Daily Papers

While reasoning models (e.g., DeepSeek R1) trained with reinforcement learning (RL), excel in textual reasoning, they struggle in scenarios requiring structured problem-solving, such as geometric reasoning, concise computation, or complex equation solving-areas where computational tools like code interpreters (CI) demonstrate distinct advantages. To bridge this gap, we propose ReTool, which enhances long-form reasoning with tool-integrated learning, including two key features: (1) dynamic interleaving of real-time code execution within natural language reasoning processes, and (2) an automated RL paradigm that allows policy rollouts with multi-turn real-time code execution and teaches the model in learning when and how to invoke tools based on outcome feedback. ReTool employs a systematic training framework, beginning with synthetic cold-start data generation to produce code-augmented long-form reasoning traces for fine-tuning base models. Subsequent RL training leverages task outcomes as rewards to iteratively refine the model's tool use strategy, enabling autonomous discovery of optimal tool invocation patterns without human priors. Experiments on the challenging MATH Olympiad benchmark AIME demonstrate ReTool's superiority: Our 32B model achieves 67% accuracy with 400 training steps, outperforming text-based RL baseline (40% accuracy, 1080 steps) in efficiency and performance. Remarkably, ReTool-32B attains 72.5% accuracy in extended settings, surpassing OpenAI's o1-preview by 27.9%. Further analysis reveals emergent behaviors such as code self-correction, signaling an ''aha moment'' in which the model autonomously masters adaptive tool use. These findings highlight the promise of outcome-driven tool integration for advancing complex mathematical reasoning and offer new insights into hybrid neuro-symbolic systems.

General-purpose agents increasingly write code, use tools, and complete complex digital tasks, raising the question of how far these capabilities carry into the physical world. To investigate this, we introduce RobotWorld, a challenging simulation testbed for robot use: turning instructions and observations into physical task execution through robot interfaces. Its 84 tasks span manipulation, mobile manipulation, locomotion, driving, and aerial control, with explicit interaction budgets and executable success checks. By analysing task outcomes alongside execution traces, we identify both the capabilities that transfer and the gaps that prevent reliable completion. Furthermore, we find that current agents can construct sophisticated perception and control workflows, including image segmentation, camera calibration, spatial estimation, and dynamics-based computation. These capabilities, however, do not consistently compose into successful behaviour: agents lose task-relevant object states despite reaching commanded poses, fail to correct ineffective actions, recover too late, or mistake unfinished tasks for completion. This uneven transfer also differs across models: Astra succeeds more often on spatial and constrained-contact goals, whereas Opus 5.5 succeeds more often on continuous-balance and timed-interaction goals. By linking these outcomes to execution behaviour, RobotWorld provides both a rigorous proving ground and an empirical account of the remaining capability gaps, thereby establishing concrete targets for training and designing more reliable physical-world agents.

Agent skill evolution seeks to improve reusable procedural guidance for large language model (LLM) agents through iterative revision. Existing methods base each revision mainly on execution trajectories or feedback, leaving recurring behavioral requirements across tasks implicit and tying revision to the behavior of the current skill. We introduce SkillFocus, which decomposes recurring task requirements into a capability space that remains fixed as the skill evolves, separating what tasks require from how the current skill behaves. SkillFocus maps current task outcomes to this space to identify the capability that leaves the most tasks unresolved, then uses that capability to determine what to revise and which evidence to use. Across four benchmarks spanning heterogeneous tasks, SkillFocus achieves the best held-out accuracy on all four, outperforming the strongest competing result by 5.7 points on average while using 24\% fewer evolution tokens on average than the closest iterative baseline. Controlled studies further show that capabilities derived from recurring task requirements outperform task-semantic and execution-derived alternatives, while randomizing task--capability assignments reduces final accuracy by up to 20.2 points. Matching evidence to the selected capability increases candidate gain by 4.4 points under prioritized revision.

Although large language models (LLMs) have demonstrated impressive potential on simple tasks, their breadth of scope, lack of transparency, and insufficient controllability can make them less effective when assisting humans on more complex tasks. In response, we introduce the concept of Chaining LLM steps together, where the output of one step becomes the input for the next, thus aggregating the gains per step. We first define a set of LLM primitive operations useful for Chain construction, then present an interactive system where users can modify these Chains, along with their intermediate results, in a modular way. In a 20-person user study, we found that Chaining not only improved the quality of task outcomes, but also significantly enhanced system transparency, controllability, and sense of collaboration. Additionally, we saw that users developed new ways of interacting with LLMs through Chains: they leveraged sub-tasks to calibrate model expectations, compared and contrasted alternative strategies by observing parallel downstream effects, and debugged unexpected model outputs by "unit-testing" sub-components of a Chain. In two case studies, we further explore how LLM Chains may be used in future applications

Real-world user requests to LLM agents are often underspecified. Agents must interact to acquire missing information and make correct downstream decisions. However, current multi-turn GRPO-based methods often rely on trajectory-level reward computation, which leads to credit assignment problems and insufficient advantage signals within rollout groups. A feasible approach is to identify valuable interaction turns at a fine granularity to drive more targeted learning. To address this, we introduce InfoPO (Information-Driven Policy Optimization), which frames multi-turn interaction as a process of active uncertainty reduction and computes an information-gain reward that credits turns whose feedback measurably changes the agent's subsequent action distribution compared to a masked-feedback counterfactual. It then combines this signal with task outcomes via an adaptive variance-gated fusion to identify information importance while maintaining task-oriented goal direction. Across diverse tasks, including intent clarification, collaborative coding, and tool-augmented decision making, InfoPO consistently outperforms prompting and multi-turn RL baselines. It also demonstrates robustness under user simulator shifts and generalizes effectively to environment-interactive tasks. Overall, InfoPO provides a principled and scalable mechanism for optimizing complex agent-user collaboration. Code is available at https://github.com/kfq20/InfoPO.

Existing benchmarks for LLM agents' social behavior typically focus on a single capability dimension and evaluate only behavioral outcomes, overlooking process signals from reasoning and communication. We present M3-BENCH, a benchmark of 24 mixed-motive games with a process-aware evaluation framework spanning three complementary views: Behavioral Trajectory Analysis (BTA), Reasoning Process Analysis (RPA), and Communication Content Analysis (CCA). Evaluating 11 frontier LLMs and a human baseline, M3-BENCH reveals substantial differences in social competence that outcome-only evaluation misses. In particular, we identify an "overthink-undercommunicate" pattern: reasoning models achieve strong internal deliberation scores but often fail to translate them into effective social communication. Although top models can surpass humans on task outcomes, humans exhibit markedly higher cross-view consistency, suggesting that current LLM agents still lack the behavioral coherence characteristic of human social competence. Our analysis further shows that the three-view decomposition surfaces safety-relevant risks, such as cooperative behavior paired with latent opportunistic reasoning, that remain hidden under outcome-only metrics.

Reinforcement learning (RL) enables generalist robot policies to improve through trial-and-error interaction, yet its effectiveness is fundamentally constrained by sparse task rewards. Existing general-purpose reward models typically alleviate this issue by learning task progress from expert demonstrations, but introduce a distribution mismatch with the mixed-quality rollouts encountered during policy optimization, making their estimates unreliable on suboptimal and failed behaviors from which the policy must learn. In this work, we introduce a demonstration-free reward learning paradigm where dense reward feedback can be learned directly from sparse task outcomes and policy experience. We theoretically show that terminal task outcomes implicitly define dense success-probability feedback at intermediate timesteps, which can be recursively learned through bootstrapping. Based on this insight, we introduce eVTA_0, which learns success probabilities from mixed-quality policy rollouts through temporal-difference-style bootstrapping, without expert demonstrations or intermediate annotations. We further introduce RL with Evolving Rewards (RLER), a closed-loop framework that adapts eVTA_0 using newly collected rollouts as the policy evolves. Experiments show that eVTA_0 provides more informative rewards than state-of-the-art reward models and achieves the best average policy performance across all LIBERO task suites under the same RL training budget, improving success rates by 5.4%-13.8% over the initial policy. In real-world manipulation, RLER further improves overall success rates by 20%-26%, with 35%-36% gains under out-of-distribution conditions. These results demonstrate the effectiveness of demonstration-free reward learning and adapting rewards as the policy evolves. Project webpage: https://duowuyms.github.io/evta0.

Effective collaboration among large language model (LLM) agents in multi-agent systems requires sharing relevant evidence, requesting missing information, and incorporating teammates' contributions into decisions. Training these behaviors is challenging because final task outcomes provide limited feedback on the value of individual messages. We introduce CoSIR (Collaboration via Surprisal-based Information Rewards), a post-training approach that uses a simple, verifiable environment to teach collaboration behaviors that generalize to more complex, unseen tasks. We propose a Bayesian evaluator that scores shared information by its reduction in team surprisal, defined as the sum of the negative log probabilities assigned to the correct option given each agent's available evidence. Requests are scored by the potential reduction in the same quantity if the requested information were truthfully provided. Separate rewards encourage agents to use the received information in their private choices and calculations. We post-train Qwen3-4B, Qwen3-8B, and Gemma-3-4B-it through self-play and evaluate transfer to seven unseen natural-language collaboration tasks. CoSIR achieves the highest aggregate score, while its performance in general tasks remains competitive. Behavioral analyses show increased disclosure of private evidence and fewer judged failures to use teammates' input.

Healthcare agents must transform longitudinal clinical evidence into reliable actions through heterogeneous tools and stateful workflows, conditions under which model-centric systems may lose context, misuse tools, fail to verify consequential transitions, or cross authorization boundaries. Existing medical-agent benchmarks primarily emphasize terminal task outcomes, making it difficult to determine whether execution failures arise from the agent model, the runtime harness, or their interaction. We present the Medical Harness Protocol, an outcome-preserving framework that independently evaluates seven healthcare-specific process dimensions: Care Action Fulfillment, Clinical Resource Competence, Patient-Evidence Alignment, Clinical Workflow Continuity, Clinical Auditability, Closed-Loop Verification, and Clinical Conduct Boundaries. A normalized intervention trace further tracks whether runtime support is available, activated, delivered, adopted by the agent, and propagated into subsequent behavior. We instantiate the protocol as MedHarnessBench, comprising 342 executable tasks across FHIR/EHR operations, healthcare-administration interfaces, and multimodal clinical tools. Across more than 16,000 controlled task executions, MedHarness yields broad improvements in terminal task success, intermediate subtask completion, and healthcare-specific process quality. We further evaluate MedHarness against multiple general-purpose harnesses and use controlled single-fault perturbations to verify the sensitivity and discrimination of the seven-dimensional evaluation. These results demonstrate that healthcare-specific runtime support can substantially improve both task completion and execution quality, while revealing failures that terminal outcomes alone cannot capture. Code and resources are available at https://github.com/Tomry1114/Awesome-Agent-Harness-for-Healthcare.

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

Large Language Models (LLMs) are becoming integral to modern software development workflows, assisting developers with code generation, API explanation, and iterative problem-solving through natural language conversations. Despite widespread adoption, there is limited understanding of how developers interact with LLMs in practice and how these conversational dynamics influence task outcomes, code quality, and software engineering workflows. To address this, we leverage CodeChat, a large dataset comprising 82,845 real-world developer-LLM conversations, containing 368,506 code snippets generated across over 20 programming languages, derived from the WildChat dataset. We find that LLM responses are substantially longer than developer prompts, with a median token-length ratio of 14:1. Multi-turn conversations account for 68% of the dataset and often evolve due to shifting requirements, incomplete prompts, or clarification requests. Topic analysis identifies web design (9.6% of conversations) and neural network training (8.7% of conversations) as the most frequent LLM-assisted tasks. Evaluation across five languages (i.e., Python, JavaScript, C++, Java, and C#) reveals prevalent and language-specific issues in LLM-generated code: generated Python and JavaScript code often include undefined variables (83.4% and 75.3% of code snippets, respectively); Java code lacks required comments (75.9%); C++ code frequently omits headers (41.1%) and C# code shows unresolved namespaces (49.2%). During a conversation, syntax and import errors persist across turns; however, documentation quality in Java improves by up to 14.7%, and import handling in Python improves by 3.7% over 5 turns. Prompts that point out mistakes in code generated in prior turns and explicitly request a fix are most effective for resolving errors.

As LLM-based agents are deployed in complex, multi-step workflows, a critical evaluation gap has emerged: most existing benchmarks judge only final outcomes, unable to distinguish reliable reasoning from lucky success or attribute failures to specific process deficiencies, hindering attribution in long-horizon tasks. In this work, we present ClawTrack, a dual-assessment benchmark that simultaneously measures what an agent achieves (Task Score) and how it achieves it (Process Score). ClawTrack comprises 320 tasks across 8 domains with 25+ deterministic mock services. A Process Grader scores each reasoning turn along four dimensions (goal alignment, efficiency, information utilization, and result verification), anchored by 12,541 task-specific rubric items. Evaluating 21 models over 16,000+ trials, we find that: (1) process scores effectively attribute success and failure to specific reasoning dimensions, filtering lucky passes invisible to outcome-only evaluation; (2) the four dimensions are complementary, with result verification as the systematic bottleneck; (3) the framework is robust to evaluator choice across different judge LLMs; and (4) process-based trajectory filtering yields consistent post-training improvements across model scales.

Latent world models learn action-conditioned dynamics in representation space and often score candidate actions by Euclidean distance to a goal representation. Joint training typically regularizes the representation to prevent collapse, but the resulting representation geometry also determines how terminal errors are weighted during planning. We show that accurate prediction and noncollapsed representations do not guarantee a task-aligned latent planning cost: isotropic Gaussian regularization can induce a geometry that ranks feasible outcomes differently from the task cost. To address this mismatch, we introduce AnisoWM with ΛReg, which replaces the fixed isotropic Gaussian target with a learnable diagonal covariance under fixed-trace and anisotropy constraints. The prediction objective, predictor architecture, and Euclidean planner remain unchanged; the target is used only during training. Our analysis characterizes the prediction-driven allocation of target variance, its dependence on the training distribution, and the conditions under which the induced metric reduces planning regret. Across four visual control environments, AnisoWM improves planning success over LeWorldModel in all four. Its latent planning cost also shows better agreement with task outcomes. Project website: https://rkdrn79.github.io/AnisoWM-page/

Major open-source serving stacks reuse the key-value tensors of a shared prefix by default. We measure what this prefix reuse costs in reproducibility. Holding model, decoding, seed and request order fixed at batch size one, we ran an eighty-episode multi-turn agentic tool-use workload with prefix reuse on and off across two engines and four weight formats. Without prefix reuse, repeated runs reproduced every token and log-probability in all ten configurations. With it, cache-hit and recompute paths diverged on both engines. On llama.cpp with GGUF weights this path divergence rises with quantization: for Qwen2.5-7B over ten independent server histories, it reached 37.4 percent of episodes at 16-bit and 82.5 percent at four-bit. Near-tie decisions became only about 1.2 times as common, while the cache-path change in top-two margins grew 4.9-fold from 16-bit to three-bit. Serving four-bit weights through 16-bit kernels left divergence not detectably different from 16-bit, placing the amplification mainly in the quantized execution path. At Q4_K_M, run-to-run divergence came mainly from llama.cpp's server-level prompt cache: under the canonical episode order it moved divergence by 37.5 points, and at its default, divergence spanned 20.0 to 28.7 percent under five random orders. With cache state restored, each path reproduced exactly while the two still differed. In the tested stacks, cached serving is irreproducible because the cache state is absent from the request and not reset by default. A single-turn bridge shows path divergence reaching task outcomes with no detectable net change in accuracy. We release the harness, logs, and analysis pipeline.

Deductive reasoning is the process of deriving conclusions strictly from the given premises, without relying on external knowledge. We define honesty in this setting as a model's ability to respond only when the conclusion is logically entailed by the premises, and to abstain otherwise. However, current language models often fail to reason honestly, producing unwarranted answers when the input is insufficient. To study this challenge, we formulate honest deductive reasoning as multi-step tasks where models must either derive the correct conclusion or abstain. We curate two datasets from graph structures, one for linear algebra and one for logical inference, and introduce unanswerable cases by randomly perturbing an edge in half of the instances. We find that prompting and existing training methods, including GRPO with or without supervised fine-tuning initialization, struggle on these tasks. In particular, GRPO optimize only for final task outcomes, leaving models vulnerable to collapse when negative rewards dominate early training. To address this, we propose ACNCHOR, a reinforcement learning method that injects ground truth trajectories into rollouts, preventing early training collapse. Our results demonstrate that this method stabilizes learning and significantly improves the overall reasoning performance, underscoring the importance of training dynamics for enabling honest deductive reasoning in language models.

Supervised fine-tuning (SFT) has emerged as one of the most effective ways to improve the performance of large language models (LLMs) in downstream tasks. However, SFT can have difficulty generalizing when the underlying data distribution changes, even when the new data does not fall completely outside the training domain. Recent reasoning-focused models such as o1 and R1 have demonstrated consistent gains over their non-reasoning counterparts, highlighting the importance of reasoning for improved generalization and reliability. However, collecting high-quality reasoning traces for SFT remains challenging -- annotations are costly, subjective, and difficult to scale. To address this limitation, we leverage Reinforcement Learning (RL) to enable models to learn reasoning strategies directly from task outcomes. We propose a pipeline in which LLMs generate reasoning steps that guide both the invocation of tools (e.g., function calls) and the final answer generation for conversational agents. Our method employs Group Relative Policy Optimization (GRPO) with rewards designed around tool accuracy and answer correctness, allowing the model to iteratively refine its reasoning and actions. Experimental results demonstrate that our approach improves both the quality of reasoning and the precision of tool invocations, achieving a 1.5% relative improvement over the SFT model (trained without explicit thinking) and a 40% gain compared to the base of the vanilla Qwen3-1.7B model. These findings demonstrate the promise of unifying reasoning and action learning through RL to build more capable and generalizable conversational agents.

AI agents are increasingly used in long, multi-turn workflows in both research and enterprise settings. As interactions grow, agent behavior often degrades due to loss of constraint focus, error accumulation, and memory-induced drift. This problem is especially visible in real-world deployments where context evolves, distractions are introduced, and decisions must remain consistent over time. A common practice is to equip agents with persistent memory through transcript replay or retrieval-based mechanisms. While convenient, these approaches introduce unbounded context growth and are vulnerable to noisy recall and memory poisoning, leading to unstable behavior and increased drift. In this work, we introduce the Agent Cognitive Compressor (ACC), a bio-inspired memory controller that replaces transcript replay with a bounded internal state updated online at each turn. ACC separates artifact recall from state commitment, enabling stable conditioning while preventing unverified content from becoming persistent memory. We evaluate ACC using an agent-judge-driven live evaluation framework that measures both task outcomes and memory-driven anomalies across extended interactions. Across scenarios spanning IT operations, cybersecurity response, and healthcare workflows, ACC consistently maintains bounded memory and exhibits more stable multi-turn behavior, with significantly lower hallucination and drift than transcript replay and retrieval-based agents. These results show that cognitive compression provides a practical and effective foundation for reliable memory control in long-horizon AI agents.

Skill self-evolution methods for LLM agents aim to turn execution trajectories into reusable skill documents. However, current pipelines typically derive skill patches from a single trajectory per task, merge them indiscriminately, and load the entire skill corpus during inference. These choices lead to unreliable evidence extraction, the accumulation of low-quality or even harmful skill edits, and inefficient use of context due to irrelevant or conflicting skill content. We propose SkillCAT, a framework that decomposes this process into three stages. (1) Contrastive Causal Extraction (CCE) samples multiple trajectories per task and contrasts same-task success/failure pairs to find the evidence that explains outcome differences. (2) Assessment-Augmented Evolution (AAE) replays each candidate patch on source-task clones, retains only those that do not damage task outcomes, and then merges the retained patches hierarchically. (3) Topology-Aware Task Execution (TTE) compiles the evolved skills into routable sub-skill topologies, so that inference loads only task-relevant capability nodes. We evaluate SkillCAT on widely-used agent benchmarks, including SpreadsheetBench, WikiTableQuestions, and DocVQA, and further assess cross-model and out-of-distribution generalization. Across these settings, SkillCAT improves the average score over the initial skill by up to 49.69%, demonstrating reliable and effective skill evolution.

In long-horizon tasks, recent agents based on Large Language Models (LLMs) face a significant challenge that sparse, outcome-based rewards make it difficult to assign credit to intermediate steps. Previous methods mainly focus on creating dense reward signals to guide learning, either through traditional reinforcement learning techniques like inverse reinforcement learning or by using Process Reward Models for step-by-step feedback. In this paper, we identify a fundamental problem in the learning dynamics of LLMs: the magnitude of policy gradients is inherently coupled with the entropy, which leads to inefficient small updates for confident correct actions and potentially destabilizes large updates for uncertain ones. To resolve this, we propose Entropy-Modulated Policy Gradients (EMPG), a framework that re-calibrates the learning signal based on step-wise uncertainty and the final task outcome. EMPG amplifies updates for confident correct actions, penalizes confident errors, and attenuates updates from uncertain steps to stabilize exploration. We further introduce a bonus term for future clarity that encourages agents to find more predictable solution paths. Through comprehensive experiments on three challenging agent tasks, WebShop, ALFWorld, and Deep Search, we demonstrate that EMPG achieves substantial performance gains and significantly outperforms strong policy gradient baselines. Project page is at https://empgseed-seed.github.io/

Training-free verbal reinforcement learning enables LLM agents to learn from world feedback -- objective signals such as dynamic task outcomes, market returns, or demand forecasts -- by extracting verbal rules from experience and injecting them as context, updating the agent's behavior without parameter changes. However, in non-stationary environments these agents face a retention-forgetting dilemma: retaining stale insights causes negative transfer, while discarding them causes catastrophic forgetting when conditions recur. We identify four requirements for navigating this dilemma -- outcome-driven evaluation, persistent structured evidence, non-monotonic knowledge lifecycle, and compositional governance -- and show that existing methods invest heavily in experience extraction while underinvesting in insight governance. We propose a three-layer architecture -- rules, evidence, and skills -- connected by a feedback-driven curation loop that closes the governance gap. Rules capture distilled experience from world outcomes; evidence logs track each rule's reliability across episodes; skills govern which rules to apply, how to resolve conflicts, and when to abstain. On financial forecasting as a case study, where world feedback is naturally abundant, noisy, and non-stationary, we show that the same accumulated experience either degrades performance below the zero-shot baseline or dramatically improves accuracy and risk-adjusted returns, depending on whether the curation loop is present.

Large language model (LLM) agents are increasingly deployed in long-horizon, interactive, and stateful environments. In these settings, a single wrong action, such as refunding the wrong purchase, can cause irreversible task failure and must be intercepted before execution. Such failures may not appear in every single run, but can emerge across repeated trials, making reliability across steps and trials critical. However, ensuring agentic reliability is challenging: even frontier LLMs struggle to explain why an action may be wrong, especially in long, intertwined trajectories governed by domain-specific policies. Much recent work relies on prompt-based critique agents, while optimization-based methods lack a systematic way to produce rich verification rationales for training. We address this gap with CAST, a critique-aware training framework that converts sparse task outcomes into action-level supervision for critique learning and policy optimization. CAST analyzes agent trajectories to synthesize structured rationales explaining action validity under partial observability. The resulting critique model is used to construct critique-aware training data for optimizing the policy model. Fine-tuning Qwen3-family models on dynamic tool-calling benchmarks, CAST improves reliability across domains, outperforming GPT-OSS-120B by over 10% pass^4 on Retail tasks and yielding an additional 9% improvement on Telehealth in an out-of-domain setting. These results demonstrate that critique-aware training improves the robustness of LLM agents in realistic dynamic environments.

Agentic robotics systems combine planners with robot harnesses to perform complex manipulation tasks. Recent work improves these systems through better harnesses and external memory, but relies on powerful closed-source models as planners and does not improve the planners' own capabilities. In contrast, coding agents have improved substantially by post-training the model itself. We therefore investigate whether post-training can also improve robot planners. However, applying it across robot harnesses requires accommodating different execution mechanisms, including VLA skill invocation and control-code generation. To this end, we introduce AgenticRoboRL, a framework for post-training robot planners across different harnesses. The training recipe combines supervised fine-tuning with multi-turn reinforcement learning. To accommodate different execution mechanisms, harness-specific adapters connect recorded planner interactions with task outcomes, enabling a common RL training pipeline while keeping execution components fixed. On LIBERO-PRO, AgenticRoboRL raises the success rate of a 9B-parameter planner from 18.1% to 71.0% without external memory, approaching the 72.1% achieved by Codex (GPT-5.5) with external memory. These results demonstrate post-training as an effective approach for improving robot planners. We will release the framework and trained planner checkpoints.

We study how to improve large foundation vision-language-action (VLA) systems through online reinforcement learning (RL) in real-world settings. Central to this process is the value function, which provides learning signals to guide VLA learning from experience. In practice, the value function is estimated from trajectory fragments collected from different data sources, including historical policies and intermittent human interventions. Estimating the value function of current behavior quality from the mixture data is inherently an off-policy evaluation problem. However, prior work often adopts conservative on-policy estimation for stability, which avoids direct evaluation of the current high-capacity policy and limits learning effectiveness. In this paper, we propose ALOE, an action-level off-policy evaluation framework for VLA post-training. ALOE applies chunking-based temporal-difference bootstrapping to evaluate individual action sequences instead of predicting final task outcomes. This design improves effective credit assignment to critical action chunks under sparse rewards and supports stable policy improvement. We evaluate our method on three real-world manipulation tasks, including smartphone packing as a high-precision task, laundry folding as a long-horizon deformable-object task, and bimanual pick-and-place involving multi-object perception. Across all tasks, ALOE improves learning efficiency without compromising execution speed, showing that off-policy RL can be reintroduced in a reliable manner for real-world VLA post-training. Videos and additional materials are available at our project website.

Vision-Language-Action (VLA) models leverage powerful perceptual priors from web-scale Vision-Language Model (VLM) pre-training, yet they remain surprisingly brittle in practice, frequently failing at simple robotic tasks. To mitigate this, we propose Contrastive Conceptor Activation Steering (COAST). COAST builds on the notion of a "conceptor", a linear operator that soft-projects data into the principal components of a target distribution. COAST uses conceptors to identify success-critical subspaces for a target robotic task from a few examples of success and failure rollouts. At inference time, it steers VLA latents into these identified success subspaces to improve task outcomes. Across three architecturally distinct neural policies (flow-matching VLA, autoregressive VLA, and Diffusion Policy), COAST improves absolute mean simulation and real-robot task success rate by over 20 and 40% respectively. The activation subspace geometry reveals that failure modes share substantial structure across tasks while success representations remain largely task-specific. When tasks share similar failure modes, this structure enables previously fitted conceptors to improve performance on new tasks without refitting. Ultimately, our results suggest that current VLAs retain substantial task-relevant knowledge in their latent representations, and that the action expert's decoding bottleneck could be mitigated by steering its residual stream toward task-relevant subspaces. COAST provides a lightweight, training-free path to unlocking these latent capabilities by steering the model towards its own "success" distributions.

We study building embodied agents for open-ended creative tasks. While existing methods build instruction-following agents that can perform diverse open-ended tasks, none of them demonstrates creativity -- the ability to give novel and diverse task solutions implicit in the language instructions. This limitation comes from their inability to convert abstract language instructions into concrete task goals in the environment and perform long-horizon planning for such complicated goals. Given the observation that humans perform creative tasks with the help of imagination, we propose a class of solutions for creative agents, where the controller is enhanced with an imaginator that generates detailed imaginations of task outcomes conditioned on language instructions. We introduce several approaches to implementing the components of creative agents. We implement the imaginator with either a large language model for textual imagination or a diffusion model for visual imagination. The controller can either be a behavior-cloning policy learned from data or a pre-trained foundation model generating executable codes in the environment. We benchmark creative tasks with the challenging open-world game Minecraft, where the agents are asked to create diverse buildings given free-form language instructions. In addition, we propose novel evaluation metrics for open-ended creative tasks utilizing GPT-4V, which holds many advantages over existing metrics. We perform a detailed experimental analysis of creative agents, showing that creative agents are the first AI agents accomplishing diverse building creation in the survival mode of Minecraft. Our benchmark and models are open-source for future research on creative agents (https://github.com/PKU-RL/Creative-Agents).

We introduce VisTA, a new reinforcement learning framework that empowers visual agents to dynamically explore, select, and combine tools from a diverse library based on empirical performance. Existing methods for tool-augmented reasoning either rely on training-free prompting or large-scale fine-tuning; both lack active tool exploration and typically assume limited tool diversity, and fine-tuning methods additionally demand extensive human supervision. In contrast, VisTA leverages end-to-end reinforcement learning to iteratively refine sophisticated, query-specific tool selection strategies, using task outcomes as feedback signals. Through Group Relative Policy Optimization (GRPO), our framework enables an agent to autonomously discover effective tool-selection pathways without requiring explicit reasoning supervision. Experiments on the ChartQA, Geometry3K, and BlindTest benchmarks demonstrate that VisTA achieves substantial performance gains over training-free baselines, especially on out-of-distribution examples. These results highlight VisTA's ability to enhance generalization, adaptively utilize diverse tools, and pave the way for flexible, experience-driven visual reasoning systems.

Online lifelong learning enables agents to accumulate experience across interactions and continually improve on long-horizon tasks. However, existing methods typically treat retrieval from past experience as a passive operation, triggering it only at task initialization or after completing a step. Consequently, agents often fail to identify knowledge gaps during interaction and proactively retrieve the most useful experience for the current decision. To address this limitation, we present ProactAgent, an experience-driven lifelong learning framework for proactive retrieval over a structured experience base. We first introduce Experience-Enhanced Online Evolution (ExpOnEvo), which enables continual improvement through both policy updates and memory refinement. The experience base organizes historical interactions into typed repositories, including factual memory, episodic memory, and behavioral skills, so that retrieval can provide both relevant evidence and actionable guidance. On top of this, we propose Proactive Reinforcement Learning-based Retrieval (ProactRL), which models retrieval as an explicit policy action and learns when and what to retrieve via paired-branch process rewards. By comparing continuations from identical interaction prefixes with and without retrieval, ProactRL provides step-level supervision for retrieval decisions, encouraging retrieval only when it leads to better task outcomes or higher efficiency. Experiments on SciWorld, AlfWorld, and StuLife show that ProactAgent consistently improves lifelong agent performance, achieving success rates of 73.50\% on SciWorld and 71.28\% on AlfWorld while substantially reducing retrieval overhead, and attains performance competitive with proprietary models on StuLife.

Terminals provide a powerful interface for AI agents by exposing diverse tools for automating complex workflows, yet existing terminal-agent benchmarks largely focus on tasks grounded in text, code, and structured files. However, many real-world workflows require practitioners to work directly with audio and video files. Working with such multimedia files calls for terminal agents not only to understand multimedia content, but also to convert auditory and visual evidence across related files into appropriate actions. To evaluate terminal agents on multimedia-file tasks, we introduce MultiMedia-TerminalBench (MMTB), a benchmark of 105 tasks across 5 meta-categories where terminal agents directly operate with audio and video files. Alongside MMTB, we propose Terminus-MM, a multimedia harness that extends Terminus-KIRA with audio and video perception for terminal agents. Together, MMTB and Terminus-MM support a controlled study of multimedia terminal agents, revealing how different forms of multimedia access shape task outcomes and determine which evidence agents rely on to construct executable terminal workflows. MMTB media and metadata are released at https://huggingface.co/datasets/mm-tbench/mmtb-media

As large language model agents tackle increasingly complex long-horizon tasks, effective post-training becomes critical. Prior work faces fundamental challenges: outcome-only rewards fail to precisely attribute credit to intermediate steps, estimated step-level rewards introduce systematic noise, and Monte Carlo sampling approaches for step reward estimation incur prohibitive computational cost. Inspired by findings that only a small fraction of high-entropy tokens drive effective RL for reasoning, we propose Critical Step Optimization (CSO), which focuses preference learning on verified critical steps, decision points where alternate actions demonstrably flip task outcomes from failure to success. Crucially, our method starts from failed policy trajectories rather than expert demonstrations, directly targeting the policy model's weaknesses. We use a process reward model (PRM) to identify candidate critical steps, leverage expert models to propose high-quality alternatives, then continue execution from these alternatives using the policy model itself until task completion. Only alternatives that the policy successfully executes to correct outcomes are verified and used as DPO training data, ensuring both quality and policy reachability. This yields fine-grained, verifiable supervision at critical decisions while avoiding trajectory-level coarseness and step-level noise. Experiments on GAIA-Text-103 and XBench-DeepSearch show that CSO achieves 37% and 26% relative improvement over the SFT baseline and substantially outperforms other post-training methods, while requiring supervision at only 16% of trajectory steps. This demonstrates the effectiveness of selective verification-based learning for agent post-training.

An assistant that serves the same user over a long horizon has to answer from what that user has revealed: which preferences still hold, which were revised, and which constraints apply now. Retaining that information is not the same as acting on it, and the two are usually optimized as if they were. Keeping the information as text makes the reader's input grow with the retained history, while compressing it into a fixed number of latent vectors bounds the interface but is typically trained to reconstruct text or imitate reference answers, both of which are scored on sequences the reader never produced. We present MemFold, which optimizes a fixed-budget soft memory by the behavior it supports. A query-conditioned textual memory is compressed into K continuous vectors that form the reader's memory interface, and the reader is then trained on its own rollouts under two complementary signals: group-relative rewards for task outcomes, and confidence-gated on-policy distillation in which a frozen textual-memory teacher re-scores the student's sampled tokens under the textual memory. The teacher is never sampled from, so supervision stays on the student's current distribution and adds no autoregressive decoding; at inference it is removed entirely. Across three Qwen backbones, MemFold attains the highest accuracy we measure on PersonaMem-32K and PersonaMem-128K, with margins that widen at the longer history length, and transfers to PrefEval and LongMemEval without target-domain training. Ablations attribute most of the task gain to the reward term and a smaller additional gain to the teacher signal, and memory interventions show that the reader depends on the instance-specific content of its soft memory.

Command-line interface (CLI) agents increasingly execute longer-horizon tasks, yet existing benchmarks either assume that user instructions remain fixed throughout execution or evaluate task outcomes without capturing runtime collaboration quality. In this paper, we present REALCLI, a paired benchmark that unifies the evaluation of task execution and human-agent collaboration quality on realistic CLI tasks. To comprehensively measure collaboration, REALCLI builds a five-dimensional framework, operationalized into 22 sub-dimensions and 51 interaction scenario templates. REALCLI provides two modes for paired comparison: REALCLI-EXEC supplies complete specifications upfront, whereas REALCLI-COLLAB introduces a scripted user that delivers progressive information, corrections, constraint revisions, and authorization decisions. Evaluations reveal profound capability gaps in frontier agents: the best configuration attains an average score of only 49.2%, and when evaluated by the joint collaboration score (J_{collab}), the peak performance reaches merely 37.6%. These results show that task effectiveness diverges sharply from collaboration quality, with models consistently struggling in steerability, trust calibration, and safety boundary management. REALCLI provides a rigorous, reproducible foundation for measuring and advancing collaborative CLI agents.

AI agents may consume massive tokens across a wide range of tasks, but how many of those tokens are really necessary? Although prior work has shown that agents can incur unexpectedly high token costs, we still lack a systematic way to identify and measure token waste. In this work, we introduce a unified framework for measuring token waste across heterogeneous agent systems. Our framework records and normalizes complete agent trajectories, identifies removable spans using an LLM-as-a-Judge (LLMaJ), and maps the identified text spans back to the original trajectory for token-level measurement. We validate the identified waste through direct interventions by removing marked spans, resuming agent execution, and measuring changes in task outcomes and subsequent actions. In online pruning experiments, we remove 41-46% of the model's reasoning while largely preserving task performance across two distinct agent benchmarks. Using this framework, we analyze agents across multiple settings. Our analysis shows that token waste has a system-level structure that aggregate token usage alone cannot reveal. First, waste varies more across tasks and agent settings than across models. Second, reasoning and visible execution can exhibit different waste patterns. Third, different harnesses can introduce distinct sources of inefficiency even under the same model and task. Overall, our framework transforms token waste into an actionable optimization signal, enabling researchers and practitioners to train more efficient models, design better reasoning strategies, and optimize agent harnesses toward reducing unnecessary computation rather than merely reducing token usage.

Recent advancements in large language models (LLMs) have led to significant successes across various applications, where the most noticeable is to a series of emerging capabilities, particularly in the areas of In-Context Learning (ICL) and Chain-of-Thought (CoT). To better understand and control model performance, many studies have begun investigating the underlying causes of these phenomena and their impact on task outcomes. However, existing explanatory frameworks predominantly focus on isolating and explaining ICL and CoT independently, leading to an incomplete understanding of their combined influence on model performance. To address this gap, we propose the Electronic Circuit Model (ECM), which provides a foundation for developing scalable, learnable policies and improving the management of AI-generated content. Specifically, ECM conceptualizes model behavior as an electronic circuit: ICL is represented as semantic magnetic field to providing an additional voltage following Faraday's Law, while CoT is modeled as series resistors to constrain the model output performance following Ohm's Law. Experimental results demonstrate that the ECM effectively predicts and explains LLM performance across a variety of prompting strategies. Furthermore, we apply ECM to advanced reasoning strategy optimization on a series of tasks, such as the International Olympiad in Informatics (IOI) and the International Mathematical Olympiad (IMO), achieving competitive performance that surpasses nearly 80% of top human competitors.

A core challenge for autonomous LLM agents in collaborative settings is balancing robust privacy understanding and preservation alongside task efficacy. Existing privacy benchmarks only focus on simplistic, single-turn interactions where private information can be trivially omitted without affecting task outcomes. In this paper, we introduce MAGPIE (Multi-AGent contextual PrIvacy Evaluation), a novel benchmark of 200 high-stakes tasks designed to evaluate privacy understanding and preservation in multi-agent collaborative, non-adversarial scenarios. MAGPIE integrates private information as essential for task resolution, forcing agents to balance effective collaboration with strategic information control. Our evaluation reveals that state-of-the-art agents, including GPT-5 and Gemini 2.5-Pro, exhibit significant privacy leakage, with Gemini 2.5-Pro leaking up to 50.7% and GPT-5 up to 35.1% of the sensitive information even when explicitly instructed not to. Moreover, these agents struggle to achieve consensus or task completion and often resort to undesirable behaviors such as manipulation and power-seeking (e.g., Gemini 2.5-Pro demonstrating manipulation in 38.2% of the cases). These findings underscore that current LLM agents lack robust privacy understanding and are not yet adequately aligned to simultaneously preserve privacy and maintain effective collaboration in complex environments.

Tool-using LLM agents increasingly coordinate real workloads by selecting and chaining third-party tools based on text-visible metadata such as tool names, descriptions, and return messages. We show that this convenience creates a supply-chain attack surface: a malicious MCP tool server can be co-registered alongside normal tools and induce overthinking loops, where individually trivial or plausible tool calls compose into cyclic trajectories that inflate end-to-end tokens and latency without any single step looking abnormal. We formalize this as a structural overthinking attack, distinguishable from token-level verbosity, and implement 14 malicious tools across three servers that trigger repetition, forced refinement, and distraction. Across heterogeneous registries and multiple tool-capable models, the attack causes severe resource amplification (up to 142.4times tokens) and can degrade task outcomes. Finally, we find that decoding-time concision controls do not reliably prevent loop induction, suggesting defenses should reason about tool-call structure rather than tokens alone.

Test-time scaling for complex reasoning tasks shows that leveraging inference-time compute, by methods such as independently sampling and aggregating multiple solutions, results in significantly better task outcomes. However, a critical bottleneck is verification: sampling is only effective if correct solutions can be reliably identified among candidates. While existing approaches typically evaluate candidates independently via scalar scoring, we demonstrate that models are substantially stronger at pairwise self-verification. Leveraging this insight, we introduce V_1, a framework that unifies generation and verification through efficient pairwise ranking. V_1 comprises two components: V_1-Infer, an uncertainty-guided algorithm using a tournament-based ranking that dynamically allocates self-verification compute to candidate pairs whose relative correctness is most uncertain; and V_1-PairRL, an RL framework that jointly trains a single model as both generator and pairwise self-verifier, ensuring the verifier adapts to the generator's evolving distribution. On code generation (LiveCodeBench, CodeContests, SWE-Bench) and math reasoning (AIME, HMMT) benchmarks, V_1-Infer improves Pass@1 by up to 10% over pointwise verification and outperforms recent test-time scaling methods while being significantly more efficient. Furthermore, V_1-PairRL achieves 7--9% test-time scaling gains over standard RL and pointwise joint training, and improves base Pass@1 by up to 8.7% over standard RL in a code-generation setting.

LLMs are now embedded in a wide range of everyday scenarios. However, their inherent hallucinations risk hiding misinformation in fluent responses, raising concerns about overreliance on AI. Detecting overreliance is challenging, as it often arises in complex, dynamic contexts and cannot be easily captured by post-hoc task outcomes. In this work, we aim to investigate how users' behavioral patterns correlate with overreliance. We collected interaction logs from 77 participants working with an LLM injected plausible misinformation across three real-world tasks and we assessed overreliance by whether participants detected and corrected these errors. By semantically encoding and clustering segments of user interactions, we identified five behavioral patterns linked to overreliance: users with low overreliance show careful task comprehension and fine-grained navigation; users with high overreliance show frequent copy-paste, skipping initial comprehension, repeated LLM references, coarse locating, and accepting misinformation despite hesitation. We discuss design implications for mitigation.

LLM routing matters most in long-horizon applications such as coding agents, deep research systems, and computer-use agents, where a single user request triggers many model calls. Routing each call to the cheapest sufficient model can cut costs without sacrificing quality, yet existing router benchmarks evaluate routers only on one-shot prompts. They never expose the router-visible prefix at an intermediate agent step, never test whether a cheaper replacement preserves downstream task success, and often rely on online LLM judges at evaluation time. We introduce TwinRouterBench, a step-level routing benchmark with two tracks. The static track provides 970 router-visible prefixes from 520 instances across SWE-bench, BFCL, mtRAG, QMSum, and PinchBench, each paired with an execution-verified target tier estimated under a released downgrade-and-cascade protocol; scoring is deterministic arithmetic over tier labels, trajectory membership, and token costs, with no online evaluator-side LLM judge. The dynamic track supplies a harness that runs routers on the full 500-case SWE-bench Verified suite; in this paper we report a 100-case held-out evaluation disjoint from the static SWE supervision split. At each LLM call the router selects a concrete model from a locked pool, and success is measured by official task resolution and realized API spend. The two tracks support fast offline iteration followed by end-to-end validation under live agent execution. Code and data are available at https://github.com/CommonstackAI/TwinRouterBench.

The ubiquitous independence assumption among symbolic concepts in neurosymbolic (NeSy) predictors is a convenient simplification: NeSy predictors use it to speed up probabilistic reasoning. Recent works like van Krieken et al. (2024) and Marconato et al. (2024) argued that the independence assumption can hinder learning of NeSy predictors and, more crucially, prevent them from correctly modelling uncertainty. There is, however, scepticism in the NeSy community around the scenarios in which the independence assumption actually limits NeSy systems (Faronius and Dos Martires, 2025). In this work, we settle this question by formally showing that assuming independence among symbolic concepts entails that a model can never represent uncertainty over certain concept combinations. Thus, the model fails to be aware of reasoning shortcuts, i.e., the pathological behaviour of NeSy predictors that predict correct downstream tasks but for the wrong reasons.

Computer-use agents must solve long-horizon tasks through repeated interaction with partially observable, multimodal desktop environments. Although imitation learning and offline trajectory refinement provide strong priors, static traces cannot cover the causal feedback loop of real computer use: each action changes the screen state, future action space, and recovery options. EvoCUA-1.5 extends self-evolving computer-use agents from offline experience learning to online reinforcement learning, where policies interact with executable sandbox environments and improve from verifiable task outcomes. Online RL in this setting requires more than directly reusing single-turn language-RL recipes. Multi-turn interaction introduces context-managed observations, sparse terminal rewards, variable-length trajectories, and slow environment feedback. EvoCUA-1.5 addresses these challenges with Step-Level Policy Optimization (STEPO), which preserves trajectory-level advantage balance after decomposition into step-level samples; policy-aware filtering and pass-rate calibration over verifiable synthesized tasks; Dynamic Tri-Adaptive Curriculum (DTAC), which combines learnable tasks, difficult positive replay, and controlled infeasible-task exposure; and a fully asynchronous RL infrastructure with staleness control and mini-group batching. Experiments show that these components improve training stability and downstream performance. EvoCUA-1.5 achieves 63.2\% success on OSWorld-Verified, outperforming comparable 32B/35B-scale open-weight baselines and even approaching models with significantly larger parameter counts. Overall, EvoCUA-1.5 provides a practical framework for scaling online RL in multi-turn computer-use agents.

Scaling computer-use agents from individual actions to complex software work requires experience that connects decisions to complete outcomes. We present NorthCUA, an integrated system for turbocharging computer-use agents through experience at scale. Its Experience Flywheel supplies diverse executable tasks to the NorthCUA Harness, which sustains long interactions, records decisions, and checks task outcomes. The Flywheel uses these records to curate verified trajectories. For unsolved tasks, recursive self-improvement evolves reusable skills and preserves verified milestones so later attempts build on earlier progress. Staged Learning combines this experience with grounding data from the Flywheel to develop GUI competence through mid-training and long-horizon supervised fine-tuning. The policy then learns from fresh attempts in the same Harness through joint desktop-web reinforcement learning, closing the loop between experience production and policy improvement. The resulting Nano, Mini, and Pro family achieves state-of-the-art results among open-source models across GUI grounding, desktop execution, and Web interaction under the reported evaluation settings. For more responsive Web workflows, NorthCUA-Ultrafast pairs direct action selection with open-ended generation on pre-RL Nano, offering a measured quality-latency tradeoff. Together, these advances establish NorthCUA as a high-performing model family and a reusable foundation for agents that carry complex software work from intent to completion.

Recent advances in large language models (LLMs) have significantly impacted data science workflows, giving rise to specialized data science agents designed to automate analytical tasks. Despite rapid adoption, systematic benchmarks evaluating the efficacy and limitations of these agents remain scarce. In this paper, we introduce a comprehensive benchmark specifically crafted to reflect real-world user interactions with data science agents by observing usage of our commercial applications. We evaluate three LLMs: Claude-4.0-Sonnet, Gemini-2.5-Flash, and OpenAI-o4-Mini across three approaches: zero-shot with context engineering, multi-step with context engineering, and with SmolAgent. Our benchmark assesses performance across a diverse set of eight data science task categories, additionally exploring the sensitivity of models to common prompting issues, such as data leakage and slightly ambiguous instructions. We further investigate the influence of temperature parameters on overall and task-specific outcomes for each model and approach. Our findings reveal distinct performance disparities among the evaluated models and methodologies, highlighting critical factors that affect practical deployment. The benchmark dataset and evaluation framework introduced herein aim to provide a foundation for future research of more robust and effective data science agents.

Embodied agents built on foundation models rely on an execution harness to organize observations, guide actions, and incorporate execution feedback into subsequent decisions. Failures can arise not only from limited model capabilities, but also from how the harness presents evidence and guides its use. These model-external mechanisms provide an adaptation target without updating model parameters, but task-specific interaction experience can be costly to collect. Task-level outcomes alone do not reveal whether relevant evidence was missing from the model input or was available but inadequately used. Moreover, these two aspects are coupled: effective action guidance may require evidence that the current context omits, while exposing that evidence alone does not ensure an appropriate response. We introduce EnactOpt, which uses limited execution feedback to jointly evolve reusable textual skills and context-construction code while keeping the underlying models and runtime fixed. It diagnoses adjacent decisions by relating action feedback to the actual model input and subsequent action. Structured episode summaries guide joint revisions through prompt-level edit operators, and candidate harnesses undergo compatibility checks and validation-based retention. Across VLABench, ALFWorld, and Embodied Agent Interface (EAI), EnactOpt improves task success by approximately 10-27 percentage points over the initial harnesses and outperforms the evaluated GEPA, TextGrad, and Trace2Skill baselines. These gains extend to task conditions not used for harness evolution or candidate selection. The results support a complementary route to embodied adaptation: turning limited execution experience into reusable behavioral improvements by jointly refining how evidence is presented and acted upon.

Task success can hide process anomalies in real-world agent executions. An agent may pass the final task oracle while still accumulating unresolved ambiguity, unsafe external writes, ignored errors, weakly grounded commitments, or capability-boundary overcommitment. We study this mismatch as the Outcome-Process Gap and introduce OpenClawBench, a large-scale dataset for measuring and supervising process-side anomalies in real agent execution processes. OpenClawBench is built from BFCL-driven OpenClaw sessions produced by 6 source models and contains 31,264 annotated trajectories. It aligns task-oracle outcomes with structured process evidence. FullTax converts the aligned trajectories into structured anomaly supervision: binary labels, supporting evidence, onset/span localization, severity, recoverability, and a 5-class anomaly taxonomy. Using OpenClawBench, we make the Outcome-Process Gap measurable. Among 31,135 oracle-passing executions, 2,904 are still labeled process-anomalous under FullTax. These results show that success-only evaluation misses a concrete class of process-side failures in real agent executions. A LoRA-fine-tuned Gemma 3 12B detector trained on the high-confidence FullTax supervised pool reaches binary F1=0.729 on the cleaner-labels held-out test split. Together, OpenClawBench turns real agent execution logs into auditable and reusable supervision for studying, diagnosing, and operationally monitoring runtime agent reliability.

Large Vision-Language Models (LVLMs) have recently advanced robotic manipulation by leveraging vision for scene perception and language for instruction following. However, existing methods rely heavily on costly human-annotated training datasets, which limits their generalization and causes them to struggle in out-of-domain (OOD) scenarios, reducing real-world adaptability. To address these challenges, we propose ManipLVM-R1, a novel reinforcement learning framework that replaces traditional supervision with Reinforcement Learning using Verifiable Rewards (RLVR). By directly optimizing for task-aligned outcomes, our method enhances generalization and physical reasoning while removing the dependence on costly annotations. Specifically, we design two rule-based reward functions targeting key robotic manipulation subtasks: an Affordance Perception Reward to enhance localization of interaction regions, and a Trajectory Match Reward to ensure the physical plausibility of action paths. These rewards provide immediate feedback and impose spatial-logical constraints, encouraging the model to go beyond shallow pattern matching and instead learn deeper, more systematic reasoning about physical interactions.

Language-model agents read attacker-writable context to solve tasks. Tool execution needs a separate authority check for protected sink fields, sink-interpreted payloads, and the invocation event. Context-to-Execution Integrity (CXI) is an execution-boundary system for this setting. Policies mark protected sink fields, typed releases carry narrow validated values from writable context to specific destinations, opaque data slots keep evidence as data, and a deterministic gate admits a call only after field authority, exact-effect authorization, and invocation authority all bind to the same action manifest. We evaluate CXI on open-weight field-projection runs, AgentDojo live episodes, a code-agent exact-effect benchmark, manifest-bound ledger faults, proposal-pressure controls, and hosted/API compatibility traces. AgentDojo covers 720 live episodes and 1,739 LLM calls; the code-agent benchmark covers 400 repository episodes with exact-effect authorization and lease-bound execution, yielding 231 safe task completions and zero observed field, effect, or invocation escapes. The accounting reports parser outcomes, authorization outcomes, and task-quality outcomes together with the admission-integrity result. Across the evaluated sinks, CXI admits execution only when field, effect, and invocation authority bind to the same action manifest.

Concerns about hallucinations in Large Language Models (LLMs) have been raised by researchers, yet their potential in areas where creativity is vital, such as drug discovery, merits exploration. In this paper, we come up with the hypothesis that hallucinations can improve LLMs in drug discovery. To verify this hypothesis, we use LLMs to describe the SMILES string of molecules in natural language and then incorporate these descriptions as part of the prompt to address specific tasks in drug discovery. Evaluated on seven LLMs and five classification tasks, our findings confirm the hypothesis: LLMs can achieve better performance with text containing hallucinations. Notably, Llama-3.1-8B achieves an 18.35% gain in ROC-AUC compared to the baseline without hallucination. Furthermore, hallucinations generated by GPT-4o provide the most consistent improvements across models. Additionally, we conduct empirical analyses and a case study to investigate key factors affecting performance and the underlying reasons. Our research sheds light on the potential use of hallucinations for LLMs and offers new perspectives for future research leveraging LLMs in drug discovery.

Decoding methods play an indispensable role in converting language models from next-token predictors into practical task solvers. Prior research on decoding methods, primarily focusing on task-specific models, may not extend to the current era of general-purpose large language models (LLMs). Moreover, the recent influx of decoding strategies has further complicated this landscape. This paper provides a comprehensive and multifaceted analysis of various decoding methods within the context of LLMs, evaluating their performance, robustness to hyperparameter changes, and decoding speeds across a wide range of tasks, models, and deployment environments. Our findings reveal that decoding method performance is notably task-dependent and influenced by factors such as alignment, model size, and quantization. Intriguingly, sensitivity analysis exposes that certain methods achieve superior performance at the cost of extensive hyperparameter tuning, highlighting the trade-off between attaining optimal results and the practicality of implementation in varying contexts.

Multi-agent coding has emerged as an increasingly active direction in software engineering, where complex development tasks are decomposed across multiple specialized agents working on different parts of the problem. Despite the shift from individual problem solving to distributed collaboration, multi-agent systems still lack a direct measure of collaboration and are largely evaluated through task-level outcomes inherited from single-agent coding, conflating individual coding capability with cross-agent coordination. We introduce AsynCodeBench, a dependency-centric benchmark for asynchronous multi-agent software engineering that represents each task with an explicit dependency graph and executable Dependency Checkers. Through this dependency-tracking process, we propose two complementary measures: Asynchronous Dependency Pass Rate (ADPR), which measures how many cross-agent dependencies are ultimately satisfied, and Dependency Resolution Step (DRS), which measures when each dependency first becomes satisfied during execution. AsynCodeBench comprises 19 tasks from real-world repositories, exposing 52 directed dependencies as explicit units for evaluating cross-agent collaboration. Experiments across model families, scales, and generations reveal a clear gap between coding and collaboration capability: improvements in coding performance do not necessarily translate into stronger collaboration, and task-level metrics can diverge substantially from dependency-level collaboration measures. Dependency-trajectory analysis further reveals that successful coordination often emerges not gradually, but through concentrated bursts in which many dependencies become resolved over a short portion of the execution trajectory, a pattern we term a hopping window.

Chemical Language Models (CLMs) pre-trained on large scale molecular data are widely used for molecular property prediction. However, the common belief that increasing training resources such as model size, dataset size, and training compute improves both pretraining loss and downstream task performance has not been systematically validated in the chemical domain. In this work, we evaluate this assumption by pretraining CLMs while scaling training resources and measuring transfer performance across diverse molecular property prediction (MPP) tasks. We find that while pretraining loss consistently decreases with increased training resources, downstream task performance shows limited improvement. Moreover, alternative metrics based on the Hessian or loss landscape also fail to estimate downstream performance in CLMs. We further identify conditions under which downstream performance saturates or degrades despite continued improvements in pretraining metrics, and analyze the underlying task dependent failure modes through parameter space visualizations. These results expose a gap between pretraining based evaluation and downstream performance, and emphasize the need for model selection and evaluation strategies that explicitly account for downstream task characteristics.

World-action models (WAMs) augment robot policies with action-conditioned environment dynamics, yet existing approaches largely rely on future observation reconstruction or generic latent prediction and lack structured, control-oriented world representations tightly coupled with action generation. We introduce Magic-W0, a world-action foundation model that jointly models structured physical state evolution and continuous actions. Magic-W0 represents interaction as a Structured World Transition consisting of Current State, Transition, and Future State. Current State combines vision-language context with Current 3D Geometry; Transition is represented by 3D Motion capturing action-induced three-dimensional changes; and Future State is represented by Future Semantics describing task-relevant outcomes. To couple prediction and control, we propose a layer-aligned world-action interaction architecture in which evolving action hypotheses condition world-transition prediction, while predicted world representations continuously inform action generation. Magic-W0 is pre-trained on large-scale egocentric human manipulation, UMI, real-robot, and simulation data, with latent supervision for geometry, 3D motion, and future semantics from pre-trained visual models. Inference-time interventions show that structured world representations respond systematically to changes in candidate actions and that action-related information propagates through shared 3D representations into future semantic predictions. On RoboDojo-Sim, Magic-W0 achieves an average Score of 27.10, the highest among the compared WAMs. Across multiple real-robot tasks, it also demonstrates strong downstream performance after fine-tuning with limited downstream data, supporting generalization and rapid adaptation.

Reasoning-enabled VLA policies expose chain-of-thought (CoT) traces that appear to explain and guide their actions, creating a potential interface for runtime safety through reasoning monitoring and correction. In this work, we define and operationalize two evaluation axes for assessing when this interface can improve embodied behavior: correctability, which measures whether unreliable reasoning can be detected and improved during generation, and actionability, which measures whether reasoning corrections produce behaviorally meaningful changes in the intended direction. To enable correctability, we introduce Token-level Reward for Utility-Steered Chain-of-Thought (TRUST), an offline-trained value model that predicts eventual reasoning correctness from partial prefixes and uses these estimates to monitor and selectively steer reasoning generation in frozen VLA policies. On the Alpamayo 1.5 driving VLA, TRUST monitors correctness with 88.9% accuracy and improves reasoning correctness from 75.9% to 90.0%. On a baseline-defined challenging subset in AlpaSim, TRUST reduces collision rate by 30.4% and maximum trajectory error by 11.5% relative to the unsteered policy, outperforming a compute-matched Best-of-4 baseline. On the DeepThinkVLA manipulation VLA, TRUST improves the correctness of grasp-state claims from 69.3% to 90.2% and action-choice claims from 68.8% to 85.9%, yet closed-loop task performance on LIBERO-Plus remains largely unchanged. Empirical analysis reveals intent-consistent behavioral effects in Alpamayo 1.5 but limited effects in DeepThinkVLA, helping interpret these different task-level outcomes. Together, our results show that gains in reasoning correctness do not automatically imply gains in embodied performance, motivating evaluation of correctability and actionability when using CoT as a runtime safety interface.

Harnesses are now central to coding-agent performance, mediating how models interact with tools and execution environments. Yet harness engineering remains a manual craft, because automating it faces a heterogeneous action space across editable components, voluminous trajectories that bury actionable signal, and edits whose effect is hard to attribute. We introduce Agentic Harness Engineering (AHE), a closed loop that addresses these challenges through three matched observability pillars: (1) component observability gives every editable harness component a file-level representation so the action space is explicit and revertible; (2) experience observability distills millions of raw trajectory tokens into a layered, drill-down evidence corpus that an evolving agent can actually consume; and (3) decision observability pairs every edit with a self-declared prediction, later verified against the next round's task-level outcomes. Together, these pillars turn every edit into a falsifiable contract, so harness evolution proceeds autonomously without collapsing into trial-and-error. Empirically, ten AHE iterations lift pass@1 on Terminal-Bench 2 from 69.7% to 77.0%, surpassing the human-designed harness Codex-CLI (71.9%) and the self-evolving baselines ACE and TF-GRPO. The frozen harness transfers without re-evolution: on SWE-bench-verified it tops aggregate success at 12% fewer tokens than the seed, and on Terminal-Bench 2 it yields +5.1 to +10.1pp cross-family gains across three alternate model families, indicating the evolved components encode general engineering experience rather than benchmark-specific tuning. Ablations localize the gain to tools, middleware, and long-term memory rather than the system prompt, suggesting factual harness structure transfers while prose-level strategy does not.

Accelerating scientific discovery requires the identification of which experiments would yield the best outcomes before committing resources to costly physical validation. While existing benchmarks evaluate LLMs on scientific knowledge and reasoning, their ability to predict experimental outcomes - a task where AI could significantly exceed human capabilities - remains largely underexplored. We introduce SciPredict, a benchmark comprising 405 tasks derived from recent empirical studies in 33 specialized sub-fields of physics, biology, and chemistry. SciPredict addresses two critical questions: (a) can LLMs predict the outcome of scientific experiments with sufficient accuracy? and (b) can such predictions be reliably used in the scientific research process? Evaluations reveal fundamental limitations on both fronts. Model accuracies are 14-26% and human expert performance is approx20%. Although some frontier models exceed human performance model accuracy is still far below what would enable reliable experimental guidance. Even within the limited performance, models fail to distinguish reliable predictions from unreliable ones, achieving only approx20% accuracy regardless of their confidence or whether they judge outcomes as predictable without physical experimentation. Human experts, in contrast, demonstrate strong calibration: their accuracy increases from approx5% to approx80% as they deem outcomes more predictable without conducting the experiment. SciPredict establishes a rigorous framework demonstrating that superhuman performance in experimental science requires not just better predictions, but better awareness of prediction reliability. For reproducibility all our data and code are provided at https://github.com/scaleapi/scipredict

We describe the SemEval task of extracting keyphrases and relations between them from scientific documents, which is crucial for understanding which publications describe which processes, tasks and materials. Although this was a new task, we had a total of 26 submissions across 3 evaluation scenarios. We expect the task and the findings reported in this paper to be relevant for researchers working on understanding scientific content, as well as the broader knowledge base population and information extraction communities.

Recent advances in large language models (LLMs) have enabled the development of autonomous agents capable of complex reasoning and multi-step problem solving. However, these agents struggle to adapt to specialized environments and do not leverage past interactions, approaching each new task from scratch regardless of their accumulated experience. We introduce Experiential Reflective Learning (ERL), a simple self-improvement framework that enables rapid environment adaptation through experiential learning. ERL reflects on task trajectories and outcomes to generate heuristics, capturing actionable lessons that transfer across tasks. At test time, relevant heuristics are retrieved based on the current task and injected into the agent's context to guide execution. On the Gaia2 benchmark, ERL improves success rate by 7.8% over a ReAct baseline, with large gains in task completion reliability, and outperforms prior experiential learning methods. Through systematic ablations, we find that selective retrieval is essential and that heuristics provide more transferable abstractions than few-shot trajectory prompting. These results demonstrate that reflecting on single-attempt experiences to extract transferable heuristics enables effective agent self-improvement.

Scientific coding agents receive equations, boundary conditions, and output requirements in writing, then must assess the programs they revise. Rules to Tools (R2T) supplies prepared executable checks of public scientific requirements. Matched SciCode repair groups share written checks, starting programs, model, and budgets; the tool group receives a callable implementation. Across two task-ID cohorts, complete repair is 26/30 with text and 29/30 with the prepared checks. Three task IDs favor tools, one favors text, and eleven tie. The eight-ID cohort scores 13/16 versus 15/16, with a task-cluster bootstrap 95% interval of [-12.5, 43.75] percentage points for the difference. The larger shared-definition SciCode cohort ties at 13/24 per group. Five development-exposed tasks with alternate starting programs score 3/10 versus 7/10. The tool group favors tasks 17, 77, and 11; initial checks flag task 17 and report no violation for tasks 77 and 11. Task 37 favors text and has no initial reported violation. A fresh source-through-Python arm also reaches 15/16, matching the dedicated command's aggregate. In a matched PDE comparison, detailed text scores 23/24 and checks score 24/24, with 31.2% lower reported model output for checks. Agent-side output savings vary by cohort, while public CPU use rises in both task-ID cohorts. These results measure task-dependent repair outcomes and agent-side costs with prepared checks.

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

AI coding agents are now submitting pull requests (PRs) to software projects, acting not just as assistants but as autonomous contributors. As these agentic contributions are rapidly increasing across real repositories, little is known about how they behave in practice and why many of them fail to be merged. In this paper, we conduct a large-scale study of 33k agent-authored PRs made by five coding agents across GitHub. (RQ1) We first quantitatively characterize merged and not-merged PRs along four broad dimensions: 1) merge outcomes across task types, 2) code changes, 3) CI build results, and 4) review dynamics. We observe that tasks related to documentation, CI, and build update achieve the highest merge success, whereas performance and bug-fix tasks perform the worst. Not-merged PRs tend to involve larger code changes, touch more files, and often do not pass the project's CI/CD pipeline validation. (RQ2) To further investigate why some agentic PRs are not merged, we qualitatively analyze 600 PRs to derive a hierarchical taxonomy of rejection patterns. This analysis complements the quantitative findings in RQ1 by uncovering rejection reasons not captured by quantitative metrics, including lack of meaningful reviewer engagement, duplicate PRs, unwanted feature implementations, and agent misalignment. Together, our findings highlight key socio-technical and human-AI collaboration factors that are critical to improving the success of future agentic workflows.

Do LLM scientific agents rely only on their initial task context, or do they adapt their decisions in response to experimental feedback? We study this question in neural operator adaptation, where a large language model (LLM) selects fine-tuning configurations under a limited trial budget. Across transfers within and between partial differential equation (PDE) families, the LLM achieves lower held-out test nRMSE than random search and Bayesian optimisation in nearly every matched comparison. Endpoint performance alone cannot distinguish what happens, so we verify each attribution with controlled interventions. Before observing any validation score, the LLM's first configuration already ranks near the top of the corresponding random-search pool, indicating a useful initial bias. A complementary cold-start intervention shows that the selected base learning rate shifts with the PDE description. Once feedback becomes available, reassigning validation scores among evaluated configurations changes the next proposal in every case tested, whereas a value-preserving rewrite produces no comparable aggregate effect. These interventions establish that the LLM's decision-level actions respond to the given task and observed outcomes, showing that it combines a task-dependent prior with sensitivity to experimental feedback.

Expert demonstrations show how to complete contact-rich tasks, but provide limited coverage of how interaction forces respond to different actions. Task-agnostic play provides broader experience of contact dynamics without specifying task goals. To fully leverage these two types of data, we introduce ForceWAM, a force world action model that decouples knowledge of task-compatible interaction outcomes from knowledge of how local actions change contact. A world model learns from successful demonstrations to jointly predict future video and six-axis force/torque trajectories. Together, these futures describe the expected evolution of motion and contact during the task. A separate inverse dynamics model (IDM) learns from both task and play, mapping current observations and recorded visual-wrench futures to their associated robot actions. Play data thus provide knowledge of force responses for goal-directed execution. At deployment, the IDM reactively translates generated futures into actions, reusing them with fresh wrench feedback between world-model updates. Across real-robot force-regulated sliding, box flipping with and without disturbances, and ring placement, ForceWAM achieves 85-95% mean success, surpassing world-modeling and reactive baselines. Our experiments further show that adding task-agnostic play data improves success by 15-45 percentage points, demonstrating its value for learning contact-rich execution. Project page and code are available at https://force-wam.github.io/.

Structured pruning of large language models (LLMs) offers substantial efficiency improvements by removing entire hidden units, yet current approaches often suffer from significant performance degradation, particularly in zero-shot settings, and necessitate costly recovery techniques such as supervised fine-tuning (SFT) or adapter insertion. To address these critical shortcomings, we introduce NIRVANA, a novel pruning method explicitly designed to balance immediate zero-shot accuracy preservation with robust fine-tuning capability. Leveraging a first-order saliency criterion derived from the Neural Tangent Kernel under Adam optimization dynamics, NIRVANA provides a theoretically grounded pruning strategy that respects essential model training behaviors. To further address the unique challenges posed by structured pruning, NIRVANA incorporates an adaptive sparsity allocation mechanism across layers and modules (attention vs. MLP), which adjusts pruning intensity between modules in a globally balanced manner. Additionally, to mitigate the high sensitivity of pruning decisions to calibration data quality, we propose a simple yet effective KL divergence-based calibration data selection strategy, ensuring more reliable and task-agnostic pruning outcomes. Comprehensive experiments conducted on Llama3, Qwen, and T5 models demonstrate that NIRVANA outperforms existing structured pruning methods under equivalent sparsity constraints, providing a theoretically sound and practical approach to LLM compression. The code is available at https://github.com/iDEA-iSAIL-Lab-UIUC/NIRVANA.

The principal goal of the TREC Neural Cross-Language Information Retrieval (NeuCLIR) track is to study the impact of neural approaches to cross-language information retrieval. The track has created four collections, large collections of Chinese, Persian, and Russian newswire and a smaller collection of Chinese scientific abstracts. The principal tasks are ranked retrieval of news in one of the three languages, using English topics. Results for a multilingual task, also with English topics but with documents from all three newswire collections, are also reported. New in this second year of the track is a pilot technical documents CLIR task for ranked retrieval of Chinese technical documents using English topics. A total of 220 runs across all tasks were submitted by six participating teams and, as baselines, by track coordinators. Task descriptions and results are presented.

Retrieval-Augmented Language Models boost task performance, owing to the retriever that provides external knowledge. Although crucial, the retriever primarily focuses on semantics relevance, which may not always be effective for generation. Thus, utility-based retrieval has emerged as a promising topic, prioritizing passages that provides valid benefits for downstream tasks. However, due to insufficient understanding, capturing passage utility accurately remains unexplored. This work proposes SCARLet, a framework for training utility-based retrievers in RALMs, which incorporates two key factors, multi-task generalization and inter-passage interaction. First, SCARLet constructs shared context on which training data for various tasks is synthesized. This mitigates semantic bias from context differences, allowing retrievers to focus on learning task-specific utility for better task generalization. Next, SCARLet uses a perturbation-based attribution method to estimate passage-level utility for shared context, which reflects interactions between passages and provides more accurate feedback. We evaluate our approach on ten datasets across various tasks, both in-domain and out-of-domain, showing that retrievers trained by SCARLet consistently improve the overall performance of RALMs.

Despite the critical need to align search targets with users' intention, retrievers often only prioritize query information without delving into the users' intended search context. Enhancing the capability of retrievers to understand intentions and preferences of users, akin to language model instructions, has the potential to yield more aligned search targets. Prior studies restrict the application of instructions in information retrieval to a task description format, neglecting the broader context of diverse and evolving search scenarios. Furthermore, the prevailing benchmarks utilized for evaluation lack explicit tailoring to assess instruction-following ability, thereby hindering progress in this field. In response to these limitations, we propose a novel benchmark,INSTRUCTIR, specifically designed to evaluate instruction-following ability in information retrieval tasks. Our approach focuses on user-aligned instructions tailored to each query instance, reflecting the diverse characteristics inherent in real-world search scenarios. Through experimental analysis, we observe that retrievers fine-tuned to follow task-style instructions, such as INSTRUCTOR, can underperform compared to their non-instruction-tuned counterparts. This underscores potential overfitting issues inherent in constructing retrievers trained on existing instruction-aware retrieval datasets.

Recent advances in applying reinforcement learning (RL) to large language models (LLMs) have led to substantial progress. In particular, a series of remarkable yet often counterintuitive phenomena have been reported in LLMs, exhibiting patterns not typically observed in traditional RL settings. For example, notable claims include that a single training example can match the performance achieved with an entire dataset, that the reward signal does not need to be very accurate, and that training solely with negative samples can match or even surpass sophisticated reward-based methods. However, the precise conditions under which these observations hold - and, critically, when they fail - remain unclear. In this work, we identify a key factor that differentiates RL observations: whether the pretrained model already exhibits strong Model-Task Alignment, as measured by pass@k accuracy on the evaluated task. Through a systematic and comprehensive examination of a series of counterintuitive claims, supported by rigorous experimental validation across different model architectures and task domains, our findings show that while standard RL training remains consistently robust across settings, many of these counterintuitive results arise only when the model and task already exhibit strong model-task alignment. In contrast, these techniques fail to drive substantial learning in more challenging regimes, where standard RL methods remain effective.

On-policy distillation (OPD) trains a student on its own responses with token-level feedback from a stronger teacher, yet the teacher can fail on questions the student already answers correctly, and how often each model succeeds varies across tasks. OPD ignores these outcomes and, on average, pushes down even the student's correct responses; gating feedback by student correctness fixes the direction but uses the teacher in the same way whether or not it succeeded. We introduce DuoOPD, in which the student's outcome sets the direction of feedback and the joint teacher-student outcome decides how the teacher supports it: when only the teacher succeeds, its verified answer becomes context for scoring the student's failed response, and when only the student succeeds, a weight shared within the task reinforces the whole response. A single rule covers all four outcome combinations without task-specific settings. Across Qwen3 and Llama, DuoOPD outperforms all five baselines in mean macro accuracy, improving over OPD by 2.58 and 5.98 percentage points, and it also leads on two further task mixtures spanning scientific calculation, instruction following, and code generation. Ablations show that outcome-based direction alone stays near the gated baseline, while the joint-outcome designs supply most of the gain.

This is the fifth year of the TREC Deep Learning track. As in previous years, we leverage the MS MARCO datasets that made hundreds of thousands of human-annotated training labels available for both passage and document ranking tasks. We mostly repeated last year's design, to get another matching test set, based on the larger, cleaner, less-biased v2 passage and document set, with passage ranking as primary and document ranking as a secondary task (using labels inferred from passage). As we did last year, we sample from MS MARCO queries that were completely held out, unused in corpus construction, unlike the test queries in the first three years. This approach yields a more difficult test with more headroom for improvement. Alongside the usual MS MARCO (human) queries from MS MARCO, this year we generated synthetic queries using a fine-tuned T5 model and using a GPT-4 prompt. The new headline result this year is that runs using Large Language Model (LLM) prompting in some way outperformed runs that use the "nnlm" approach, which was the best approach in the previous four years. Since this is the last year of the track, future iterations of prompt-based ranking can happen in other tracks. Human relevance assessments were applied to all query types, not just human MS MARCO queries. Evaluation using synthetic queries gave similar results to human queries, with system ordering agreement of τ=0.8487. However, human effort was needed to select a subset of the synthetic queries that were usable. We did not see clear evidence of bias, where runs using GPT-4 were favored when evaluated using synthetic GPT-4 queries, or where runs using T5 were favored when evaluated on synthetic T5 queries.

Evaluation efforts such as TREC, CLEF, NTCIR and FIRE, alongside public leaderboard such as MS MARCO, are intended to encourage research and track our progress, addressing big questions in our field. However, the goal is not simply to identify which run is "best", achieving the top score. The goal is to move the field forward by developing new robust techniques, that work in many different settings, and are adopted in research and practice. This paper uses the MS MARCO and TREC Deep Learning Track as our case study, comparing it to the case of TREC ad hoc ranking in the 1990s. We show how the design of the evaluation effort can encourage or discourage certain outcomes, and raising questions about internal and external validity of results. We provide some analysis of certain pitfalls, and a statement of best practices for avoiding such pitfalls. We summarize the progress of the effort so far, and describe our desired end state of "robust usefulness", along with steps that might be required to get us there.

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

Recent neural retrieval mainly focuses on ranking short texts and is challenged with long documents. Existing work mainly evaluates either ranking passages or whole documents. However, there are many cases where the users want to find a relevant passage within a long document from a huge corpus, e.g. legal cases, research papers, etc. In this scenario, the passage often provides little document context and thus challenges the current approaches to finding the correct document and returning accurate results. To fill this gap, we propose and name this task Document-Aware Passage Retrieval (DAPR) and build a benchmark including multiple datasets from various domains, covering both DAPR and whole-document retrieval. In experiments, we extend the state-of-the-art neural passage retrievers with document-level context via different approaches including prepending document summary, pooling over passage representations, and hybrid retrieval with BM25. The hybrid-retrieval systems, the overall best, can only improve on the DAPR tasks marginally while significantly improving on the document-retrieval tasks. This motivates further research in developing better retrieval systems for the new task. The code and the data are available at https://github.com/kwang2049/dapr

Task semantics can be expressed by a set of input-to-output examples or a piece of textual instruction. Conventional machine learning approaches for natural language processing (NLP) mainly rely on the availability of large-scale sets of task-specific examples. Two issues arise: first, collecting task-specific labeled examples does not apply to scenarios where tasks may be too complicated or costly to annotate, or the system is required to handle a new task immediately; second, this is not user-friendly since end-users are probably more willing to provide task description rather than a set of examples before using the system. Therefore, the community is paying increasing interest in a new supervision-seeking paradigm for NLP: learning from task instructions. Despite its impressive progress, there are some common issues that the community struggles with. This survey paper tries to summarize and provide insights into the current research on instruction learning, particularly by answering the following questions: (i) What is task instruction, and what instruction types exist? (ii) How to model instructions? (iii) What factors influence and explain the instructions' performance? (iv) What challenges remain in instruction learning? To our knowledge, this is the first comprehensive survey about textual instructions.

We describe our system for LitTraceQA (GroundLM @ EMNLP 2026): given a research question, retrieve the relevant papers from a pool of 27,487, cite the page and the table or figure where the answer lives, and answer in a requested format. Our main finding is that evidence grounding is limited by retrieval, not by reading. The page selector put the annotator's page, which we call the gold page, in front of the model that locates evidence only about half the time (52.6% gold-page recall), while that model, given the page, cited the right one in 45 of the 48 locators it emitted (94%). When the page was missing it rarely said so: of 45 such cases it returned nothing 14 times, a wrong page 24 times, and a correct page 7 times, so the pipeline failed quietly almost twice as often as it failed visibly. Since the failure was that the right page was never shown, the fix is to stop choosing: each retrieved paper fits in the model's context, so we show it whole. Page ranking survives only as a fallback inside papers too long to fit, which no test-split paper was, and gold-page recall reaches 100% on the papers we can parse. Separately, questions that identify their target by position rather than content, such as "the first author of the 24th reference", are served by parsing rather than retrieval: we resolve the bibliography into an addressable list, which also supplies identifiers the evidence metric scores. The final system scores 0.762 paper F_1, 0.441 evidence F_1 and 0.920 multiple-choice accuracy on the held-out test split. Because the pipeline depends on a closed model without seed control, we release a harness that verifies the paper's central claims against committed artifacts.

We study the problem of retrieval with instructions, where users of a retrieval system explicitly describe their intent along with their queries. We aim to develop a general-purpose task-aware retrieval system using multi-task instruction tuning, which can follow human-written instructions to find the best documents for a given query. We introduce the first large-scale collection of approximately 40 retrieval datasets with instructions, BERRI, and present TART, a multi-task retrieval system trained on BERRI with instructions. TART shows strong capabilities to adapt to a new retrieval task via instructions and advances the state of the art on two zero-shot retrieval benchmarks, BEIR and LOTTE, outperforming models up to three times larger. We further introduce a new evaluation setup, X^2-Retrieval to better reflect real-world scenarios, where diverse domains and tasks are pooled and a system needs to find documents aligning users' intents. In this setup, TART significantly outperforms competitive baselines, further demonstrating the effectiveness of guiding retrieval with instructions.

Context: Recent research indicates that Web queries written by software developers are not very successful in retrieving relevant results, performing measurably worse compared to general purpose Web queries. Most approaches up to this point have addressed this problem with software engineering-specific automated query reformulation techniques, which work without developer involvement but are limited by the content of the original query. In other words, these techniques automatically improve the existing query but can not contribute new, previously unmentioned, concepts. Objective: In this paper, we propose a technique to guide software developers in manually improving their own Web search queries. We examine a conversational approach that follows unsuccessful queries with a clarification question aimed at eliciting additional query terms, thus providing to the developer a clear dimension along which the query could be improved. Methods: We describe a set of clarification questions derived from a corpus of software developer queries and a neural approach to recommending them for a newly issued query. Results: Our evaluation indicates that the recommendation technique is accurate, predicting a valid clarification question 80% of the time and outperforms simple baselines, as well as, state-of-the-art Learning To Rank (LTR) baselines. Conclusion: As shown in the experimental results, the described approach is capable at recommending appropriate clarification questions to software developers and considered useful by a sample of developers ranging from novices to experienced professionals.

We here summarize our experience running a challenge with open data for musical genre recognition. Those notes motivate the task and the challenge design, show some statistics about the submissions, and present the results.

Instruction-tuned language models (LM) are able to respond to imperative commands, providing a more natural user interface compared to their base counterparts. In this work, we present Promptriever, the first retrieval model able to be prompted like an LM. To train Promptriever, we curate and release a new instance-level instruction training set from MS MARCO, spanning nearly 500k instances. Promptriever not only achieves strong performance on standard retrieval tasks, but also follows instructions. We observe: (1) large gains (reaching SoTA) on following detailed relevance instructions (+14.3 p-MRR / +3.1 nDCG on FollowIR), (2) significantly increased robustness to lexical choices/phrasing in the query+instruction (+12.9 Robustness@10 on InstructIR), and (3) the ability to perform hyperparameter search via prompting to reliably improve retrieval performance (+1.4 average increase on BEIR). Promptriever demonstrates that retrieval models can be controlled with prompts on a per-query basis, setting the stage for future work aligning LM prompting techniques with information retrieval.

When reading, we often have specific information that interests us in a text. For example, you might be reading this paper because you are curious about LLMs for eye movements in reading, the experimental design, or perhaps you only care about the question ``but does it work?''. More broadly, in daily life, people approach texts with any number of text-specific goals that guide their reading behavior. In this work, we ask, for the first time, whether open-ended reading goals can be automatically decoded from eye movements in reading. To address this question, we introduce goal classification and goal reconstruction tasks and evaluation frameworks, and use large-scale eye tracking for reading data in English with hundreds of text-specific information seeking tasks. We develop and compare several discriminative and generative multimodal LLMs that combine eye movements and text for goal classification and goal reconstruction. Our experiments show considerable success on both tasks, suggesting that LLMs can extract valuable information about the readers' text-specific goals from eye movements.

Finding relevant literature underpins the practice of evidence-based medicine. From 2014 to 2016, TREC conducted a clinical decision support track, wherein participants were tasked with finding articles relevant to clinical questions posed by physicians. In total, 87 teams have participated over the past three years, generating 395 runs. During this period, each team has trialled a variety of methods. While there was significant overlap in the methods employed by different teams, the results were varied. Due to the diversity of the platforms used, the results arising from the different techniques are not directly comparable, reducing the ability to build on previous work. By using a stable platform, we have been able to compare different document and query processing techniques, allowing us to experiment with different search parameters. We have used our system to reproduce leading teams runs, and compare the results obtained. By benchmarking our indexing and search techniques, we can statistically test a variety of hypotheses, paving the way for further research.

Search agents powered by large language models can autonomously decompose queries, retrieve information, and synthesize answers through multi-step reasoning. However, the rapid growth of training methods has outpaced controlled comparison: existing works differ in retrieval corpora, reward designs, and training protocols, making it unclear what actually drives improvements. We present a controlled empirical study that isolates three under-explored dimensions of search agent training. First, we identify a critical data-coverage issue in the widely used Wikipedia 2018 corpus and show that correcting it alone yields larger gains than the differences between training algorithms. Second, we systematically compare outcome-based and process-based reward methods across three base models, finding that the simplest outcome-based approach achieves competitive or superior performance in most settings, and that process-level credit assignment can over-correct agent behavior. Third, we analyze training data diversity, off-policy data utilization, and search budget scaling, distilling practical guidelines for training effective search agents. Our code is available at https://github.com/YiboZhao624/SearchAgentReview.

In the information retrieval (IR) domain, evaluation plays a crucial role in optimizing search experiences and supporting diverse user intents. In the recent LLM era, research has been conducted to automate document relevance labels, as these labels have traditionally been assigned by crowd-sourced workers - a process that is both time and consuming and costly. This study focuses on LLM-generated usefulness labels, a crucial evaluation metric that considers the user's search intents and task objectives, an aspect where relevance falls short. Our experiment utilizes task-level, query-level, and document-level features along with user search behavior signals, which are essential in defining the usefulness of a document. Our research finds that (i) pre-trained LLMs can generate moderate usefulness labels by understanding the comprehensive search task session, (ii) pre-trained LLMs perform better judgement in short search sessions when provided with search session contexts. Additionally, we investigated whether LLMs can capture the unique divergence between relevance and usefulness, along with conducting an ablation study to identify the most critical metrics for accurate usefulness label generation. In conclusion, this work explores LLM-generated usefulness labels by evaluating critical metrics and optimizing for practicality in real-world settings.

We study prompting-based approaches with Large Language Models (LLMs) for modifying documents so as to promote their ranking in a competitive search setting. Our methods are inspired by prior work on leveraging LLMs as rankers. We evaluate our approach by deploying it as a bot in previous ranking competitions and in competitions we organized. Our findings demonstrate that our approach effectively improves document ranking while preserving high levels of faithfulness to the original content and maintaining overall document quality.

Terminal-agent benchmarks have become a primary signal for measuring the coding and system-administration capabilities of large language models. As the market for evaluation environments grows, so does the pressure to ship tasks quickly, often without thorough adversarial review of the verification logic. This paper is a guideline for writing good benchmark tasks, drawn from over a year of contributing to and reviewing tasks for Terminal Bench. Most people write benchmark tasks the way they write prompts. They shouldn't. A prompt is designed to help the agent succeed; a benchmark is designed to find out if it can. We argue that good tasks are adversarial, difficult, and legible, and that a large class of common failure modes -- AI-generated instructions, over-prescriptive specifications, clerical difficulty, oracle solutions that assume hidden knowledge, tests that validate the wrong things, and reward-hackable environments -- are predictable consequences of treating task authoring as prompt authoring. We catalog these failure modes, argue that real difficulty is conceptual rather than environmental, and discuss recent empirical evidence that over 15% of tasks in popular terminal-agent benchmarks are reward-hackable. We hope this serves as a useful reference for benchmark maintainers, task contributors, and researchers using benchmark scores as evidence.

To enable building and testing models on long-document comprehension, we introduce QuALITY, a multiple-choice QA dataset with context passages in English that have an average length of about 5,000 tokens, much longer than typical current models can process. Unlike in prior work with passages, our questions are written and validated by contributors who have read the entire passage, rather than relying on summaries or excerpts. In addition, only half of the questions are answerable by annotators working under tight time constraints, indicating that skimming and simple search are not enough to consistently perform well. Our baseline models perform poorly on this task (55.4%) and significantly lag behind human performance (93.5%).

The principal goal of the TREC Neural Cross-Language Information Retrieval (NeuCLIR) track is to study the effect of neural approaches on cross-language information access. The track has created test collections containing Chinese, Persian, and Russian news stories and Chinese academic abstracts. NeuCLIR includes four task types: Cross-Language Information Retrieval (CLIR) from news, Multilingual Information Retrieval (MLIR) from news, Report Generation from news, and CLIR from technical documents. A total of 274 runs were submitted by five participating teams (and as baselines by the track coordinators) for eight tasks across these four task types. Task descriptions and the available results are presented.

We present our submission to the AXOLOTL-24 shared task. The shared task comprises two subtasks: identifying new senses that words gain with time (when comparing newer and older time periods) and producing the definitions for the identified new senses. We implemented a conceptually simple and computationally inexpensive solution to both subtasks. We trained adapter-based binary classification models to match glosses with usage examples and leveraged the probability output of the models to identify novel senses. The same models were used to match examples of novel sense usages with Wiktionary definitions. Our submission attained third place on the first subtask and the first place on the second subtask.

Neural networks -- especially those that use large, pre-trained language models -- have improved search engines in various ways. Most prominently, they can estimate the relevance of a passage or document to a user's query. In this work, we depart from this direction by exploring whether neural networks can effectively predict which of a document's passages are unlikely to be relevant to any query submitted to the search engine. We refer to this query-agnostic estimation of passage relevance as a passage's quality. We find that our novel methods for estimating passage quality allow passage corpora to be pruned considerably while maintaining statistically equivalent effectiveness; our best methods can consistently prune >25% of passages in a corpora, across various retrieval pipelines. Such substantial pruning reduces the operating costs of neural search engines in terms of computing resources, power usage, and carbon footprint -- both when processing queries (thanks to a smaller index size) and when indexing (lightweight models can prune low-quality passages prior to the costly dense or learned sparse encoding step). This work sets the stage for developing more advanced neural "learning-what-to-index" methods.

This paper contains what the Georgetown InfoSense group has done in regard to solving the challenges presented by TREC iKAT 2023. Our submitted runs outperform the median runs by a significant margin, exhibiting superior performance in nDCG across various cut numbers and in overall success rate. Our approach uses a Generate-Retrieve-Generate method, which we've found to greatly outpace Retrieve-Then-Generate approaches for the purposes of iKAT. Our solution involves the use of Large Language Models (LLMs) for initial answers, answer grounding by BM25, passage quality filtering by logistic regression, and answer generation by LLMs again. We leverage several purpose-built Language Models, including BERT, Chat-based, and text-to-transfer-based models, for text understanding, classification, generation, and summarization. The official results of the TREC evaluation contradict our initial self-evaluation, which may suggest that a decrease in the reliance on our retrieval and classification methods is better. Nonetheless, our findings suggest that the sequence of involving these different components matters, where we see an essentiality of using LLMs before using search engines.

One long-term goal of machine learning research is to produce methods that are applicable to reasoning and natural language, in particular building an intelligent dialogue agent. To measure progress towards that goal, we argue for the usefulness of a set of proxy tasks that evaluate reading comprehension via question answering. Our tasks measure understanding in several ways: whether a system is able to answer questions via chaining facts, simple induction, deduction and many more. The tasks are designed to be prerequisites for any system that aims to be capable of conversing with a human. We believe many existing learning systems can currently not solve them, and hence our aim is to classify these tasks into skill sets, so that researchers can identify (and then rectify) the failings of their systems. We also extend and improve the recently introduced Memory Networks model, and show it is able to solve some, but not all, of the tasks.

Dense retrieval has become a prominent method to obtain relevant context or world knowledge in open-domain NLP tasks. When we use a learned dense retriever on a retrieval corpus at inference time, an often-overlooked design choice is the retrieval unit in which the corpus is indexed, e.g. document, passage, or sentence. We discover that the retrieval unit choice significantly impacts the performance of both retrieval and downstream tasks. Distinct from the typical approach of using passages or sentences, we introduce a novel retrieval unit, proposition, for dense retrieval. Propositions are defined as atomic expressions within text, each encapsulating a distinct factoid and presented in a concise, self-contained natural language format. We conduct an empirical comparison of different retrieval granularity. Our results reveal that proposition-based retrieval significantly outperforms traditional passage or sentence-based methods in dense retrieval. Moreover, retrieval by proposition also enhances the performance of downstream QA tasks, since the retrieved texts are more condensed with question-relevant information, reducing the need for lengthy input tokens and minimizing the inclusion of extraneous, irrelevant information.

We assessed the performance of commercial Large Language Models (LLMs) GPT-3.5-Turbo and GPT-4 on tasks from the 2023 BioASQ challenge. In Task 11b Phase B, which is focused on answer generation, both models demonstrated competitive abilities with leading systems. Remarkably, they achieved this with simple zero-shot learning, grounded with relevant snippets. Even without relevant snippets, their performance was decent, though not on par with the best systems. Interestingly, the older and cheaper GPT-3.5-Turbo system was able to compete with GPT-4 in the grounded Q&A setting on factoid and list answers. In Task 11b Phase A, focusing on retrieval, query expansion through zero-shot learning improved performance, but the models fell short compared to other systems. The code needed to rerun these experiments is available through GitHub.

Passage ranking involves two stages: passage retrieval and passage re-ranking, which are important and challenging topics for both academics and industries in the area of Information Retrieval (IR). However, the commonly-used datasets for passage ranking usually focus on the English language. For non-English scenarios, such as Chinese, the existing datasets are limited in terms of data scale, fine-grained relevance annotation and false negative issues. To address this problem, we introduce T2Ranking, a large-scale Chinese benchmark for passage ranking. T2Ranking comprises more than 300K queries and over 2M unique passages from real-world search engines. Expert annotators are recruited to provide 4-level graded relevance scores (fine-grained) for query-passage pairs instead of binary relevance judgments (coarse-grained). To ease the false negative issues, more passages with higher diversities are considered when performing relevance annotations, especially in the test set, to ensure a more accurate evaluation. Apart from the textual query and passage data, other auxiliary resources are also provided, such as query types and XML files of documents which passages are generated from, to facilitate further studies. To evaluate the dataset, commonly used ranking models are implemented and tested on T2Ranking as baselines. The experimental results show that T2Ranking is challenging and there is still scope for improvement. The full data and all codes are available at https://github.com/THUIR/T2Ranking/

Using Large Language Models (LLMs) for relevance assessments offers promising opportunities to improve Information Retrieval (IR), Natural Language Processing (NLP), and related fields. Indeed, LLMs hold the promise of allowing IR experimenters to build evaluation collections with a fraction of the manual human labor currently required. This could help with fresh topics on which there is still limited knowledge and could mitigate the challenges of evaluating ranking systems in low-resource scenarios, where it is challenging to find human annotators. Given the fast-paced recent developments in the domain, many questions concerning LLMs as assessors are yet to be answered. Among the aspects that require further investigation, we can list the impact of various components in a relevance judgment generation pipeline, such as the prompt used or the LLM chosen. This paper benchmarks and reports on the results of a large-scale automatic relevance judgment evaluation, the LLMJudge challenge at SIGIR 2024, where different relevance assessment approaches were proposed. In detail, we release and benchmark 42 LLM-generated labels of the TREC 2023 Deep Learning track relevance judgments produced by eight international teams who participated in the challenge. Given their diverse nature, these automatically generated relevance judgments can help the community not only investigate systematic biases caused by LLMs but also explore the effectiveness of ensemble models, analyze the trade-offs between different models and human assessors, and advance methodologies for improving automated evaluation techniques. The released resource is available at the following link: https://llm4eval.github.io/LLMJudge-benchmark/

Knowledge-intensive language tasks (KILTs) benefit from retrieving high-quality relevant contexts from large external knowledge corpora. Learning task-specific retrievers that return relevant contexts at an appropriate level of semantic granularity, such as a document retriever, passage retriever, sentence retriever, and entity retriever, may help to achieve better performance on the end-to-end task. But a task-specific retriever usually has poor generalization ability to new domains and tasks, and it may be costly to deploy a variety of specialised retrievers in practice. We propose a unified generative retriever (UGR) that combines task-specific effectiveness with robust performance over different retrieval tasks in KILTs. To achieve this goal, we make two major contributions: (i) To unify different retrieval tasks into a single generative form, we introduce an n-gram-based identifier for relevant contexts at different levels of granularity in KILTs. And (ii) to address different retrieval tasks with a single model, we employ a prompt learning strategy and investigate three methods to design prompt tokens for each task. In this way, the proposed UGR model can not only share common knowledge across tasks for better generalization, but also perform different retrieval tasks effectively by distinguishing task-specific characteristics. We train UGR on a heterogeneous set of retrieval corpora with well-designed prompts in a supervised and multi-task fashion. Experimental results on the KILT benchmark demonstrate the effectiveness of UGR on in-domain datasets, out-of-domain datasets, and unseen tasks.

Recent work finds that retrieval-augmented generation with large language models is prone to be influenced by the order of retrieved documents in the context. However, the lack of in-depth analysis limits the use of this phenomenon for prompt engineering in practice. In this study, we posit that likelihoods serve as an effective gauge for language model performance. Through experiments on two question-answering datasets with a variety of state-of-the-art language models, we reveal correlations between answer accuracy and the likelihood of the question at both the corpus level and the instance level. In addition, we find that question likelihood can also indicate the position of the task-relevant information in the context. Based on these findings, we propose two methods that use question likelihood as a gauge for selecting and constructing prompts that lead to better performance. We demonstrate their effectiveness with experiments. In addition, our likelihood-based methods are efficient, as they only need to compute the likelihood of the input, requiring much fewer language model passes than heuristic prompt engineering methods that require generating responses. Our analysis deepens our understanding of how input prompts affect model performance and provides a promising direction for efficient prompt optimization.

A key component of successfully reading a passage of text is the ability to apply knowledge gained from the passage to a new situation. In order to facilitate progress on this kind of reading, we present ROPES, a challenging benchmark for reading comprehension targeting Reasoning Over Paragraph Effects in Situations. We target expository language describing causes and effects (e.g., "animal pollinators increase efficiency of fertilization in flowers"), as they have clear implications for new situations. A system is presented a background passage containing at least one of these relations, a novel situation that uses this background, and questions that require reasoning about effects of the relationships in the background passage in the context of the situation. We collect background passages from science textbooks and Wikipedia that contain such phenomena, and ask crowd workers to author situations, questions, and answers, resulting in a 14,322 question dataset. We analyze the challenges of this task and evaluate the performance of state-of-the-art reading comprehension models. The best model performs only slightly better than randomly guessing an answer of the correct type, at 61.6% F1, well below the human performance of 89.0%.

We demonstrate that, with the availability of distributed computation platforms such as Amazon Web Services and open-source tools, it is possible for a small engineering team to build, launch and maintain a cost-effective, large-scale visual search system with widely available tools. We also demonstrate, through a comprehensive set of live experiments at Pinterest, that content recommendation powered by visual search improve user engagement. By sharing our implementation details and the experiences learned from launching a commercial visual search engines from scratch, we hope visual search are more widely incorporated into today's commercial applications.

Interest in neural machine translation has grown rapidly as its effectiveness has been demonstrated across language and data scenarios. New research regularly introduces architectural and algorithmic improvements that lead to significant gains over "vanilla" NMT implementations. However, these new techniques are rarely evaluated in the context of previously published techniques, specifically those that are widely used in state-of-theart production and shared-task systems. As a result, it is often difficult to determine whether improvements from research will carry over to systems deployed for real-world use. In this work, we recommend three specific methods that are relatively easy to implement and result in much stronger experimental systems. Beyond reporting significantly higher BLEU scores, we conduct an in-depth analysis of where improvements originate and what inherent weaknesses of basic NMT models are being addressed. We then compare the relative gains afforded by several other techniques proposed in the literature when starting with vanilla systems versus our stronger baselines, showing that experimental conclusions may change depending on the baseline chosen. This indicates that choosing a strong baseline is crucial for reporting reliable experimental results.

Did you try out the new Bing Search? Or maybe you fiddled around with Google AI~Overviews? These might sound familiar because the modern-day search stack has recently evolved to include retrieval-augmented generation (RAG) systems. They allow searching and incorporating real-time data into large language models (LLMs) to provide a well-informed, attributed, concise summary in contrast to the traditional search paradigm that relies on displaying a ranked list of documents. Therefore, given these recent advancements, it is crucial to have an arena to build, test, visualize, and systematically evaluate RAG-based search systems. With this in mind, we propose the TREC 2024 RAG Track to foster innovation in evaluating RAG systems. In our work, we lay out the steps we've made towards making this track a reality -- we describe the details of our reusable framework, Ragnar\"ok, explain the curation of the new MS MARCO V2.1 collection choice, release the development topics for the track, and standardize the I/O definitions which assist the end user. Next, using Ragnar\"ok, we identify and provide key industrial baselines such as OpenAI's GPT-4o or Cohere's Command R+. Further, we introduce a web-based user interface for an interactive arena allowing benchmarking pairwise RAG systems by crowdsourcing. We open-source our Ragnar\"ok framework and baselines to achieve a unified standard for future RAG systems.

We introduce Legal RAG Bench, a benchmark and evaluation methodology for assessing the end-to-end performance of legal RAG systems. As a benchmark, Legal RAG Bench consists of 4,876 passages from the Victorian Criminal Charge Book alongside 100 complex, hand-crafted questions demanding expert knowledge of criminal law and procedure. Both long-form answers and supporting passages are provided. As an evaluation methodology, Legal RAG Bench leverages a full factorial design and novel hierarchical error decomposition framework, enabling apples-to-apples comparisons of the contributions of retrieval and reasoning models in RAG. We evaluate three state-of-the-art embedding models (Isaacus' Kanon 2 Embedder, Google's Gemini Embedding 001, and OpenAI's Text Embedding 3 Large) and two frontier LLMs (Gemini 3.1 Pro and GPT-5.2), finding that information retrieval is the primary driver of legal RAG performance, with LLMs exerting a more moderate effect on correctness and groundedness. Kanon 2 Embedder, in particular, had the largest positive impact on performance, improving average correctness by 17.5 points, groundedness by 4.5 points, and retrieval accuracy by 34 points. We observe that many errors attributed to hallucinations in legal RAG systems are in fact triggered by retrieval failures, concluding that retrieval sets the ceiling for the performance of many modern legal RAG systems. We document why and how we built Legal RAG Bench alongside the results of our evaluations. We also openly release our code and data to assist with reproduction of our findings.

We introduce a large scale MAchine Reading COmprehension dataset, which we name MS MARCO. The dataset comprises of 1,010,916 anonymized questions---sampled from Bing's search query logs---each with a human generated answer and 182,669 completely human rewritten generated answers. In addition, the dataset contains 8,841,823 passages---extracted from 3,563,535 web documents retrieved by Bing---that provide the information necessary for curating the natural language answers. A question in the MS MARCO dataset may have multiple answers or no answers at all. Using this dataset, we propose three different tasks with varying levels of difficulty: (i) predict if a question is answerable given a set of context passages, and extract and synthesize the answer as a human would (ii) generate a well-formed answer (if possible) based on the context passages that can be understood with the question and passage context, and finally (iii) rank a set of retrieved passages given a question. The size of the dataset and the fact that the questions are derived from real user search queries distinguishes MS MARCO from other well-known publicly available datasets for machine reading comprehension and question-answering. We believe that the scale and the real-world nature of this dataset makes it attractive for benchmarking machine reading comprehension and question-answering models.

Making the relevance judgments for a TREC-style test collection can be complex and expensive. A typical TREC track usually involves a team of six contractors working for 2-4 weeks. Those contractors need to be trained and monitored. Software has to be written to support recording relevance judgments correctly and efficiently. The recent advent of large language models that produce astoundingly human-like flowing text output in response to a natural language prompt has inspired IR researchers to wonder how those models might be used in the relevance judgment collection process. At the ACM SIGIR 2024 conference, a workshop ``LLM4Eval'' provided a venue for this work, and featured a data challenge activity where participants reproduced TREC deep learning track judgments, as was done by Thomas et al (arXiv:2408.08896, arXiv:2309.10621). I was asked to give a keynote at the workshop, and this paper presents that keynote in article form. The bottom-line-up-front message is, don't use LLMs to create relevance judgments for TREC-style evaluations.

We introduce Browsing Lost Unformed Recollections, a tip-of-the-tongue known-item search and reasoning benchmark for general AI assistants. BLUR introduces a set of 573 real-world validated questions that demand searching and reasoning across multi-modal and multilingual inputs, as well as proficient tool use, in order to excel on. Humans easily ace these questions (scoring on average 98%), while the best-performing system scores around 56%. To facilitate progress toward addressing this challenging and aspirational use case for general AI assistants, we release 350 questions through a public leaderboard, retain the answers to 250 of them, and have the rest as a private test set.

This is the first year of the TREC Neural CLIR (NeuCLIR) track, which aims to study the impact of neural approaches to cross-language information retrieval. The main task in this year's track was ad hoc ranked retrieval of Chinese, Persian, or Russian newswire documents using queries expressed in English. Topics were developed using standard TREC processes, except that topics developed by an annotator for one language were assessed by a different annotator when evaluating that topic on a different language. There were 172 total runs submitted by twelve teams.

Although LLMs are increasing the productivity of professional programmers, existing work shows that beginners struggle to prompt LLMs to solve text-to-code tasks. Why is this the case? This paper explores two competing hypotheses about the cause of student-LLM miscommunication: (1) students simply lack the technical vocabulary needed to write good prompts, and (2) students do not understand the extent of information that LLMs need to solve code generation tasks. We study (1) with a causal intervention experiment on technical vocabulary and (2) by analyzing graphs that abstract how students edit prompts and the different failures that they encounter. We find that substance beats style: a poor grasp of technical vocabulary is merely correlated with prompt failure; that the information content of prompts predicts success; that students get stuck making trivial edits; and more. Our findings have implications for the use of LLMs in programming education, and for efforts to make computing more accessible with LLMs.

We propose several small modifications to Duet---a deep neural ranking model---and evaluate the updated model on the MS MARCO passage ranking task. We report significant improvements from the proposed changes based on an ablation study.

Outcome RL raises pass@1 but not pass@k at large k (Yue et al., 2025). But every pass@1 is scored under a token budget. Every recipe also trains under one. On physics and math problems at fixed token budgets, we compare the pass@1 of 20 released RL models (1.5B to 32B) with that of the thinking models they started from. We ask whether each pass@1 gain is changed behaviour or new knowledge. We divide each change in pass@1 into three parts. Finishing measures answering before the budget runs out. Reliability measures being right once finished. Reach measures problems RL solves that its base answered wrongly in all eight tries, minus problems the base solves that RL answered wrongly in all eight. We find four things. (1) Where the base often hits the token budget before answering, most of the change, gain or loss, is finishing (82% at 4k and 8k). Given more tokens, the base closes most of the finishing gap itself. (2) Where the base has enough tokens, the gain is reliability: the RL model is graded right more often, even on problems both finish. (3) By our definition, RL rarely solves what its base cannot. Reach adds at most 2.1 percentage points on any of our four physics and math sets. We checked 144 physics problems that only the RL model solved. Of these, 3.5% hold up when resampled or read. In a random sample, nine of ten new RL steps a human confirmed also appear in the base's samples. (4) In a small-scale RL run on an 8B model, the gain on new problems is finishing, with reliability flat. Supplying each problem's physical setup was not shown to improve on plain RL.

The paper focuses on deep learning semantic search algorithms applied in the HR domain. The aim of the article is developing a novel approach to training a Siamese network to link the skills mentioned in the job ad with the title. It has been shown that the title normalization process can be based either on classification or similarity comparison approaches. While classification algorithms strive to classify a sample into predefined set of categories, similarity search algorithms take a more flexible approach, since they are designed to find samples that are similar to a given query sample, without requiring pre-defined classes and labels. In this article semantic similarity search to find candidates for title normalization has been used. A pre-trained language model has been adapted while teaching it to match titles and skills based on co-occurrence information. For the purpose of this research fifty billion title-descriptions pairs had been collected for training the model and thirty three thousand title-description-normalized title triplets, where normalized job title was picked up manually by job ad creator for testing purposes. As baselines FastText, BERT, SentenceBert and JobBert have been used. As a metric of the accuracy of the designed algorithm is Recall in top one, five and ten model's suggestions. It has been shown that the novel training objective lets it achieve significant improvement in comparison to other generic and specific text encoders. Two settings with treating titles as standalone strings, and with included skills as additional features during inference have been used and the results have been compared in this article. Improvements by 10% and 21.5% have been achieved using VacancySBERT and VacancySBERT (with skills) respectively. The benchmark has been developed as open-source to foster further research in the area.

Modern web interfaces are unnecessarily complex to use as they overwhelm users with excessive text and visuals unrelated to their current goals. This problem particularly impacts screen reader users (SRUs), who navigate content sequentially and may spend minutes traversing irrelevant elements before reaching desired information compared to vision users (VUs) who visually skim in seconds. We present Task Mode, a system that dynamically filters web content based on user-specified goals using large language models to identify and prioritize relevant elements while minimizing distractions. Our approach preserves page structure while offering multiple viewing modes tailored to different access needs. Our user study with 12 participants (6 VUs, 6 SRUs) demonstrates that our approach reduced task completion time for SRUs while maintaining performance for VUs, decreasing the completion time gap between groups from 2x to 1.2x. 11 of 12 participants wanted to use Task Mode in the future, reporting that Task Mode supported completing tasks with less effort and fewer distractions. This work demonstrates how designing new interactions simultaneously for visual and non-visual access can reduce rather than reinforce accessibility disparities in future technology created by human-computer interaction researchers and practitioners.

We present TransEvalnia, a prompting-based translation evaluation and ranking system that uses reasoning in performing its evaluations and ranking. This system presents fine-grained evaluations based on a subset of the Multidimensional Quality Metrics (https://themqm.org/), returns an assessment of which translation it deems the best, and provides numerical scores for the various dimensions and for the overall translation. We show that TransEvalnia performs as well as or better than the state-of-the-art MT-Ranker (Moosa et al. 2024) on our own English-Japanese data as well as several language pairs from various WMT shared tasks. Using Anthropic's Claude-3.5-Sonnet and Qwen-2.5-72B-Instruct as the evaluation LLMs, we show that the evaluations returned are deemed highly acceptable to human raters, and that the scores assigned to the translations by Sonnet, as well as other LLMs, correlate well with scores assigned by the human raters. We also note the sensitivity of our system -- as well as MT-Ranker -- to the order in which the translations are presented, and we propose methods to address this position bias. All data, including the system's evaluation and reasoning, human assessments, as well as code is released.

This paper presents the overview of the AuTexTification shared task as part of the IberLEF 2023 Workshop in Iberian Languages Evaluation Forum, within the framework of the SEPLN 2023 conference. AuTexTification consists of two subtasks: for Subtask 1, participants had to determine whether a text is human-authored or has been generated by a large language model. For Subtask 2, participants had to attribute a machine-generated text to one of six different text generation models. Our AuTexTification 2023 dataset contains more than 160.000 texts across two languages (English and Spanish) and five domains (tweets, reviews, news, legal, and how-to articles). A total of 114 teams signed up to participate, of which 36 sent 175 runs, and 20 of them sent their working notes. In this overview, we present the AuTexTification dataset and task, the submitted participating systems, and the results.

A law practitioner has to go through numerous lengthy legal case proceedings for their practices of various categories, such as land dispute, corruption, etc. Hence, it is important to summarize these documents, and ensure that summaries contain phrases with intent matching the category of the case. To the best of our knowledge, there is no evaluation metric that evaluates a summary based on its intent. We propose an automated intent-based summarization metric, which shows a better agreement with human evaluation as compared to other automated metrics like BLEU, ROUGE-L etc. in terms of human satisfaction. We also curate a dataset by annotating intent phrases in legal documents, and show a proof of concept as to how this system can be automated. Additionally, all the code and data to generate reproducible results is available on Github.

To facilitate research in the direction of fine-tuning foundation models from human feedback, we held the MineRL BASALT Competition on Fine-Tuning from Human Feedback at NeurIPS 2022. The BASALT challenge asks teams to compete to develop algorithms to solve tasks with hard-to-specify reward functions in Minecraft. Through this competition, we aimed to promote the development of algorithms that use human feedback as channels to learn the desired behavior. We describe the competition and provide an overview of the top solutions. We conclude by discussing the impact of the competition and future directions for improvement.

In populous countries, pending legal cases have been growing exponentially. There is a need for developing NLP-based techniques for processing and automatically understanding legal documents. To promote research in the area of Legal NLP we organized the shared task LegalEval - Understanding Legal Texts at SemEval 2023. LegalEval task has three sub-tasks: Task-A (Rhetorical Roles Labeling) is about automatically structuring legal documents into semantically coherent units, Task-B (Legal Named Entity Recognition) deals with identifying relevant entities in a legal document and Task-C (Court Judgement Prediction with Explanation) explores the possibility of automatically predicting the outcome of a legal case along with providing an explanation for the prediction. In total 26 teams (approx. 100 participants spread across the world) submitted systems paper. In each of the sub-tasks, the proposed systems outperformed the baselines; however, there is a lot of scope for improvement. This paper describes the tasks, and analyzes techniques proposed by various teams.

Reinforcement learning (RL) has demonstrated strong potential in training large language models (LLMs) capable of complex reasoning for real-world problem solving. More recently, RL has been leveraged to create sophisticated LLM-based search agents that adeptly combine reasoning with search engine use. While the use of RL for training search agents is promising, the optimal design of such agents remains not fully understood. In particular, key factors -- such as (1) reward formulation, (2) the choice and characteristics of the underlying LLM, and (3) the role of the search engine in the RL process -- require further investigation. In this work, we conduct comprehensive empirical studies to systematically investigate these and offer actionable insights. We highlight several key findings: format rewards are effective in improving final performance, whereas intermediate retrieval rewards have limited impact; the scale and initialization of the LLM (general-purpose vs. reasoning-specialized) significantly influence RL outcomes; and the choice of search engine plays a critical role in shaping RL training dynamics and the robustness of the trained agent during inference. These establish important guidelines for successfully building and deploying LLM-based search agents in real-world applications. Code is available at https://github.com/PeterGriffinJin/Search-R1.

We describe our single submission to task 1 of COLIEE 2021. Our vanilla BM25 got second place, well above the median of submissions. Code is available at https://github.com/neuralmind-ai/coliee.

Many NLP tasks, such as summarization and extractive question answering, reduce to retrieving relevant content from documents under two constraints: coverage, retaining enough pertinent information to achieve some goal, and conciseness, removing as much irrelevant information as possible. Conformal prediction methods have been used to guarantee coverage, and must be optimized for conciseness through design of a score function. State-of-the-art scoring functions use hand-engineered LLM prompts asking the model to rate the importance of content, but manual prompt engineering is labor-intensive and task-specific. We introduce the Conformal Relevance framework which uses in-context learning example curation and ensembling to create a score function which maintains coverage while improving conciseness with minimal manual input. We demonstrate this framework's application on seven NLP tasks, and also theoretically study the impact of diversity for ensembled conformal scores, giving a complementarity condition that characterizes when ensembling improves worst-case sentence scores, and a saturation bound on ensemble improvement.

Deep search agents answer difficult information-seeking questions by iteratively issuing search queries to gather supporting evidence, but it remains unclear whether and how greater search effort leads to better answers. We study these questions through a trajectory-level diagnosis of long-horizon search agents. Using human-annotated document-level relevance judgments, we evaluate the evidence retrieved at each search step and separate two stages of agent behavior: what evidence an agent retrieves and how effectively it uses that evidence. This distinction further allows us to decompose failures into retrieval gaps, where the necessary evidence is never found, and utilization gaps, where relevant evidence is retrieved but not used correctly. With the retrieval model and evaluation harness held fixed, we compare six agents on BrowseComp-Plus and further validate our findings on BrowseComp with an open-web search API. Across settings, we find that search effort and answer quality are only weakly aligned. Answer accuracy is better correlated with the quality of retrieved evidence, especially cumulative retrieval recall, than with the number of searches or the amount of context consumed. Useful evidence often appears early in the trajectory, yet agents tend to continue searching, producing a long tail of low-yield retrieval steps. At the query level, exploratory reformulations remain useful, but the best-performing agents issue far fewer redundant queries. Overall, by systematically characterizing the search behavior and failure modes of long-horizon search agents, this work points to practical directions for building better deep research systems, including stronger query formulation, more effective evidence selection and context management, and stopping criteria based on whether sufficient supporting evidence has been retrieved.

Text generation has become more accessible than ever, and the increasing interest in these systems, especially those using large language models, has spurred an increasing number of related publications. We provide a systematic literature review comprising 244 selected papers between 2017 and 2024. This review categorizes works in text generation into five main tasks: open-ended text generation, summarization, translation, paraphrasing, and question answering. For each task, we review their relevant characteristics, sub-tasks, and specific challenges (e.g., missing datasets for multi-document summarization, coherence in story generation, and complex reasoning for question answering). Additionally, we assess current approaches for evaluating text generation systems and ascertain problems with current metrics. Our investigation shows nine prominent challenges common to all tasks and sub-tasks in recent text generation publications: bias, reasoning, hallucinations, misuse, privacy, interpretability, transparency, datasets, and computing. We provide a detailed analysis of these challenges, their potential solutions, and which gaps still require further engagement from the community. This systematic literature review targets two main audiences: early career researchers in natural language processing looking for an overview of the field and promising research directions, as well as experienced researchers seeking a detailed view of tasks, evaluation methodologies, open challenges, and recent mitigation strategies.

Large language models (LLMs) have made significant advancements in various natural language processing tasks, including question answering (QA) tasks. While incorporating new information with the retrieval of relevant passages is a promising way to improve QA with LLMs, the existing methods often require additional fine-tuning which becomes infeasible with recent LLMs. Augmenting retrieved passages via prompting has the potential to address this limitation, but this direction has been limitedly explored. To this end, we design a simple yet effective framework to enhance open-domain QA (ODQA) with LLMs, based on the summarized retrieval (SuRe). SuRe helps LLMs predict more accurate answers for a given question, which are well-supported by the summarized retrieval that could be viewed as an explicit rationale extracted from the retrieved passages. Specifically, SuRe first constructs summaries of the retrieved passages for each of the multiple answer candidates. Then, SuRe confirms the most plausible answer from the candidate set by evaluating the validity and ranking of the generated summaries. Experimental results on diverse ODQA benchmarks demonstrate the superiority of SuRe, with improvements of up to 4.6% in exact match (EM) and 4.0% in F1 score over standard prompting approaches. SuRe also can be integrated with a broad range of retrieval methods and LLMs. Finally, the generated summaries from SuRe show additional advantages to measure the importance of retrieved passages and serve as more preferred rationales by models and humans.