one-shot suggestions. Across 1,197 headline-run trials plus 600 Parameter Golf control trials after one-time setup and launch, humans did not choose proposals, edit recipes, override scores, or repair failed trials during the search. In the three headline runs, the same submitted-trial loop reduces Parameter Golf validation bpb by 0.81%, raises NanoChat-D12 CORE by 38.7%, and reduces CIFAR-10 Airbench96 wallclock by 4.59%, with each task measured by its own external evaluator and legality checks. The trace includes a strict architecture-domain audit of 157 headline-run submissions and program rewrites such as a NanoChat attention-kernel path change. Within this scope the loop autonomously writes code, submits experiments, absorbs feedback, applies and combines known techniques inside each environment, and improves public starting recipes.

Evaluations should do more than measure a models current performance. They should tell us what to fix for the next model iteration and provide a way to generate targeted post training data. Most evaluation pipelines identify weak examples, topics, or categories, but they leave the underlying capability failure implicit: they say where a model fails, not why. We introduce CRAFT, a method that converts any rubric based evaluation dataset into a model specific diagnosis of weak capabilities. CRAFT treats each grading criterion as a capability probe: it extracts a capability description from every prompt rubric pair, clusters these descriptions into a hierarchical capability tree, scores the target model at every node, and selects low performing nodes dynamically across tree levels, at the granularity where each failure is clearest. The selected weak capabilities then direct the generation of targeted supervised finetuning data. Holding the data generation, finetuning, and evaluation setup fixed, we compare CRAFT against prompt level EvalTree clustering and untargeted random generation on four open source models, two professional domains (finance and legal), and 13 held out benchmarks disjoint from the diagnostic data. CRAFT achieves the strongest finance domain average for all four models under repeated temperature decoding; on legal domain, it is strongest for three of four models and remains within the decoding variance bands of the best baseline on the fourth. Diagnosing weaknesses at the level of rubric criteria, rather than prompts or categories, thus yields both a sharper picture of what a model cannot do and measurably better models after finetuning on that diagnosis.

As executable environments for language agents scale, generating and verifying their rollouts becomes increasingly expensive. Under finite rollout budgets, this raises an important question: can data management at the environment level make agentic reinforcement learning (RL) more efficient by organizing, evaluating, and scheduling task instances together with the environments that determine their tools, states, transitions, and verification? To address this question, we introduce EnvSched (Environment-aware Scheduling), an online scheduler that learns from rollout history. EnvSched assigns scheduling value through two complementary modeling choices: exploitation prioritizes tasks likely to provide informative reward differences, while coverage spreads each batch across underobserved environments and interaction patterns. EnvSched organizes environment--task pairs as nodes in a dual-relation graph, with within-environment edges capturing task semantics and shared tool, state, and verifier structure, and cross-environment edges capturing abstract execution and reward topology. By propagating historical rollout statistics along both edge types, it predicts within-task reward variance. These predictions guide a scheduler that allocates each batch to tasks with high predicted reward variance and underobserved environments and tasks. Environment quotas and a similarity penalty further limit concentration and redundancy, yielding informative and diverse batches without additional pilot rollouts. In RL on Qwen3-8B, EnvSched achieves an average score of 48.81 across BFCL-v3, tau-Bench, and ACEBench-Agent after 40 training steps, exceeding the highest score attained by random task selection after 120 steps. It also improves downstream performance across model sizes at the same training budget.

Evaluating large language models at scale remains a practical bottleneck for many organizations. While existing evaluation frameworks work well for thousands of examples, they struggle when datasets grow to hundreds of thousands or millions of samples. This scale is common when assessing model behavior across diverse domains or conducting comprehensive regression testing. We present Spark-LLM-Eval, a distributed evaluation framework built natively on Apache Spark. The system treats evaluation as a data-parallel problem, partitioningexamplesacrossexecutorsandaggregatingresultswithproperstatistical accounting. Beyond raw throughput, we emphasize statistical rigor: every reported metric includes bootstrap confidence intervals, and model comparisons come with appropriate significance tests (paired t-tests, McNemar's test, or Wilcoxon signed-rank, depending on the metric type). The framework also addresses the cost problem inherent in LLM evaluation through content-addressable response caching backed by Delta Lake, which allows iterating on metric definitions without re-running inference. We describe the system architecture, the statistical methodology, and report benchmark results showing linear scaling with cluster size. The framework and all evaluation code are available as open source.

Skill-augmented reinforcement learning improves language agents by storing reusable procedural knowledge acquired from past experience. Existing methods typically use strong language models to analyze trajectories, generate skills, and update a retrievable skill bank during online training. However, they rarely assess whether a newly generated skill is useful before it is stored and reused. We find that this assumption is unreliable: even skills generated by proprietary frontier LLMs exhibit highly mixed utility, with many providing little benefit or even degrading performance. Once such skills enter the bank, their effects are difficult to identify, because subsequent rollout feedback is delayed and usually reflects the combined effect of multiple retrieved skills rather than the marginal contribution of any individual skill. We propose an online reinforcement learning framework for pre-storage skill validation. The framework estimates whether a candidate skill contributes useful information beyond the skills already retrieved for the current task. It uses the standard rollout budget to form two matched groups under the same task and retrieval context: base rollouts conditioned on the currently retrieved skills, and skill-augmented rollouts conditioned on the same skills plus one candidate skill induced from the base trajectories. The reward gap between these two groups estimates the candidate skill's context-dependent marginal utility, enabling the framework to promote useful skills while filtering ineffective or harmful ones without additional rollout overhead. The framework further uses this marginal-utility signal to train the policy itself as a skill generator, reducing reliance on repeated calls to proprietary models. The learned skill-generation likelihood serves as a context-dependent score for retrieval-time reranking and outdated-skill pruning as the policy evolves.

Jev is a fast, low-cost decision model that answers natural-language questions with choices, binary judgments, and scores. As its public ecosystem grows rapidly, it remains unclear how Jev is used across applications and how public attention relates to project distribution. To answer these questions, we conduct a large-scale, data-driven analysis of 2,170 publicly available Jev projects collected from GitHub as of September 22, 2026. We find rapid early growth in Jev's public ecosystem, with both new projects and integration into existing repositories. Across diverse domains, projects use Jev for multiple decision purposes and combine its interfaces. Attribute judgment and scoring are widely used, while the use of action selection, content filtering, and model and tool selection varies across domains. These patterns suggest that Jev serves as a reusable decision component whose functionality varies with the surrounding workflow. Meanwhile, public attention is concentrated in routing and interface agents and does not track project counts. Our findings provide a quantitative view of Jev's emerging ecosystem and inform the design and evaluation of general-purpose decision models across diverse application contexts.

Large Language Models (LLMs) are being used more and more extensively for automated evaluation in various scenarios. Previous studies have attempted to fine-tune open-source LLMs to replicate the evaluation explanations and judgments of powerful proprietary models, such as GPT-4. However, these methods are largely limited to text-based analyses under predefined general criteria, resulting in reduced adaptability for unseen instructions and demonstrating instability in evaluating adherence to quantitative and structural constraints. To address these limitations, we propose a novel evaluation framework, ARJudge, that adaptively formulates evaluation criteria and synthesizes both text-based and code-driven analyses to evaluate LLM responses. ARJudge consists of two components: a fine-tuned Analyzer that generates multi-faceted evaluation analyses and a tuning-free Refiner that combines and refines all analyses to make the final judgment. We construct a Composite Analysis Corpus that integrates tasks for evaluation criteria generation alongside text-based and code-driven analysis generation to train the Analyzer. Our results demonstrate that ARJudge outperforms existing fine-tuned evaluators in effectiveness and robustness. Furthermore, it demonstrates the importance of multi-faceted evaluation and code-driven analyses in enhancing evaluation capabilities.

Constructing large-scale datasets for the GitHub issue resolution task is crucial for both training and evaluating the software engineering capabilities of Large Language Models (LLMs). However, the traditional process for creating such benchmarks is notoriously challenging and labor-intensive, particularly in the stages of setting up evaluation environments, grading test outcomes, and validating task instances. In this paper, we propose SWE-Factory, an automated pipeline designed to address these challenges. To tackle these issues, our pipeline integrates three core automated components. First, we introduce SWE-Builder, a multi-agent system that automates evaluation environment construction, which employs four specialized agents that work in a collaborative, iterative loop and leverages an environment memory pool to enhance efficiency. Second, we introduce a standardized, exit-code-based grading method that eliminates the need for manually writing custom parsers. Finally, we automate the fail2pass validation process using these reliable exit code signals. Experiments on 671 issues across four programming languages show that our pipeline can effectively construct valid task instances; for example, with GPT-4.1-mini, our SWE-Builder constructs 269 valid instances at 0.045 per instance, while with Gemini-2.5-flash, it achieves comparable performance at the lowest cost of 0.024 per instance. We also demonstrate that our exit-code-based grading achieves 100% accuracy compared to manual inspection, and our automated fail2pass validation reaches a precision of 0.92 and a recall of 1.00. We hope our automated pipeline will accelerate the collection of large-scale, high-quality GitHub issue resolution datasets for both training and evaluation. Our code and datasets are released at https://github.com/DeepSoftwareAnalytics/swe-factory.

Real-world evaluation is the gold standard for robot policies because it tests them against the physical conditions and deployment challenges they are ultimately designed to handle. However, real-world evaluation is also the bottleneck for iterating on robot policies: it is costly, difficult to reproduce, and often too sparse to reliably compare nearby model variants. A straightforward proxy for performance is validation loss on expert demonstrations, but this proxy is often poorly correlated with real-world performance. In this paper, we introduce Critical Interval MSE (CI-MSE), an intuitively simple yet effective offline validation metric. CI-MSE restricts error computation to task-critical segments and pairs it with simple action-alignment procedures that better match rollout-time behavior. Across simulation and real-world experiments, CI-MSE yields a stronger correlation between validation error and rollout performance than raw MSE. Across a wide range of policy checkpoints, CI-MSE achieves a Spearman's rank correlation of -0.87, much closer to the ideal value of -1 than raw MSE's -0.61, demonstrating a significant improvement. We show through sensitivity analysis that our metric is robust to a wide range of hyperparameters. We further study the effectiveness of CI-MSE under evaluation distribution shifts and suggest design boundaries when using this metric. In summary, this paper provides a simple and reliable offline validation tool for accelerating policy iteration. Project webpage: https://ci-mse.github.io/

Computer use agents (CUAs) have advanced rapidly in desktop automation, and a growing number of users deploy CUAs such as OpenClaw on Mac Mini for always-on automation. However, existing benchmarks, including those for macOS, evaluate agents without framework augmentation and rely on binary evaluation. As a result, they fail to capture both the framework capabilities leveraged by modern CUAs and the partial progress on long-horizon, multi-application tasks. We present MacAgentBench, a comprehensive macOS agent benchmark comprising 676 tasks across 25 applications, with nearly 60% involving both GUI and CLI interaction. The benchmark adopts deterministic rule-based evaluation and introduces fine-grained multi-checkpoint scoring with capability annotations for multi-application tasks. Experiments across three frameworks and 16 models show that the best configuration, Claude Opus 4.6 on OpenClaw, attains 73.7% Pass@1, while this advantage is primarily driven by the skill library rather than by framework design. Fine-grained metrics further reveal that models with similar Pass@1 can differ substantially in sub-goal completion. Our code and data are publicly available at https://github.com/JetAstra/MacAgentBench.

Existing browser agent benchmarks face a fundamental trilemma: real-website benchmarks lack reproducibility due to content drift, controlled environments sacrifice realism by omitting real-web noise, and both require costly manual curation that limits scalability. We present WebForge, the first fully automated framework that resolves this trilemma through a four-agent pipeline -- Plan, Generate, Refine, and Validate -- that produces interactive, self-contained web environments end-to-end without human annotation. A seven-dimensional difficulty control framework structures task design along navigation depth, visual complexity, reasoning difficulty, and more, enabling systematic capability profiling beyond single aggregate scores. Using WebForge, we construct WebForge-Bench, a benchmark of 934 tasks spanning 7 domains and 3 difficulty levels. Multi-model experiments show that difficulty stratification effectively differentiates model capabilities, while cross-domain analysis exposes capability biases invisible to aggregate metrics. Together, these results confirm that multi-dimensional evaluation reveals distinct capability profiles that a single aggregate score cannot capture. Code and benchmark are publicly available at https://github.com/yuandaxia2001/WebForge.

Traditional fixed test sets fall short in evaluating open-ended capabilities of foundation models. To address this, we propose ONEBench(OpeN-Ended Benchmarking), a new testing paradigm that consolidates individual evaluation datasets into a unified, ever-expanding sample pool. ONEBench allows users to generate custom, open-ended evaluation benchmarks from this pool, corresponding to specific capabilities of interest. By aggregating samples across test sets, ONEBench enables the assessment of diverse capabilities beyond those covered by the original test sets, while mitigating overfitting and dataset bias. Most importantly, it frames model evaluation as a collective process of selecting and aggregating sample-level tests. The shift from task-specific benchmarks to ONEBench introduces two challenges: (1)heterogeneity and (2)incompleteness. Heterogeneity refers to the aggregation over diverse metrics, while incompleteness describes comparing models evaluated on different data subsets. To address these challenges, we explore algorithms to aggregate sparse measurements into reliable model scores. Our aggregation algorithm ensures identifiability(asymptotically recovering ground-truth scores) and rapid convergence, enabling accurate model ranking with less data. On homogenous datasets, we show our aggregation algorithm provides rankings that highly correlate with those produced by average scores. We also demonstrate robustness to ~95% of measurements missing, reducing evaluation cost by up to 20x with little-to-no change in model rankings. We introduce ONEBench-LLM for language models and ONEBench-LMM for vision-language models, unifying evaluations across these domains. Overall, we present a technique for open-ended evaluation, which can aggregate over incomplete, heterogeneous sample-level measurements to continually grow a benchmark alongside the rapidly developing foundation models.

While vision-language-action (VLA) models have advanced rapidly, running them at every control step remains expensive. Prior work accelerates VLA inference using techniques like action chunking and visual-token pruning, typically evaluating based on latency and average task success. However, acceleration may discard information and break tasks the original policy would solve, a risk hidden by average metrics. Measuring these failures is challenging because action deviations compound over closed-loop trajectories, meaning task failure is only observable across full episodes. We therefore define an acceleration-induced failure via paired rollouts from identical initial conditions, tracking when the reference succeeds but the accelerated policy fails. To manage this, we introduce CARE, an approach for certified accelerator selection. CARE uses paired rollouts on a calibration set to provide finite-sample guarantees that acceleration-induced failure risk stays below a user-specified budget. It deploys the fastest certified candidate, falling back to the reference if none qualify. By relying only on terminal outcomes and measured compute, CARE applies unchanged across diverse acceleration mechanisms, while sequential testing and failure-triggered reference rollouts keep certification affordable. On four LIBERO suites with OpenVLA-OFT, CARE certifies 9.0--10.8times speedups while guaranteeing (at 95% confidence) that at least 85.8% of reference-solved episodes are preserved. Under tight budgets, selectors without guarantees exceed the budget in up to 75% of trials, whereas CARE stays within budget and its sequential form uses 78.9% fewer rollouts than exhaustive evaluation. CARE further generalizes to flow-step reduction for π_{0.5}, and to Qwen3.5-9B and Llama-3.1-8B agents in Crafter.

Comprehensive, unbiased, and comparable evaluation of modern generalist policies is uniquely challenging: existing approaches for robot benchmarking typically rely on heavy standardization, either by specifying fixed evaluation tasks and environments, or by hosting centralized ''robot challenges'', and do not readily scale to evaluating generalist policies across a broad range of tasks and environments. In this work, we propose RoboArena, a new approach for scalable evaluation of generalist robot policies in the real world. Instead of standardizing evaluations around fixed tasks, environments, or locations, we propose to crowd-source evaluations across a distributed network of evaluators. Importantly, evaluators can freely choose the tasks and environments they evaluate on, enabling easy scaling of diversity, but they are required to perform double-blind evaluations over pairs of policies. Then, by aggregating preference feedback from pairwise comparisons across diverse tasks and environments, we can derive a ranking of policies. We instantiate our approach across a network of evaluators at seven academic institutions using the DROID robot platform. Through more than 600 pairwise real-robot evaluation episodes across seven generalist policies, we demonstrate that our crowd-sourced approach can more accurately rank the performance of existing generalist policies than conventional, centralized evaluation approaches, while being more scalable, resilient, and trustworthy. We open our evaluation network to the community and hope that it can enable more accessible comparisons of generalist robot policies.

Reinforcement learning with verifiable rewards has become a standard recipe for improving the reasoning abilities of large language models. Existing algorithms face a tradeoff between computational efficiency and sample efficiency in value estimation and policy learning. We introduce BASIS, a critic-free post-training algorithm designed to address this tradeoff. At each online training step, BASIS samples only one rollout per prompt, but leverages rich information across prompts in the entire batch to improve value function estimation. Our experiments demonstrate that BASIS reduces MSE in value function estimation by 69% compared to REINFORCE++, a representative single-rollout baseline, and achieves lower MSE with one rollout than group mean estimators with 8 rollouts. This improvement in value estimation translates to better policy optimization: using substantially less training time, BASIS achieves performance close to multi-rollout GRPO-type baselines and often outperforms single-rollout REINFORCE-type baselines.

We present Miles v0.1, a full-stack, production-ready system for frontier post-training. Building upon the clean design of slime, Miles designs each stage of the reinforcement-learning (RL) training loop around a single principle: components should be verified, clean, and customizable. With accuracy, efficiency, reliability, and scalability as first-class goals, Miles aims to make frontier-scale RL accessible to researchers and enterprises alike. This report walks through the system end to end: rollout engines built on SGLang, a trainer with a choice of two backends (NVIDIA Megatron-LM and PyTorch FSDP), and three weight-synchronization transports for different deployment topologies. Beyond full-parameter RL, Miles also supports LoRA RL, on-policy distillation, supervised fine-tuning, and true-on-policy rollout-training alignment, and extends the same architecture to diffusion models. We close with an end-to-end case study: fully asynchronous agentic RL on a GLM-5.2 744B-A40B model over terminal-use coding tasks, running on 64 NVIDIA GB300 GPUs with a median step time of 263 seconds over the first 30 measured steps. Miles is open-sourced at https://github.com/radixark/miles, with the project website at https://miles.radixark.com.

Large Language Models (LLMs) based autonomous agents demonstrate multifaceted capabilities to contribute substantially to economic production. However, existing benchmarks remain focused on single agentic capability, failing to capture long-horizon real-world scenarios. Moreover, the reliance on human-in-the-loop feedback for realistic tasks creates a scalability bottleneck, hindering automated rollout collection and evaluation. To bridge this gap, we introduce AgencyBench, a comprehensive benchmark derived from daily AI usage, evaluating 6 core agentic capabilities across 32 real-world scenarios, comprising 138 tasks with specific queries, deliverables, and rubrics. These scenarios require an average of 90 tool calls, 1 million tokens, and hours of execution time to resolve. To enable automated evaluation, we employ a user simulation agent to provide iterative feedback, and a Docker sandbox to conduct visual and functional rubric-based assessment. Experiments reveal that closed-source models significantly outperform open-source models (48.4% vs 32.1%). Further analysis reveals significant disparities across models in resource efficiency, feedback-driven self-correction, and specific tool-use preferences. Finally, we investigate the impact of agentic scaffolds, observing that proprietary models demonstrate superior performance within their native ecosystems (e.g., Claude-4.5-Opus via Claude-Agent-SDK), while open-source models exhibit distinct performance peaks, suggesting potential optimization for specific execution frameworks. AgencyBench serves as a critical testbed for next-generation agents, highlighting the necessity of co-optimizing model architecture with agentic frameworks. We believe this work sheds light on the future direction of autonomous agents, and we release the full benchmark and evaluation toolkit at https://github.com/GAIR-NLP/AgencyBench.

We identify a new dimension for enhancing rollout diversity in Group Relative Policy Optimization (GRPO) for LLMs. While GRPO relies on diverse rollouts, prevailing strategies primarily increase diversity by injecting more token-level randomness, which may introduce step-wise noise and lead to incoherent trajectories. We uncover that smaller models within the same model family inherently exhibit higher policy-level diversity, indicated by their superior pass@k relative to larger counterparts as sample counts increase. Unlike token-level noise, this diversity is temporally correlated, preserves logical consistency, and provides structured exploration signals for gradient estimation. We thus propose S2L-PO (Small-to-Large Policy Optimization), a framework that leverages fixed small models as natural explorers to train larger models. To balance exploration and exploitation, we design a progressive annealing strategy that transitions from offline small-model rollouts to the large learner's own sampling. This shift elegantly avoids mid-training performance drops caused by the small model's capacity limits, achieving faster convergence and unlocking a higher performance ceiling. S2L-PO improves accuracy on diverse mathematical reasoning benchmarks (e.g., +8.8% on AIME 24 using a 1.7B explorer to guide the 8B model) while reducing rollout compute.

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.

The rapid development of large language models and multimodal large language models has accelerated the emergence of proactive agents capable of operating everyday tools and assisting users in real-world environments. However, existing benchmarks struggle to evaluate such agents effectively, as they often rely on sandboxed environments and single-turn evaluation paradigms. Moreover, their scenario-based task taxonomies mix multiple model capabilities within the same task category, making it difficult to identify the root causes of agent failures. To address these limitations, we introduce UniClawBench, the first capability-driven benchmark designed to evaluate proactive agents in dynamic, real-world settings. UniClawBench is built around five foundational model capabilities: Skill Usage, Exploration, Long-Context Reasoning, Multimodal Understanding, and Cross-Platform Coordination. Based on these capabilities, we design 400 bilingual real-world tasks. Unlike previous benchmarks that rely on static, pre-recorded answers, our benchmark evaluates agents in live Docker containers using fine-grained, step-by-step completion checkpoints. Furthermore, we design a closed-loop evaluation strategy comprising an executor agent, a hidden supervisor agent, and a user agent to simulate realistic multi-turn human feedback without leaking grading criteria. To disentangle base model capabilities from framework-level design choices, we evaluate state-of-the-art models under multiple agent frameworks. Through comprehensive comparisons across both models and frameworks, we show how base model capabilities and agent framework designs jointly shape performance in real-world environments. To facilitate future research, we make our benchmark and code publicly available at https://github.com/HKU-MMLab/UniClawBench.

Behavior-cloned diffusion policies are expressive but remain vulnerable to covariate shift: small deviations from demonstrated states can compound into task failure. Existing methods address this either by expanding the training distribution through expert corrections or synthetic augmentation, or by steering a frozen policy at test time with guidance from a learned model. The former can be expensive or assumption-dependent, while the latter discards the corrected trajectories after execution. We introduce ReGuide, a self-improving framework that treats guided rollouts as reusable on-policy recovery data. ReGuide first uses Phase-Conditioned Guidance (PCG) to generate corrective rollouts: it constructs phase-specific latent targets, applies guidance only in the drifted-but-recoverable regime, and guides through the estimated clean action to match the dynamics model's training distribution. Successful guided rollouts are then absorbed back into the policy through ReGuide-FT, which fine-tunes the current checkpoint, or ReGuide-FS, which retrains from scratch on the augmented dataset; the two can also be composed and iterated. On Robomimic Can, Square, Transport, and Tool Hang, ReGuide improves base-policy success by 1.3--7.7times, outperforms LPB in the test-time-only setting, and matched-data ablations show that the gains come from guided recovery data rather than additional rollouts alone.

Video foundation models aim to integrate video understanding, generation, editing, and instruction following within a single framework, making them a central direction for next-generation multimodal systems. However, existing evaluation benchmarks remain fragmented and limited in scope, as they each target a single task, rely on task-specific metrics, and typically use short or simple video clips. As a result, they do not capture the unified capabilities that these models are designed to deliver. To address this gap, we introduce UniVBench, a benchmark purpose-built for evaluating video foundation models across four core abilities: video understanding, video generation, video editing, and a newly proposed task, video reconstruction, which assesses how faithfully a model can reproduce video content it has encountered. Our benchmark substantially expands the complexity of evaluation by incorporating 200 high-quality, diverse and multi-shot videos, each paired with detailed captions, multi-format editing instructions, and reference images. All videos are human-created and carefully validated, offering richer cinematic information than prior benchmarks. In addition, we develop a unified agentic evaluation system (UniV-Eval) that standardizes prompting, instruction parsing, and scoring across all tasks, enabling fair, scalable, and reproducible comparisons of unified video models. By grounding evaluation in instruction-based multi-shot video tasks, UniVBench provides the first framework for measuring the integrated capabilities that video foundation models aim to achieve. Extensive human annotations ensure our evaluation aligns with human judgment, enabling rigorous assessment and accelerating progress toward robust video intelligence.

The development of state-of-the-art systems in different applied areas of machine learning (ML) is driven by benchmarks, which have shaped the paradigm of evaluating generalisation capabilities from multiple perspectives. Although the paradigm is shifting towards more fine-grained evaluation across diverse tasks, the delicate question of how to aggregate the performances has received particular interest in the community. In general, benchmarks follow the unspoken utilitarian principles, where the systems are ranked based on their mean average score over task-specific metrics. Such aggregation procedure has been viewed as a sub-optimal evaluation protocol, which may have created the illusion of progress. This paper proposes Vote'n'Rank, a framework for ranking systems in multi-task benchmarks under the principles of the social choice theory. We demonstrate that our approach can be efficiently utilised to draw new insights on benchmarking in several ML sub-fields and identify the best-performing systems in research and development case studies. The Vote'n'Rank's procedures are more robust than the mean average while being able to handle missing performance scores and determine conditions under which the system becomes the winner.

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

We present RepoST, a scalable method to construct environments that provide execution feedback for repository-level code generation for both training and evaluation. Unlike existing works that aim to build entire repositories for execution, which is challenging for both human and LLMs, we provide execution feedback with sandbox testing, which isolates a given target function and its dependencies to a separate script for testing. Sandbox testing reduces the complexity of external dependencies and enables constructing environments at a large scale. We use our method to construct RepoST-Train, a large-scale train set with 7,415 functions from 832 repositories. Training with the execution feedback provided by RepoST-Train leads to a performance gain of 5.5% Pass@1 on HumanEval and 3.5% Pass@1 on RepoEval. We also build an evaluation dataset, RepoST-Eval, and benchmark 12 code generation models.

Recent efforts have aimed to automate scientific diagram generation from paper content (Lin et al., 2026; Zhu et al., 2026a). However, fully satisfying an author's visual and communicative preferences in a single turn is challenging: in our formative user study (N = 14), all participants requested further revisions after viewing an initial draft, and 86% of them rated the refined diagrams as more satisfactory. Despite the clear demand, the multi-turn workflow remains largely underexplored. To bridge this gap, we present MTPaperBananaBench, a benchmark for multi-turn diagram generation containing 292 images annotated with 3,518 user requirements. To reduce expensive human studies and enable scalable benchmarking, we construct a user simulator that, at each turn, identifies unsatisfied requirements and converts k of them into natural language feedback. Evaluating both requirement satisfaction and overall diagram quality reveals two key failure modes shared across baseline multiturn systems: (1) quality drift, where diagram quality progressively declines over turns, and (2) forgetting, where previously implemented features are lost in subsequent turns. To address these issues, we introduce PaperBanana-Interact, a multi-agent system that refines diagrams via an internal critique-and-refine loop. PaperBanana-Interact consistently improves rather than degrades diagram quality across turns, outperforming baselines by 11.9-18.6 points in quality score and reducing forgetting by 3.7-6.2 points.

Large multimodal models (LMMs) are increasingly adopted as judges in multimodal evaluation systems due to their strong instruction following and consistency with human preferences. However, their ability to follow diverse, fine-grained evaluation criteria remains underexplored. We develop Multi-Crit, a benchmark for evaluating multimodal judges on their capacity to follow pluralistic criteria and produce reliable criterion-level judgments. Covering both open-ended generation and verifiable reasoning tasks, Multi-Crit is built through a rigorous data curation pipeline that gathers challenging response pairs with multi-criterion human annotations. It further introduces three novel metrics for systematically assessing pluralistic adherence, criterion-switching flexibility, and the ability to recognize criterion-level preference conflicts. Comprehensive analysis of 25 LMMs reveals that 1) proprietary models still struggle to maintain consistent adherence to pluralistic criteria--especially in open-ended evaluation; 2) open-source models lag further behind in flexibly following diverse criteria; and 3) critic fine-tuning with holistic judgment signals enhances visual grounding but fails to generalize to pluralistic criterion-level judgment. Additional analyses on reasoning fine-tuning, test-time scaling, and boundary consistency between open-source and proprietary models further probe the limits of current multimodal judges. As a pioneering study, Multi-Crit lays the foundation for building reliable and steerable multimodal AI evaluation.

Reinforcement Learning with Verifiable Rewards (RLVR) has significantly advanced the reasoning capabilities of Large Language Models (LLMs). However, methods such as GRPO and DAPO suffer from substantial computational cost, since they rely on sampling many rollouts for each prompt. Moreover, in RLVR the relative advantage is often sparse: many samples become nearly all-correct or all-incorrect, yielding low within-group reward variance and thus weak learning signals. In this paper, we introduce arrol (Accelerating RLVR via online Rollout Pruning), an online rollout pruning method that prunes rollouts during generation while explicitly steering the surviving ones more correctness-balanced to enhance learning signals. Specifically, arrol trains a lightweight quality head on-the-fly to predict the success probability of partial rollouts and uses it to make early pruning decisions. The learned quality head can further weigh candidates to improve inference accuracy during test-time scaling. To improve efficiency, we present a system design that prunes rollouts inside the inference engine and re-batches the remaining ones for log-probability computation and policy updates. Across GRPO and DAPO on Qwen-3 and LLaMA-3.2 models (1B-8B), arrol improves average accuracy by +2.30 to +2.99 while achieving up to 1.7x training speedup, and yielding up to +8.33 additional gains in average accuracy in test-time scaling. The code is available at https://github.com/Hsu1023/ARRoL.

Establishing fair and robust benchmarks is essential for evaluating intelligent code generation by large language models (LLMs). Our survey of 35 existing benchmarks uncovers three major imbalances: 85.7% focus on a single programming language; 94.3% target only function-level or statement-level tasks; and over 80% include fewer than ten test cases on average. To address these gaps, we propose MultiOOP, a multi-language object-oriented programming benchmark covering six popular languages (Python, PHP, C++, C#, Java, JavaScript) with 267 tasks per language. We design a translator that extends an existing single-language OOP benchmark and the pass@o metric to a multilingual setting. Moreover, we propose an automated framework for augmenting test cases to ensure the reliability of the evaluation results. We evaluate 14 mainstream LLMs under zero-shot prompting and report three key findings: 1) Substantial performance degradation: pass@1 scores on MultiOOP drop by up to 65.6 percentage points compared to function-level tasks (e.g., HumanEval). 2) Cross-language variability: GPT-4o mini achieves pass@1 of 48.06% in Python but only 0.12%-15.26% in other languages, indicating limited multilingual generalization. 3) Conceptual gaps: pass@o scores are consistently 1.1-19.2 points lower than pass@k, demonstrating that LLMs often generate executable code without fully capturing core OOP concepts. Our benchmark, metric extensions, and evaluation scripts will be publicly released to foster a more balanced and comprehensive assessment of LLMs in object-oriented code generation. Our code and data will be released at https://github.com/alphadl/OOP-eval and https://huggingface.co/datasets/codeai-dteam/MultiOOP respectively.

This paper introduces LegalRikai: Open Benchmark, a new benchmark comprising four complex tasks that emulate Japanese corporate legal practices. The benchmark was created by legal professionals under the supervision of an attorney. This benchmark has 100 samples that require long-form, structured outputs, and we evaluated them against multiple practical criteria. We conducted both human and automated evaluations using leading LLMs, including GPT-5, Gemini 2.5 Pro, and Claude Opus 4.1. Our human evaluation revealed that abstract instructions prompted unnecessary modifications, highlighting model weaknesses in document-level editing that were missed by conventional short-text tasks. Furthermore, our analysis reveals that automated evaluation aligns well with human judgment on criteria with clear linguistic grounding, and assessing structural consistency remains a challenge. The result demonstrates the utility of automated evaluation as a screening tool when expert availability is limited. We propose a dataset evaluation framework to promote more practice-oriented research in the legal domain.

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

The rapid deployment of AI agents in commercial settings has outpaced the development of evaluation methodologies that reflect production realities. Existing benchmarks measure agent capabilities through retrospectively curated tasks with well-specified requirements and deterministic metrics -- conditions that diverge fundamentally from production environments where requirements contain implicit constraints, inputs are heterogeneous multi-modal documents with information fragmented across sources, tasks demand undeclared domain expertise, outputs are long-horizon professional deliverables, and success is judged by domain experts whose standards evolve over time. We present AlphaEval, a production-grounded benchmark of 94 tasks sourced from seven companies deploying AI agents in their core business, spanning six O*NET (Occupational Information Network) domains. Unlike model-centric benchmarks, AlphaEval evaluates complete agent products -- Claude Code, Codex, etc. -- as commercial systems, capturing performance variations invisible to model-level evaluation. Our evaluation framework covers multiple paradigms (LLM-as-a-Judge, reference-driven metrics, formal verification, rubric-based assessment, automated UI testing, etc.), with individual domains composing multiple paradigms. Beyond the benchmark itself, we contribute a requirement-to-benchmark construction framework -- a systematic methodology that transforms authentic production requirements into executable evaluation tasks in minimal time. This framework standardizes the entire pipeline from requirement to evaluation, providing a reproducible, modular process that any organization can adopt to construct production-grounded benchmarks for their own domains.

The large-scale training of multi-modal models on data scraped from the web has shown outstanding utility in infusing these models with the required world knowledge to perform effectively on multiple downstream tasks. However, one downside of scraping data from the web can be the potential sacrifice of the benchmarks on which the abilities of these models are often evaluated. To safeguard against test data contamination and to truly test the abilities of these foundation models we propose LiveXiv: A scalable evolving live benchmark based on scientific ArXiv papers. LiveXiv accesses domain-specific manuscripts at any given timestamp and proposes to automatically generate visual question-answer pairs (VQA). This is done without any human-in-the-loop, using the multi-modal content in the manuscripts, like graphs, charts, and tables. Moreover, we introduce an efficient evaluation approach that estimates the performance of all models on the evolving benchmark using evaluations of only a subset of models. This significantly reduces the overall evaluation cost. We benchmark multiple open and proprietary Large Multi-modal Models (LMMs) on the first version of our benchmark, showing its challenging nature and exposing the models true abilities, avoiding contamination. Lastly, in our commitment to high quality, we have collected and evaluated a manually verified subset. By comparing its overall results to our automatic annotations, we have found that the performance variance is indeed minimal (<2.5%). Our dataset is available online on HuggingFace, and our code will be available here.

As LLM-based agents continue to advance, their evaluation has become increasingly multifaceted: a capable agent must not only achieve high task completion accuracy but also perform well in interaction quality, safety, and efficiency, raising a central question: can these objectives be optimized simultaneously? Existing methods have considered multiple objectives, but many collapse heterogeneous measurements into a fixed scalar score. Such scalarization depends on metric normalization and preference weights and may discard candidates that represent useful deployment trade-offs. We introduce Multi-Objective Agent Evolution (MOAE), which organizes iterative in-context refinement as a Pareto-preserving evolutionary search over complete agent rollouts. Given a limited rollout budget, MOAE maintains an empirical archive of non-dominated candidates, uses objective-specific diagnostics to guide offspring generation, and applies constraint-aware selection only at deployment. This separates candidate preservation during search from the preference used to return a final solution. The procedure requires no parameter updates and allows each objective to be replaced by any measurable property, which we instantiate as task performance, trajectory quality, and safety. Experiments on TravelPlanner and AgentDojo show that MOAE consistently improves task performance and trajectory quality while maintaining strong safety under matched rollout budgets. Search-behavior analysis further shows that Pareto preservation expands the attainable objective region and increases the frequency of joint improvement. These results demonstrate the potential of Pareto-preserving in-context evolution for optimizing multiple agent properties without committing to a fixed scalarization during search.

Model merging combines multiple fine-tuned checkpoints into a single model without additional training, offering an attractive approach to reusing models and efficiently improving performance. However, it remains unclear whether the advantages reported for settings where all merged experts have distinct roles and are tuned on clearly separated tasks also hold in settings where the merged experts do not have clearly distinct roles, but are trained on overlapping or even conflicting objectives. To evaluate this setting, we present a large-scale, systematic evaluation of "in-the-wild" model merging of heterogeneous experts, that may have been trained on overlapping or conflicting objectives. Concretely, we evaluate six state-of-the-art merging methods, including recent subspace methods, across four open-weight LLMs, twelve fine-tuned checkpoints per base model, and sixteen standard LLM benchmarks. Evaluating through standardized benchmarks, we measure both the probability that a model merged from a heterogeneous set of experts outperforms the base model and we measure relative gains over the best individual checkpoint. Our results show that the oldest and simplest method, Task Arithmetic, is the only approach that reliably yields performance gains on LLMs in this "in-the-wild" setting. Other interference-aware and subspace merging methods typically do not result in notable improvements over the base model. Our findings indicate that current merging techniques mostly do not enable extracting useful weight updates from heterogeneous and potentially conflicting versions. This motivates the design of LLM-specific merging algorithms and merging-aware fine-tuning methods.

The emergence of long-context language models with context windows extending to millions of tokens has created new opportunities for sophisticated code understanding and software development evaluation. We propose LoCoBench, a comprehensive benchmark specifically designed to evaluate long-context LLMs in realistic, complex software development scenarios. Unlike existing code evaluation benchmarks that focus on single-function completion or short-context tasks, LoCoBench addresses the critical evaluation gap for long-context capabilities that require understanding entire codebases, reasoning across multiple files, and maintaining architectural consistency across large-scale software systems. Our benchmark provides 8,000 evaluation scenarios systematically generated across 10 programming languages, with context lengths spanning 10K to 1M tokens, a 100x variation that enables precise assessment of long-context performance degradation in realistic software development settings. LoCoBench introduces 8 task categories that capture essential long-context capabilities: architectural understanding, cross-file refactoring, multi-session development, bug investigation, feature implementation, code comprehension, integration testing, and security analysis. Through a 5-phase pipeline, we create diverse, high-quality scenarios that challenge LLMs to reason about complex codebases at unprecedented scale. We introduce a comprehensive evaluation framework with 17 metrics across 4 dimensions, including 8 new evaluation metrics, combined in a LoCoBench Score (LCBS). Our evaluation of state-of-the-art long-context models reveals substantial performance gaps, demonstrating that long-context understanding in complex software development represents a significant unsolved challenge that demands more attention. LoCoBench is released at: https://github.com/SalesforceAIResearch/LoCoBench.

We show that domain-general automatic evaluators can significantly improve the performance of agents for web navigation and device control. We experiment with multiple evaluation models that trade off between inference cost, modularity of design, and accuracy. We validate the performance of these models in several popular benchmarks for digital agents, finding between 74.4 and 92.9% agreement with oracle evaluation metrics. Finally, we use these evaluators to improve the performance of existing agents via fine-tuning and inference-time guidance. Without any additional supervision, we improve state-of-the-art performance by 29% on the popular benchmark WebArena, and achieve a 75% relative improvement in a challenging domain transfer scenario.

As language models become increasingly capable, users expect them to provide not only accurate responses but also behaviors aligned with diverse human preferences across a variety of scenarios. To achieve this, Reinforcement learning (RL) pipelines have begun incorporating multiple rewards, each capturing a distinct preference, to guide models toward these desired behaviors. However, recent work has defaulted to apply Group Relative Policy Optimization (GRPO) under multi-reward setting without examining its suitability. In this paper, we demonstrate that directly applying GRPO to normalize distinct rollout reward combinations causes them to collapse into identical advantage values, reducing the resolution of the training signal and resulting in suboptimal convergence and, in some cases, early training failure. We then introduce Group reward-Decoupled Normalization Policy Optimization (GDPO), a new policy optimization method to resolve these issues by decoupling the normalization of individual rewards, more faithfully preserving their relative differences and enabling more accurate multi-reward optimization, along with substantially improved training stability. We compare GDPO with GRPO across three tasks: tool calling, math reasoning, and coding reasoning, evaluating both correctness metrics (accuracy, bug ratio) and constraint adherence metrics (format, length). Across all settings, GDPO consistently outperforms GRPO, demonstrating its effectiveness and generalizability for multi-reward reinforcement learning optimization.

We study whether demonstration-curation metrics that detect defective training episodes also improve the downstream behavior-cloning policy that trains on the curated data. On a contact-rich LIBERO pick-and-place benchmark with a controlled structural defect (early gripper release during the carry phase), we find that the two quantities are sharply decoupled. The metric with the highest defect-detection AUROC (0.804) produces the worst curated policy (13.3% task success), while a metric with a substantially lower AUROC (0.638) produces a policy that nearly matches the oracle trained on ground-truth clean data (90.0% vs. 93.3%). We further show that five of the seven metrics we evaluate exploit episode length as a trivial proxy for the defect label, a confound that inflates reported AUROCs to near-perfect values and disappears once episode length is controlled. Across all conditions, the contaminated baseline succeeds on only 3.3% of rollouts, and the two best curation methods close this to within 3 percentage points of the 93.3% oracle ceiling. Our results argue that curation methods should be evaluated by the policy they produce, not the defects they flag, and that any curation benchmark must control for episode length before reporting detection accuracy. We release the testbed, all metric implementations, and the evaluation pipeline.

Test-time scaling has become a powerful way to improve large language models. However, existing methods are best suited to short, bounded outputs that can be directly compared, ranked or refined. Long-horizon coding agents violate this premise: each attempt produces an extended trajectory of actions, observations, errors, and partial progress taken by the agent. In this setting, the main challenge is no longer generating more attempts, but representing prior experience in a form that can be effectively selected from and reused. We propose a test-time scaling framework for agentic coding based on compact representations of rollout trajectories. Our framework converts each rollout into a structured summary that preserves its salient hypotheses, progress, and failure modes while discarding low-signal trace details. This representation enables two complementary forms of inference-time scaling. For parallel scaling, we introduce Recursive Tournament Voting (RTV), which recursively narrows a population of rollout summaries through small-group comparisons. For sequential scaling, we adapt Parallel-Distill-Refine (PDR) to the agentic setting by conditioning new rollouts on summaries distilled from prior attempts. Our method consistently improves the performance of frontier coding agents across SWE-Bench Verified and Terminal-Bench v2.0. For example, by using our method Claude-4.5-Opus improves from 70.9% to 77.6% on SWE-Bench Verified (mini-SWE-agent) and 46.9% to 59.1% on Terminal-Bench v2.0 (Terminus 1). Our results suggest that test-time scaling for long-horizon agents is fundamentally a problem of representation, selection, and reuse.

Reinforcement Learning (RL) has become critical for advancing modern Large Language Models (LLMs), yet existing synchronous RL systems face severe performance bottlenecks. The rollout phase, which dominates end-to-end iteration time, suffers from substantial long-tail latency and poor resource utilization due to inherent workload imbalance. We present Seer, a novel online context learning system that addresses these challenges by exploiting previously overlooked similarities in output lengths and generation patterns among requests sharing the same prompt. Seer introduces three key techniques: divided rollout for dynamic load balancing, context-aware scheduling, and adaptive grouped speculative decoding. Together, these mechanisms substantially reduce long-tail latency and improve resource efficiency during rollout. Evaluations on production-grade RL workloads demonstrate that Seer improves end-to-end rollout throughput by 74% to 97% and reduces long-tail latency by 75% to 93% compared to state-of-the-art synchronous RL systems, significantly accelerating RL training iterations.

Confidence-based voting aggregates parallel LLM rollouts by weighting each with internal signals such as token log probabilities, and has been actively studied for single-turn reasoning. However, modern LLMs increasingly act as multi-turn search agents that retrieve and condition on external documents. In this paper, we show that confidence-based voting transfers poorly to this multi-turn setting, and identify the underlying failure reason as copy inflation: when retrieved documents are appended to an agent's context, tokens copied from those documents receive systematically inflated log probabilities. This flattens confidence scores within each question and weakens the resulting weighted vote. To address this issue, we propose Retrieval-Grounded Voting (RGV), which scores each rollout by the lexical overlap between its final answer and the documents it retrieved. By computing the signal outside the contaminated context, RGV sidesteps both token log probabilities and additional LLM calls. Across four search-agent benchmarks and five LLMs, RGV consistently outperforms confidence-based voting, with gains of up to +5.4% accuracy and +35% on minority-correct questions, where the correct answer appears in only 1-2 of 8 rollouts.

Automated issue solving aims to resolve real-world issues in software repositories. The most popular benchmarks for automated issue solving are SWE-bench and its human-filtered subset SWE-bench Verified. These benchmarks leverage testing to validate generated patches. However, because testing is rarely exhaustive, a patch may pass the tests but nevertheless fail to match the developers' expectations. Unfortunately, it is currently unclear to what extent evaluations performed with SWE-bench suffer from such plausible but incorrect patches. This paper presents an in-depth empirical study of the correctness of plausible patches generated by three state-of-the-art issue-solving tools evaluated on SWE-bench Verified. We extensively test and inspect generated patches, and compare them against human-written ground truth patches. The core of our methodology is a novel technique PatchDiff for differential patch testing, which automatically exposes behavioral discrepancies between two patches. Our findings reveal critical weaknesses in SWE-bench's patch validation mechanism, which causes 7.8% of all patches to count as correct while failing the developer-written test suite. Moreover, our novel automated technique reveals that even more (29.6%) plausible patches induce different behavior than the ground truth patches. These behavioral differences are often due to similar, but divergent implementations (46.8%) and due to generated patches that adapt more behavior than the ground truth patches (27.3%). Our manual inspection shows that 28.6% of behaviorally divergent patches are certainly incorrect. Combined, the different weaknesses lead to an inflation of reported resolution rates by 6.2 absolute percent points. Our findings are a call to arms for more robust and reliable evaluation of issue-solving tools. We envision our automated differential patch testing technique to be useful for this purpose.

Reinforcement learning (RL) has become a cornerstone in advancing large-scale pre-trained language models (LLMs). Successive generations, including GPT-o series, DeepSeek-R1, Kimi-K1.5, Grok 4, and GLM-4.5, have relied on large-scale RL training to enhance reasoning and coding capabilities. To meet the community's growing RL needs, numerous RL frameworks have been proposed. However, RL training remains computationally expensive, with rollout generation accounting for more than 90% of total runtime. In addition, its efficiency is often constrained by the long-tail distribution of rollout response lengths, where a few lengthy responses stall entire batches, leaving GPUs idle and underutilized. As model and rollout sizes continue to grow, this bottleneck increasingly limits scalability. To address this challenge, we propose Active Partial Rollouts in Reinforcement Learning (APRIL), which mitigates long-tail inefficiency. In the rollout phase, APRIL over-provisions rollout requests, terminates once the target number of responses is reached, and recycles incomplete responses for continuation in future steps. This strategy ensures that no rollouts are discarded while substantially reducing GPU idle time. Experiments show that APRIL improves rollout throughput by 22.5% on average (at most 44%) across commonly used RL algorithms (GRPO, DAPO, GSPO), accelerates convergence, and achieves 2.1% on average(at most 8%) higher final accuracy across tasks. Moreover, APRIL is both framework and hardware agnostic, already integrated into the slime RL framework, and deployable on NVIDIA and AMD GPUs alike. Taken together, this work unifies system-level and algorithmic considerations in proposing APRIL, with the aim of advancing RL training efficiency and inspiring further optimizations in RL systems. Our codebase is available at https://github.com/RLsys-Foundation/APRIL

When a professional web developer's code fails a test, they do not simply re-read the stack trace. They open the application in a browser, click buttons, inspect computed styles, and run diagnostic commands to understand what went wrong. Existing feedback mechanisms for coding agents rely on screenshots, LLM-as-a-judge scoring, or natural-language corrections, but few interact with the live application the way a developer would. We introduce WatchPoint, a simulated-user system that mimics real developer behavior by generating and executing diagnostic scripts against the running application, producing structured observations that guide the coding model's retry. Unlike prior approaches that target single-file edits or evaluate using non-executable metrics, we operate on Web-Bench, a benchmark of 50 multi-file web projects comprising 1,000 sequentially dependent tasks, verified by deterministic end-to-end tests. WatchPoint recovers 57.6% of the tasks it diagnoses, and a controlled user study confirms the simulation's realism: human testers achieve a comparable recovery rate (54.5%), providing evidence that automated diagnostic scripts can substitute for interactive human testing on sequential web development tasks. We further identify a pattern of capability gaps that governs when simulated-user feedback is helpful and when it should be withheld.

This report evaluates PDF-to-Markdown conversion using recent Vision-Language Models (VLMs) on challenging French documents. Document parsing is a critical step for Retrieval-Augmented Generation (RAG) pipelines, where transcription and layout errors propagate to downstream retrieval and grounding. Existing benchmarks often emphasize English or Chinese and can over-penalize benign formatting and linearization choices (e.g., line breaks, list segmentation, alternative table renderings) that are largely irrelevant for downstream use. We introduce a French-focused benchmark of difficult pages selected via model-disagreement sampling from a corpus of 60{,}000 documents, covering handwritten forms, complex layouts, dense tables, and graphics-rich pages. Evaluation is performed with unit-test-style checks that target concrete failure modes (text presence, reading order, and local table constraints) combined with category-specific normalization designed to discount presentation-only variance. Across 15 models, we observe substantially higher robustness for the strongest proprietary models on handwriting and forms, while several open-weights systems remain competitive on standard printed layouts.