Daily Papers

We present MineNPC-Task, a user-authored benchmark and evaluation harness for testing memory-aware, mixed-initiative LLM agents in open-world Minecraft. Rather than relying on synthetic prompts, tasks are elicited through formative and summative co-play with expert players, then normalized into parametric templates with explicit preconditions and dependency structure. These tasks are paired with machine-checkable validators under a bounded-knowledge policy that forbids out-of-world shortcuts. The harness captures plan, action, and memory events, including plan previews, targeted clarifications, memory reads and writes, precondition checks, and repair attempts, and reports outcomes relative to the total number of attempted subtasks using only in-world evidence. As an initial snapshot, we instantiate the framework with GPT-4o and evaluate 216 subtasks across 8 experienced players. We observe recurring breakdown patterns in code execution, inventory and tool handling, referencing, and navigation, alongside successful recoveries supported by mixed-initiative clarifications and lightweight memory use. Participants rated interaction quality and interface usability positively, while noting the need for stronger memory persistence across tasks. We release the complete task suite, validators, logs, and evaluation harness to support transparent and reproducible evaluation of future memory-aware embodied agents.

Foundation-model policies for robotic manipulation are advancing rapidly on task success, but rigorous evaluation of whether they succeed safely is still lacking. We introduce ManiGuard, a specification-grounded framework for evaluating and improving the safety of foundation-model manipulation, comprising the ManiGuard-Bench task suite and a paired safety-annotated trajectory-generation pipeline. ManiGuard-Bench organizes six contact-rich household task families into 200 locked base tasks along a skill times constraint taxonomy, with safety specified independently of task success. Each task is evaluated under one in-distribution and four single-axis out-of-distribution perturbations that hold the safety specification fixed, giving 1,000 locked scenarios. Every rollout is runtime-checked by LTL_f-grounded automaton monitors over physics-grounded predicates rather than learned classifiers or LLM judges, in simulation and on a physical Franka platform. The pipeline pairs an automated motion-planning generator with human teleoperation, annotated by the same per-step monitor, and directly supports safety-aware fine-tuning; we release 8,000 safety-annotated demonstrations, 40 per base task. Benchmarking zero-shot and fine-tuned VLAs across more than 23,000 rollouts, we find: (i) safety must be evaluated independently of task success, as 6-21% of successful rollouts violate the specification; (ii) fine-tuning on our suite raises safe task completion from near zero to 7.5-29.8% and engaged-and-safe behavior from 16-40% to 51-72%; but (iii) a gap remains that scaling demonstrations does not close, with 21-42% of engaged rollouts still violating, two of six families below 2% safe success for every policy, and these failures persisting under distribution shift and on hardware.

Conformal prediction guarantees marginal coverage, but a single calibration threshold can still spread that coverage unevenly, over-covering easy regions and under-covering hard ones. SimplexUQ is, to our knowledge, the first benchmark and reproducible protocol for measuring this allocation problem on simplex-valued predictions; it compares existing conformal wrappers rather than proposing a new one. Its task suite, SimplexTasks-12, combines six controlled synthetic regimes with six frozen-predictor real tasks spanning class probabilities, topic mixtures, spectral abundances, cell-type fractions, age distributions, and emotion mixtures. Each comparison fixes the predictor, score, and response-free stratification map, varies only the wrapper, and reports marginal coverage, worst-stratum coverage, max disparity, and within-task radius and compute. Global calibration can look valid while failing badly: on CIFAR-10 it attains 0.900 marginal coverage but only 0.542 in the worst entropy stratum, and Mondrian calibration raises that stratum to 0.886 while reducing max disparity from 0.358 to 0.022. No wrapper dominates, however. Under smooth synthetic heterogeneity, several repairs are competitive; fixed-map analyses show that rankings depend on the evaluation groups and protocol; and in a 12-task comparison, Mondrian has lower disparity on its single target partition for all 12 tasks, whereas BatchMVP has lower disparity over overlapping groups on five. These are empirical comparisons, not new coverage guarantees. A controlled predictor-bias sweep shows that removing predictor bias only partly reduces global-threshold disparity. We release task cards, result provenance, permitted derived arrays, and rebuild instructions, and treat wrapper selection as a diagnostic comparison rather than a universal ranking.

Capturing a loose object inside a spacecraft requires a robot to approach a target that can translate and rotate, and whose motion changes when contact is made. These conditions motivate evaluation beyond manipulation of supported objects. We introduce AstroBench, an open benchmark for free-floating capture and space robotic manipulation with a fixed-base arm. The task suite combines four initial-motion tracks with sample-storage operations based on MELFI hardware and cap and captive-screw servicing based on the Robotic Refueling Mission. The two workcells provide nine task definitions and 29 task-target combinations. A metadata-filtered Objaverse inventory supplies 8,414 object instances for asset expansion, alongside five cabin visual contexts and versioned task instructions. We report two acquisition-focused experiments using existing checkpoints. A frozen π_{0.5} DROID joint-position policy captures 2 of 200 static targets and 4 of 200 moving targets across twenty object identities. A cube-specific ACT policy captures 8 of 10 static targets and 6 of 10 moving targets in a separate experiment. Each experiment retains its recorded observation, action and capture configuration. The results document difficult direct transfer for the tested generalist checkpoint and successful task-specific capture behavior, while identifying concrete settings for studying motion-aware perception and contact-sensitive control. Code, task definitions and editable workcell assets are available at https://github.com/AstroUpLab/AstroBench.

Three challenges limit the progress of robot learning research: robots are expensive (few labs can participate), everyone uses different robots (findings do not generalize across labs), and we lack internet-scale robotics data. We take on these challenges via a new benchmark: Train Offline, Test Online (TOTO). TOTO provides remote users with access to shared robotic hardware for evaluating methods on common tasks and an open-source dataset of these tasks for offline training. Its manipulation task suite requires challenging generalization to unseen objects, positions, and lighting. We present initial results on TOTO comparing five pretrained visual representations and four offline policy learning baselines, remotely contributed by five institutions. The real promise of TOTO, however, lies in the future: we release the benchmark for additional submissions from any user, enabling easy, direct comparison to several methods without the need to obtain hardware or collect data.

Mathematical reasoning---a core ability within human intelligence---presents some unique challenges as a domain: we do not come to understand and solve mathematical problems primarily on the back of experience and evidence, but on the basis of inferring, learning, and exploiting laws, axioms, and symbol manipulation rules. In this paper, we present a new challenge for the evaluation (and eventually the design) of neural architectures and similar system, developing a task suite of mathematics problems involving sequential questions and answers in a free-form textual input/output format. The structured nature of the mathematics domain, covering arithmetic, algebra, probability and calculus, enables the construction of training and test splits designed to clearly illuminate the capabilities and failure-modes of different architectures, as well as evaluate their ability to compose and relate knowledge and learned processes. Having described the data generation process and its potential future expansions, we conduct a comprehensive analysis of models from two broad classes of the most powerful sequence-to-sequence architectures and find notable differences in their ability to resolve mathematical problems and generalize their knowledge.

One promise that Vision-Language-Action (VLA) models hold over traditional imitation learning for robotics is to leverage the broad generalization capabilities of large Vision-Language Models (VLMs) to produce versatile, "generalist" robot policies. However, current evaluations of VLAs remain insufficient. Traditional imitation learning benchmarks are unsuitable due to the lack of language instructions. Emerging benchmarks for VLAs that incorporate language often come with limited evaluation tasks and do not intend to investigate how much VLM pretraining truly contributes to the generalization capabilities of the downstream robotic policy. Meanwhile, much research relies on real-world robot setups designed in isolation by different institutions, which creates a barrier for reproducibility and accessibility. To address this gap, we introduce a unified probing suite of 50 simulation-based tasks across 10 subcategories spanning language instruction, vision, and objects. We systematically evaluate several state-of-the-art VLA architectures on this suite to understand their generalization capability. Our results show that while VLM backbones endow VLAs with robust perceptual understanding and high level planning, which we refer to as good intentions, this does not reliably translate into precise motor execution: when faced with out-of-distribution observations, policies often exhibit coherent intentions, but falter in action execution. Moreover, finetuning on action data can erode the original VLM's generalist reasoning abilities. We release our task suite and evaluation code to serve as a standardized benchmark for future VLAs and to drive research on closing the perception-to-action gap. More information, including the source code, can be found at https://ai4ce.github.io/INT-ACT/

Recursive self-improvement (RSI) asks whether an AI system can improve the process that produces AI systems, so that the next system inherits the improvement. That process is the training algorithm: a better objective or update rule improves the compute-capability exchange rate for every subsequent run, including the one that produces the next agent. Whether RSI is feasible therefore turns on whether an agent can design training algorithms. No benchmark isolates that ability: existing suites are won by collecting data or by tuning hyperparameters, and none tells a change to how a run is executed apart from a change to how the model learns. We present AI4AI-Bench, 10 frozen research repositories spanning 10 training algorithm families. In each task, an agent has 4 hours on one B300 to rewrite the training algorithm; its code is then rerun from scratch for up to 12 hours and scored by a fixed evaluator hidden from the agent, against the repository's original algorithm under the same procedure. Because the 10 metrics are incommensurable, every task is mapped onto one scale on which 0 is an uninformative model, 0.1 is the algorithm the repository ships, and 1.0 is the task optimum. Across 29 configurations of 6 systems on all 10 tasks the mean score is 0.166, and the best system reaches 0.250: even the strongest closes under a fifth of the distance between the algorithm that was already there and the optimum. The submissions show where that distance went: most never change how the model learns at all, and the minority that do average 0.226 against 0.126 for the rest. More reasoning effort mostly buys the willingness to go there, taking that minority from 8% of submissions to 64% and the mean score from 0.094 to 0.196. We release the task suite, the evaluators and every scored submission, so that the measurement can be repeated as these systems change.

Robot foundation models are trained and evaluated predominantly in English, and robot demonstration corpora do not exist for most languages. We study the addition of Greek to an open vision-language-action stack using only machine-rephrased instructions and no architecture changes. The main challenge is measurement rather than translation. Several plausible instruments produce false conclusions: a color-histogram metric rewards noise, a single-goal benchmark scores 84.6% under correct Greek and 82.6% under deliberately wrong instructions, training loss fails to predict Greek success, and single-run comparisons are dominated by seed variation. On a discriminative ninety-task suite with three seeds per arm, a multilingual text tower without Greek demonstrations remains at its wrong-instruction floor, while Greek-only training exceeds its control by at most 2.7 points. Bilingual training yields a consistent 6.7-7.1 point margin over its control and reaches about two fifths of English performance. The policy also overfits the translator's phrasing; training on seven phrasings per task approximately halves this penalty. Warm-starting from a language-adapted world model and unfreezing the text tower both degrade performance. The results support two practical requirements for low-resource robot-policy localization: build a guaranteed null before trusting a metric, and replicate low-resource-language results across seeds.

Making an open-weight language model generate text is not the same as making it finish coding work. This retrospective empirical systems experience report follows my engineering work from July to October 2026 across coding clients, model-specific serving recipes, vLLM, SGLang, llama.cpp, and MLX. Models behind a nominally compatible endpoint still differed in tool-call formatting, turn control, context behavior, cache reuse, and reliability. A served model, coherent text, executed tool calls, and code that passes independent behavioral checks proved to be distinct levels of progress. The evidence includes 4,165 coding-agent gateway turns from my own sessions, 623 debug jobs, and three separate delegation studies with hidden behavioral checks and a fixed Claude Opus 5.5 parent. Saved cold-prefix requests had a median time to first token of about 69.4 s, versus 2.8 s for high-reuse requests (observational). In a 45-trial study, direct coding passed 15/15 and delegation to a quantized Qwen3-Coder 30B-A3B worker passed 12/15 and 14/15. On a second 15-task suite under a fixed artifact contract, the completed replacement cohort had 45/45 for direct coding and 12/45 for a Nemotron Cascade 2 worker (11 worker substitutions). Sixteen sessions stopped by a subscription limit before any model call were replaced; counting them as failures gives 37/45 and 8/45. A Qwen3-Coder worker passed 0/45. In both studies with a direct arm, direct coding had lower API list-price-equivalent frontier cost per verified success. A post-hoc diagnostic found that 19 of 36 artifacts rejected only for renamed parameters passed the hidden checks unchanged; they are not counted as successes. The contribution is an evidence-qualified account of serving, agent protocol, session continuity, artifact validation, and verification in practice. The studies are not pooled and do not rank models or runtimes.

Vision-Language-Action (VLA) models achieve strong performance on standard manipulation benchmarks, but most evaluations assume that task-relevant objects are fully visible. This assumption often fails in realistic settings, where occlusion makes manipulation partially observable. In this paper, we study scene-induced occlusion as a fundamental challenge for VLA models and introduce LIBERO-Occ, an occlusion-oriented extension of LIBERO. Experiments show that state-of-the-art VLAs suffer substantial performance degradation under occlusion. To address this issue, we propose Viewpoint Imagination (VIM), which generates a complementary view from an occluded primary observation and conditions action prediction on both observed and imagined evidence. VIM improves robustness across task suites, occlusion types, and severity levels without requiring additional cameras at deployment time, suggesting that viewpoint imagination is an promising mechanism for perception completion in partially observable manipulation. Our benchmark and corresponding code are available at: https://github.com/litsh/Libero-Occ{https://github.com/litsh/Libero-Occ}.

When a large language model (LLM) agent executes the same task, token consumption can vary by over an order of magnitude across runs. The agent chooses its next steps based on tool feedback and intermediate results, while the growing context steadily inflates the input size of every subsequent call. The total consumption of a task is therefore hard to predict before execution and the prediction must be revised as the run unfolds. In this paper, we propose TokenCast, which learns a composable cost representation for each execution segment, recording its own consumption and the context growth it introduces. Composing adjacent segments yields a cumulative estimate that captures the extra input cost incurred when context from earlier segments is re-read by every later call. As execution unfolds, newly observed evidence refreshes the forecast, requiring no additional LLM calls and incurring a mean cumulative prediction time of 32.8 ms per run on SWE-bench Verified. Across 4 task suites and 6 agent models, TokenCast's mean absolute error reduction against the strongest comparator averages 14.5% over 96 evaluated combinations. In offline budget-control replay, TokenCast uses 21.3% fewer tokens on average than a fixed-budget policy at matched trace completion. The code is available at https://github.com/DEFENSE-SEU/TokenCast.

Personalization is essential for general-purpose robots to transition into household environments and achieve true practicality. While general robotic manipulation has been extensively studied, personalized robotic manipulation (PeRM) lacks a controlled benchmark for separating task completion from preference satisfaction. We introduce LIBERO-PeRM, a benchmark for preference learning in embodied agents. LIBERO-PeRM studies five questions: 1) how policies follow preferences in zero-shot settings; 2) how they learn single preferences from demonstrations; 3) how efficiently they acquire preferences as demonstrations increase; 4) how learned preferences generalize across layouts and tasks; and 5) how policies compose preferences and behave under conflicts. We propose a controlled formulation and taxonomy of observable preferences in simulation, covering motion, skill, and user-profile choices. This formulation is restricted to language-expressible preferences that can be instantiated in LIBERO without changing the task goal, evaluated automatically, and reused across tasks. Building on LIBERO, our procedural pipeline generates paired demonstrations with and without preferences. We create five task suites (312 tasks) with high-quality demonstrations, rich metadata, and preference-satisfaction metrics. Experiments show that current VLA policies struggle with zero-shot personalization, while SFT makes preferences learnable. Mixed training achieves better no-preference retention than sequential adaptation and improves performance with contextual prompts compared with preference-only training, but can weaken motion-preference learning; preference generalization, decomposition, and conflict handling remain difficult. The code, datasets, and demos can be found on the project website: https://libero-perm.github.io/.

We present HourVideo, a benchmark dataset for hour-long video-language understanding. Our dataset consists of a novel task suite comprising summarization, perception (recall, tracking), visual reasoning (spatial, temporal, predictive, causal, counterfactual), and navigation (room-to-room, object retrieval) tasks. HourVideo includes 500 manually curated egocentric videos from the Ego4D dataset, spanning durations of 20 to 120 minutes, and features 12,976 high-quality, five-way multiple-choice questions. Benchmarking results reveal that multimodal models, including GPT-4 and LLaVA-NeXT, achieve marginal improvements over random chance. In stark contrast, human experts significantly outperform the state-of-the-art long-context multimodal model, Gemini Pro 1.5 (85.0% vs. 37.3%), highlighting a substantial gap in multimodal capabilities. Our benchmark, evaluation toolkit, prompts, and documentation are available at https://hourvideo.stanford.edu

Professional agent tasks often depend on conventions that are absent from public corpora, yet benchmarks rarely control whether an agent has access to those conventions. We introduce a knowledge-gated task-construction protocol that separates a task instruction from a compact artefact containing private conventions, reference tables, and utility operators. Construction-time provenance, byte-identical task instructions across the provided- and withheld-artefact conditions, leak audits, and executable witnesses make dependence on the artefact explicit and testable. Across fifteen calibration tasks, one frontier agent configuration achieves a 68.0% pass rate with the artefact and 0% without it; on one task, a plausible but incorrect artefact also yields 0% across five trials. Deterministic solvers and rule corpora provide exact ground truth for structured tasks, while named criterion-level rubrics support outputs that cannot be checked by a single executable oracle. A configuration-relative calibration screen retains seven tasks satisfying our five-trial empirical knowledge-gating screen. These experiments validate the behavior of the construction protocol; they do not establish that the retained tasks improve post-training. We publicly release part of the task suite and supporting tooling at https://github.com/DatagridsAI/Knowledge-Gated-Task-Construction.

In this work, we introduce dual goal representations for goal-conditioned reinforcement learning (GCRL). A dual goal representation characterizes a state by "the set of temporal distances from all other states"; in other words, it encodes a state through its relations to every other state, measured by temporal distance. This representation provides several appealing theoretical properties. First, it depends only on the intrinsic dynamics of the environment and is invariant to the original state representation. Second, it contains provably sufficient information to recover an optimal goal-reaching policy, while being able to filter out exogenous noise. Based on this concept, we develop a practical goal representation learning method that can be combined with any existing GCRL algorithm. Through diverse experiments on the OGBench task suite, we empirically show that dual goal representations consistently improve offline goal-reaching performance across 20 state- and pixel-based tasks.

Enterprise agent programs are moving from prototypes into production, where reusable skills, tools, and workflow packages must be reviewed with evidence rather than prose. Current gates often scan these artifacts for structure, style, and security, but they do not answer the deployment question: does the capability package help a live agent complete enterprise tasks under the same model, sandbox, and grading policy? We present ACES (Agentic Continuous Evaluation of Skills), a repository-native framework for evaluating skills and product capability packages as executable agent artifacts. ACES runs paired live trials with and without a target skill, normalizes trajectories into the Agent Trajectory Interchange Format (ATIF), grades six default runtime metrics, and reports Skill Lift: the target skill's added value for a fixed task, harness, workspace, and scorer. The same protocol supports product-owned task suites that compare baseline, skill, bundle, team-skill, and plugin targets. On 145 real skills from internal enterprise repositories and public catalogs, scan-only gates surface useful authoring issues but measure complementary facets (structural versus LLM-judge Spearman ρ= 0.14). Across 947 scored paired cases from 58 of 64 production skills and four primary harnesses, mean composite Skill Lift is 0.2134 (95\% paired-case CI [0.1967, 0.2301]); mean outcome-only lift, the average of accuracy and goal accuracy, is 0.1799. Composite lift is positive in 72.8\% of paired cases. The largest process-metric gains appear in skill execution, behavior check, and skill efficiency---signals about discovery, routing, workflow following, and tool use that document scans cannot observe. An open-source implementation of the methodology is available in NVIDIA SkillEvaluator.

Advances in large-scale neural recording have made it possible to collect data across many animals and distributed brain regions, raising the question of whether this scale can be exploited to learn general-purpose neural representations transferable across diverse downstream tasks. Yet, progress toward this goal has been limited by fragmented evaluation protocols and a narrow focus on individual task domains. Here, we present BrainWideBench, a benchmark for evaluating across-animal transfer on multi-region neural recordings, built on the International Brain Laboratory Brainwide Map dataset of neural and behavioral recordings spanning 276 brain regions from 139 mice performing a sensory-guided decision-making task. The benchmark is organized around three complementary task suites that evaluate whether learned representations support downstream decoding of behavior, can predict masked or future neural activity, and can recover biologically meaningful anatomical organization. With this benchmark, we systematically evaluate pretraining methods across transfer settings, including finetuning on downstream objectives and zero-shot generalization to unseen animals. Our results confirm pretraining improves performance over matched single-session baselines, but we show current methods exhibit heterogeneity in transfer capabilities: gains depend strongly on the alignment between pretraining objectives and downstream tasks. No single approach performs uniformly well across all three suites, and most methods are designed to only address a subset of them. Together, these findings suggest that learning representations that jointly generalize across behavior, dynamics, and anatomy remains an open challenge. By providing a unified and reproducible evaluation suite, BrainWideBench establishes a framework for measuring progress toward general-purpose models of the mouse brain.

The rapid development of mobile GUI agents has stimulated growing research interest in long-horizon task automation. However, building agents for these tasks faces a critical bottleneck: the reliance on ever-expanding interaction history incurs substantial context overhead. Existing context management and compression techniques often fail to preserve vital semantic information, leading to degraded task performance. We propose AgentProg, a program-guided approach for agent context management that reframes the interaction history as a program with variables and control flow. By organizing information according to the structure of program, this structure provides a principled mechanism to determine which information should be retained and which can be discarded. We further integrate a global belief state mechanism inspired by Belief MDP framework to handle partial observability and adapt to unexpected environmental changes. Experiments on AndroidWorld and our extended long-horizon task suite demonstrate that AgentProg has achieved the state-of-the-art success rates on these benchmarks. More importantly, it maintains robust performance on long-horizon tasks while baseline methods experience catastrophic degradation. Our system is open-sourced at https://github.com/MobileLLM/AgentProg.

Causal Transformer language models suffer from strictly sequential decoding and a quadratic per-step attention cost. While linear-time causal models and discrete diffusion models each address these weaknesses, their integration remains inherently inconsistent: diffusion requires bidirectional attention, while causal models are unidirectional. To unify these architectures, we propose B^3D-RWKV, a diffusion RWKV variant that integrates the model's O(L) inference efficiency with parallel, bidirectional discrete-diffusion through a triplet-block layout method. B^3D-RWKV-7.2B reaches comparable accuracy on an 8-task suite versus existing models while significantly outperforming baselines in decoding throughput with an average of 1.6times speedup.

Modern coding agents expose multiple tool surfaces -- IDE primitives, bash, and Model Context Protocol (MCP) code-execution -- and the field has shipped three contradictory claims about which one matters. We run the missing crossed comparison: an integrity-clean three-arm ablation (baseline / bash_only / code_only) on synthetic computation tasks and SWE-bench Mini modification tasks, holding model, harness, and prompts fixed, with two agents (Claude Code, OpenAI Codex CLI) so the comparison spans both regime and agent-design axes. Across the four resulting (regime, agent) cells, restricting the agent to a single execute_code MCP tool is cheaper than -- or statistically tied with -- its cheapest tool-rich rival in three cells (significantly on Artifact/Claude and SWE-bench/Codex; directionally on Artifact/Codex), with pass rates statistically tied within each cell. The lone exception is SWE-bench/Claude, where code_only is directionally costlier (+14.4%, not significant); a conditional-cost analysis localizes that gap to failure-cost on doomed-run trajectories, not a per-edit tax on successful runs. Two implications: the cheapest tool surface is jointly determined by task regime and agent design rather than by either axis alone, and the headline cost signal lives in cache-adjusted cost -- not pass rate, which is invariant across surfaces at the model sizes we evaluate. The benchmark harness, task suite, and analysis code are available at https://github.com/hyang0129/onlycodes.

We study behavioral alignment and representation dynamics of large language model (LLM) agents in financial decision environments. TradeArena, an auditable trading-agent testbed with risk reports, execution simulation, memory, and replayable trajectories, lets us analyze how rationales, positions, and interventions evolve under market stress. Code and data artifacts are available through the https://github.com/weich97/TradeArena.git{TradeArena repository}. We find pre-failure signatures: planning embeddings drift from normal centroids, fused plan-risk representations separate normal from pre-drawdown states, and local manifolds exhibit effective-rank contraction. Across 80 rolling failure anchors and eight LLM trajectories, this pattern persists across hash, LSA, Transformer, and white-box hidden-state probes. Stress tests with CoT-free target weights, lexical controls, OHLCV noise, and false audits show that rationale-level contraction can vanish without rationales, while intent-space and fused signatures remain informative. Structured risk feedback can act as an external alignment signal without fine-tuning, but not as a universal performance enhancer: true audit feedback improves calibration for some models, returns for others, and exposes cases where placebo or hidden feedback has higher short-horizon return but weaker alignment diagnostics. A 51-stock intraday experiment reveals a correlation blind spot: LLM rationales justify exposure to coupled assets that the risk layer clips. Finally, a financial-audit task suite shifts comparison from ``which model trades best'' to whether models can audit trajectories, respect execution boundaries, reproduce artifacts, and avoid claim overreach. These results support a research claim, not a profitability claim: auditable risk feedback and representation trajectories reveal when LLM financial reasoning is aligning, drifting, or failing.

While spatial foundation models have demonstrated impressive performance on standard datasets, a critical question remains: are they truly all-round players capable of generalizing robustly across diverse downstream tasks, arbitrary viewpoints, shifting scene domains, varying input densities, and specific hardware constraints? Answering this overarching question requires a holistic assessment, yet current models are mainly evaluated on specific domains for which they were specifically designed or trained. Such evaluations are intrinsically limited by narrow paradigm coverage, limited scene domains, and arbitrary frame sampling, making it fundamentally difficult to assess their true generalization capabilities. To address this gap, we present SpatialBench, a cross-paradigm, domain-diverse benchmark for spatial foundation models with deterministic sampling. SpatialBench features unprecedented scale and rigorous deterministic design, comprising 19 datasets and 546 scenes across 5 diverse spatial domains. It comprehensively evaluates 41 models across 6 paradigms on 5 task suites under 4 different input density settings. Our extensive evaluation reveals that current models are not yet all-round players, and uncovers crucial insights for future advancement. Specifically, we demonstrate that full-context attention maximizes accuracy while bounded-memory strategies unlock long-sequence scalability. Moreover, our empirical evaluations in challenging embodied and egocentric tasks demonstrate that strict domain alignment and high data quality are far more critical to performance than simple dataset scaling. Furthermore, to address the largest data gap identified in our analysis, we go beyond evaluation by introducing a large-scale dataset, DA-Next-5M, and a strong baseline model, DA-Next, pushing the boundaries of spatial representation learning.

Indirect prompt injection (IPI) is a major security threat to LLM-powered agents. Thus, a growing body of work have proposed a variety of defensive approaches against IPI. These can be grouped into three broad categories: 1) prompt-based (using prompting as a way to prevent agents from following malicious instructions), 2) detection-based (identifying and filtering malicious instructions), and 3) system-level (using systems insights, such as control and data isolation, for defense). However, commonly used benchmarks for evaluating defense, such as AgentDojo, are inherently static, generating a fixed distribution of IPI attacks. Consequently, static benchmarks do not usefully evaluate defense robustness to adaptive threats. We address this issue by developing AutoDojo, an adaptive extension of AgentDojo that optimizes IPI against a given defense. Using AutoDojo against state-of-the-art IPI defenses across three task suites and five target models, we make two key observations. First, many defenses offer only limited protection: a cheap, black-box adaptive attack using a frontier LLM to iteratively optimize the injection raises attack success rate (ASR) well above the level achieved by static injections against nearly all evaluated defenses. Against a filter that reduces static ASR to 0\%, AutoDojo recovers 28\% overall and 64\% on action-open tasks. Second, for prompt-level and filter-based defenses, ASR is substantially higher on action-open tasks -- where the user's request delegates the action itself to attacker-controlled content -- than on precisely specified tasks. This is a structural limit: on such tasks the injection can pose as ordinary data rather than an explicit instruction, bypassing defenses that rely on detecting instruction-like text. AutoDojo is publicly available at https://github.com/xhOwenMa/AutoDojo.

General-purpose robot policies must work across robots, tasks, and environments, yet the variation that makes physical trials informative also makes them difficult to reproduce. We introduce RoboColosseum, a distributed evaluation framework that combines unified policy serving and standardized protocols across a network of locally operated robots in diverse environments. Our initial study evaluates five vision-language-action (VLA) policies and three frontier AI agents across four embodiments (SO-101, Franka, Bimanual-YAM, and G1 Humanoid). The Open Track compares models without task-specific adaptation in blinded, pairwise trials under matched local conditions. The Fine-tune Track measures how well VLA policies adapt to 20 tasks across all four embodiments when each policy receives the same 100 demonstrations per task. Across 5,699 pairwise Open Track comparisons at 124 sites, frontier agents rival specialized VLA policies, while all systems struggle to complete open-world tasks reliably. After fine-tuning, GR00T N1.7 leads on SO-101 and pi0.5 leads on Franka and Bimanual-YAM. Policy rankings show weak agreement across embodiment-specific task suites (Kendall W = 0.24). Open Track rankings poorly predict relative performance after fine-tuning, with substantial rank reversals on all three embodiments shared between tracks. Measuring partial progress reveals differences that task success alone misses. On the G1 Humanoid, no policy completes any of the five tasks, yet average progress ranges from 13.3% to 23.2% across policies. Open Track rankings stabilize as trials accumulate, and active matchmaking achieves comparable ranking agreement with substantially fewer physical trials than uniform sampling. These results show how distributed physical evaluation can reveal where robot policies generalize, how well they adapt to new tasks, and how far they progress on tasks they fail to complete.

Long-horizon language-model agents are dominated, in lines of code and in operational complexity, not by their underlying model but by the harness that wraps it: context compaction, tool caching, semantic memory, trajectory reuse, speculative tool prediction, and the glue that binds the model to a sandboxed execution environment. We argue that harness design is a first-class machine-learning problem and that automated configuration search dominates manual stacking once the flag space exceeds a handful of bits. We defend this claim in two steps. First, we formalize automated harness optimization as constrained noisy Bayesian optimization over a mixed-variable, cost-heterogeneous configuration space with cold-start-corrected rewards and a posterior chance-constrained safety check, and give a reference solver, HARBOR (Harness Axis-aligned Regularized Bayesian Optimization Routine), built from a block-additive SAAS surrogate, multi-fidelity cost-aware acquisition, and TuRBO trust regions. Second, we instantiate the problem in a flag-gated harness over a production coding agent and report a controlled four-round manual-tuning case study against a fixed task suite and an end-to-end HARBOR run. The formulation itself is task-class agnostic: the configuration space, reward correction, acquisition, and safety check apply to any agent harness with a bounded flag space and a reproducible task suite.

Building agents with large language models (LLMs) for computer control is a burgeoning research area, where the agent receives computer states and performs actions to complete complex tasks. Previous computer agents have demonstrated the benefits of in-context learning (ICL); however, their performance is hindered by several issues. First, the limited context length of LLMs and complex computer states restrict the number of exemplars, as a single webpage can consume the entire context. Second, the exemplars in current methods, such as high-level plans and multi-choice questions, cannot represent complete trajectories, leading to suboptimal performance in long-horizon tasks. Third, existing computer agents rely on task-specific exemplars and overlook the similarity among tasks, resulting in poor generalization to novel tasks. To address these challenges, we introduce Synapse, a computer agent featuring three key components: i) state abstraction, which filters out task-irrelevant information from raw states, allowing more exemplars within the limited context, ii) trajectory-as-exemplar prompting, which prompts the LLM with complete trajectories of the abstracted states and actions to improve multi-step decision-making, and iii) exemplar memory, which stores the embeddings of exemplars and retrieves them via similarity search for generalization to novel tasks. We evaluate Synapse on MiniWoB++, a standard task suite, and Mind2Web, a real-world website benchmark. In MiniWoB++, Synapse achieves a 99.2% average success rate (a 10% relative improvement) across 64 tasks using demonstrations from only 48 tasks. Notably, Synapse is the first ICL method to solve the book-flight task in MiniWoB++. Synapse also exhibits a 56% relative improvement in average step success rate over the previous state-of-the-art prompting scheme in Mind2Web.

Data science agents promise to accelerate discovery and insight-generation by turning data into executable analyses and findings. Yet existing data science benchmarks fall short due to fragmented evaluation interfaces that make cross-benchmark comparison difficult, narrow task coverage and a lack of rigorous data grounding. In particular, we show that a substantial portion of tasks in current benchmarks can be solved without using the actual data. To address these limitations, we introduce DSGym, a standardized framework for evaluating and training data science agents in self-contained execution environments. Unlike static benchmarks, DSGym provides a modular architecture that makes it easy to add tasks, agent scaffolds, and tools, positioning it as a live, extensible testbed. We curate DSGym-Tasks, a holistic task suite that standardizes and refines existing benchmarks via quality and shortcut solvability filtering. We further expand coverage with (1) DSBio: expert-derived bioinformatics tasks grounded in literature and (2) DSPredict: challenging prediction tasks spanning domains such as computer vision, molecular prediction, and single-cell perturbation. Beyond evaluation, DSGym enables agent training via execution-verified data synthesis pipeline. As a case study, we build a 2,000-example training set and trained a 4B model in DSGym that outperforms GPT-4o on standardized analysis benchmarks. Overall, DSGym enables rigorous end-to-end measurement of whether agents can plan, implement, and validate data analyses in realistic scientific context.

Spatio-temporal reasoning is a core capability for Multimodal Large Language Models (MLLMs) operating in the real world. As such, evaluating it precisely has become an essential challenge. However, existing spatio-temporal reasoning benchmark datasets primarily rely on static image sets or passively curated video data, which limits the evaluation of fine-grained reasoning capabilities. In this paper, we introduce VGenST-Bench, a video benchmark that employs generative models to actively synthesize highly controlled and diverse evaluation scenarios. To construct VGenST-Bench, we propose a multi-agent pipeline incorporating a human quality control stage, ensuring the quality of all generated videos and QA pairs. We establish a comprehensive 3x2x2 video taxonomy, encompassing Spatial Scale, Perspective, and Scene Dynamics to span diverse scenarios. Furthermore, we design a hierarchical task suite that decouples low-level visual perception from high-level spatio-temporal reasoning. By shifting the paradigm from passive curation to active synthesis, VGenST-Bench enables fine-grained diagnosis of spatio-temporal understanding in MLLMs.

Lifelong learning offers a promising paradigm of building a generalist agent that learns and adapts over its lifespan. Unlike traditional lifelong learning problems in image and text domains, which primarily involve the transfer of declarative knowledge of entities and concepts, lifelong learning in decision-making (LLDM) also necessitates the transfer of procedural knowledge, such as actions and behaviors. To advance research in LLDM, we introduce LIBERO, a novel benchmark of lifelong learning for robot manipulation. Specifically, LIBERO highlights five key research topics in LLDM: 1) how to efficiently transfer declarative knowledge, procedural knowledge, or the mixture of both; 2) how to design effective policy architectures and 3) effective algorithms for LLDM; 4) the robustness of a lifelong learner with respect to task ordering; and 5) the effect of model pretraining for LLDM. We develop an extendible procedural generation pipeline that can in principle generate infinitely many tasks. For benchmarking purpose, we create four task suites (130 tasks in total) that we use to investigate the above-mentioned research topics. To support sample-efficient learning, we provide high-quality human-teleoperated demonstration data for all tasks. Our extensive experiments present several insightful or even unexpected discoveries: sequential finetuning outperforms existing lifelong learning methods in forward transfer, no single visual encoder architecture excels at all types of knowledge transfer, and naive supervised pretraining can hinder agents' performance in the subsequent LLDM. Check the website at https://libero-project.github.io for the code and the datasets.

AI agents -- systems that combine foundation models with reasoning, planning, memory, and tool use -- are rapidly becoming a practical interface between natural-language intent and real-world computation. This survey synthesizes the emerging landscape of AI agent architectures across: (i) deliberation and reasoning (e.g., chain-of-thought-style decomposition, self-reflection and verification, and constraint-aware decision making), (ii) planning and control (from reactive policies to hierarchical and multi-step planners), and (iii) tool calling and environment interaction (retrieval, code execution, APIs, and multimodal perception). We organize prior work into a unified taxonomy spanning agent components (policy/LLM core, memory, world models, planners, tool routers, and critics), orchestration patterns (single-agent vs.\ multi-agent; centralized vs.\ decentralized coordination), and deployment settings (offline analysis vs.\ online interactive assistance; safety-critical vs.\ open-ended tasks). We discuss key design trade-offs -- latency vs.\ accuracy, autonomy vs.\ controllability, and capability vs.\ reliability -- and highlight how evaluation is complicated by non-determinism, long-horizon credit assignment, tool and environment variability, and hidden costs such as retries and context growth. Finally, we summarize measurement and benchmarking practices (task suites, human preference and utility metrics, success under constraints, robustness and security) and identify open challenges including verification and guardrails for tool actions, scalable memory and context management, interpretability of agent decisions, and reproducible evaluation under realistic workloads.

Developing embodied agents capable of performing complex interactive tasks in real-world scenarios remains a fundamental challenge in embodied AI. Although recent advances in simulation platforms have greatly enhanced task diversity to train embodied Vision Language Models (VLMs), most platforms rely on simplified robot morphologies and bypass the stochastic nature of low-level execution, which limits their transferability to real-world robots. To address these issues, we present a physics-based simulation platform DualTHOR for complex dual-arm humanoid robots, built upon an extended version of AI2-THOR. Our simulator includes real-world robot assets, a task suite for dual-arm collaboration, and inverse kinematics solvers for humanoid robots. We also introduce a contingency mechanism that incorporates potential failures through physics-based low-level execution, bridging the gap to real-world scenarios. Our simulator enables a more comprehensive evaluation of the robustness and generalization of VLMs in household environments. Extensive evaluations reveal that current VLMs struggle with dual-arm coordination and exhibit limited robustness in realistic environments with contingencies, highlighting the importance of using our simulator to develop more capable VLMs for embodied tasks. The code is available at https://github.com/ds199895/DualTHOR.git.

Enterprise Resource Planning (ERP) systems record transactions reliably but still delegate almost all operational decision-making to human specialists, because classical rule-based automation cannot reason about exceptions and monolithic AI assistants degrade when asked to coordinate across functional boundaries. This paper presents Agentic ERP, an expert-system architecture that combines role-aligned large-language-model (LLM) agents with a risk-tiered human-in-the-loop harness and a graph-based orchestrator to execute end-to-end business workflows on a production ERP backend. First, autonomous ERP operation is formulated as a constrained sequential-decision problem over a structured enterprise state, with a decomposition argument linking role-aligned agents to a measurable reduction in per-step tool-selection complexity. Second, a graph-based Planner--Executor--Reflector--Responder orchestration decouples generation from evaluation through externalised grading criteria and sprint contracts, packaging recent harness-engineering principles as inspectable expert-system artefacts. Third, the system is evaluated at three levels: a scenario-based task suite, a comprehensive comparison of six orchestration paradigms on cross-functional crisis tasks, and a 365-day agent-in-the-loop simulation against rule-based RPA and no-intervention baselines. Across these levels the proposed multi-agent method is significantly better than the baseline, and the system sustains a simulated year of operation with zero stockouts while the rule-based baseline accumulates hundreds under the same demand stream. The work shows that role-aligned LLM agents under human oversight can move an ERP system from passively recording transactions to actively executing operational decisions, and it provides a reference architecture and an evaluation protocol for autonomous enterprise resource planning.

Few-shot in-context learning (ICL), the capacity of a model to infer abstract patterns from input-output examples provided in its prompt and apply them to new inputs, has been extensively studied in large language models trained for next-token prediction on human text. Recently, few-shot ICL has been demonstrated in autoregressive genomic models as well. This raises a question: does ICL emerge broadly across domains, and if so, what common structure is shared? To address both, we develop a controlled cross-modality framework that instantiates the same task suite in a variety of modalities to test what we call the Convergent Emergence Hypothesis: the idea that few-shot ICL, when it emerges, shares a common cross-modality difficulty profile - i.e., tasks that benefit from ICL in one modality tend to benefit in others. We show that paired-mapping ICL emerges across six modalities (language, genome, integer sequences, time series, images, and proteins), surpasses controlled baselines, and has correlated per-task effects across five of them. Together, these results provide support for the Convergent Emergence Hypothesis in some modalities, but not all.

Imitation learning with human data has demonstrated remarkable success in teaching robots in a wide range of skills. However, the inherent diversity in human behavior leads to the emergence of multi-modal data distributions, thereby presenting a formidable challenge for existing imitation learning algorithms. Quantifying a model's capacity to capture and replicate this diversity effectively is still an open problem. In this work, we introduce simulation benchmark environments and the corresponding Datasets with Diverse human Demonstrations for Imitation Learning (D3IL), designed explicitly to evaluate a model's ability to learn multi-modal behavior. Our environments are designed to involve multiple sub-tasks that need to be solved, consider manipulation of multiple objects which increases the diversity of the behavior and can only be solved by policies that rely on closed loop sensory feedback. Other available datasets are missing at least one of these challenging properties. To address the challenge of diversity quantification, we introduce tractable metrics that provide valuable insights into a model's ability to acquire and reproduce diverse behaviors. These metrics offer a practical means to assess the robustness and versatility of imitation learning algorithms. Furthermore, we conduct a thorough evaluation of state-of-the-art methods on the proposed task suite. This evaluation serves as a benchmark for assessing their capability to learn diverse behaviors. Our findings shed light on the effectiveness of these methods in tackling the intricate problem of capturing and generalizing multi-modal human behaviors, offering a valuable reference for the design of future imitation learning algorithms.

Financial tasks are pivotal to global economic stability; however, their execution faces challenges including labor intensive processes, low error tolerance, data fragmentation, and tool limitations. Although large language models (LLMs) have succeeded in various natural language processing tasks and have shown potential in automating workflows through reasoning and contextual understanding, current benchmarks for evaluating LLMs in finance lack sufficient domain-specific data, have simplistic task design, and incomplete evaluation frameworks. To address these gaps, this article presents FinMaster, a comprehensive financial benchmark designed to systematically assess the capabilities of LLM in financial literacy, accounting, auditing, and consulting. Specifically, FinMaster comprises three main modules: i) FinSim, which builds simulators that generate synthetic, privacy-compliant financial data for companies to replicate market dynamics; ii) FinSuite, which provides tasks in core financial domains, spanning 183 tasks of various types and difficulty levels; and iii) FinEval, which develops a unified interface for evaluation. Extensive experiments over state-of-the-art LLMs reveal critical capability gaps in financial reasoning, with accuracy dropping from over 90% on basic tasks to merely 40% on complex scenarios requiring multi-step reasoning. This degradation exhibits the propagation of computational errors, where single-metric calculations initially demonstrating 58% accuracy decreased to 37% in multimetric scenarios. To the best of our knowledge, FinMaster is the first benchmark that covers full-pipeline financial workflows with challenging tasks. We hope that FinMaster can bridge the gap between research and industry practitioners, driving the adoption of LLMs in real-world financial practices to enhance efficiency and accuracy.

Vision-Language-Action (VLA) models have achieved high task success rates on robot manipulation task benchmarks. More recently, there has been an emphasis on evaluating the robustness of VLA models to perturbations. However, this robustness is still predominantly measured through Task Success Rate (TSR). In this work, we propose a benchmark-agnostic evaluation framework to measure the behavioural robustness of models by characterising how successful trajectories are executed under perturbation. We implement this methodology by extending the widely-used LIBERO and LIBERO-Plus benchmarks. Across three state-of-the-art VLA models, four LIBERO task suites and seven perturbation conditions, we evaluate changes in both typical successful behaviour and its variability, including metrics of motion smoothness, efficiency and gripper behaviour. We find that perturbations can alter the behaviour of successful trajectories, a phenomenon which cannot necessarily be inferred from TSR alone. Across LIBERO suites, we identify cases where state-of-the-art VLA models achieve comparable TSR under the same perturbation condition, yet behaviour on successful trajectories diverges substantially. Therefore, to have a more robust assessment of task performance, we argue that suitable measures of robustness should capture not only whether a task is completed, but also how the robot behaves while completing it. When evaluating the robustness of VLA models, TSR may be complemented by behavioural evaluation metrics that characterise the nature and variability of successful task execution by robots.

Neural Network movement controllers promise a variety of advantages over conventional control methods however they are not widely adopted due to their inability to produce reliably precise movements. This research explores a bilateral neural network architecture as a control system for motor tasks. We aimed to achieve hemispheric specialisation similar to what is observed in humans across different tasks; the dominant system (usually the right hand, left hemisphere) excels at tasks involving coordination and efficiency of movement, and the non-dominant system performs better at tasks requiring positional stability. Specialisation was achieved by training the hemispheres with different loss functions tailored toward the expected behaviour of the respective hemispheres. We compared bilateral models with and without specialised hemispheres, with and without inter-hemispheric connectivity (representing the biological Corpus Callosum), and unilateral models with and without specialisation. The models were trained and tested on two tasks common in the human motor control literature: the random reach task, suited to the dominant system, a model with better coordination, and the hold position task, suited to the non-dominant system, a model with more stable movement. Each system out-performed the non-favoured system in its preferred task. For both tasks, a bilateral model outperforms the 'non-preferred' hand, and is as good or better than the 'preferred' hand. The Corpus Callosum tends to improve performance, but not always for the specialised models.

AI models capable of comprehending humor hold real-world promise -- for example, enhancing engagement in human-machine interactions. To gauge and diagnose the capacity of multimodal large language models (MLLMs) for humor understanding, we introduce v-HUB, a novel video humor understanding benchmark. v-HUB comprises a curated collection of non-verbal short videos, reflecting real-world scenarios where humor can be appreciated purely through visual cues. We pair each video clip with rich annotations to support a variety of evaluation tasks and analyses, including a novel study of environmental sound that can enhance humor. To broaden its applicability, we construct an open-ended QA task, making v-HUB readily integrable into existing video understanding task suites. We evaluate a diverse set of MLLMs, from specialized Video-LLMs to versatile OmniLLMs that can natively process audio, covering both open-source and proprietary domains. The experimental results expose the difficulties MLLMs face in comprehending humor from visual cues alone. Our findings also demonstrate that incorporating audio helps with video humor understanding, highlighting the promise of integrating richer modalities for complex video understanding tasks.

Multimodal Large Language Models (MLLMs) have shown great potential in revolutionizing Graphical User Interface (GUI) automation. However, existing GUI models mostly rely on learning from nearly error-free offline trajectories, thus lacking reflection and error recovery capabilities. To bridge this gap, we propose GUI-Reflection, a novel framework that explicitly integrates self-reflection and error correction capabilities into end-to-end multimodal GUI models throughout dedicated training stages: GUI-specific pre-training, offline supervised fine-tuning (SFT), and online reflection tuning. GUI-reflection enables self-reflection behavior emergence with fully automated data generation and learning processes without requiring any human annotation. Specifically, 1) we first propose scalable data pipelines to automatically construct reflection and error correction data from existing successful trajectories. While existing GUI models mainly focus on grounding and UI understanding ability, we propose the GUI-Reflection Task Suite to learn and evaluate reflection-oriented abilities explicitly. 2) Furthermore, we built a diverse and efficient environment for online training and data collection of GUI models on mobile devices. 3) We also present an iterative online reflection tuning algorithm leveraging the proposed environment, enabling the model to continuously enhance its reflection and error correction abilities. Our framework equips GUI agents with self-reflection and correction capabilities, paving the way for more robust, adaptable, and intelligent GUI automation, with all data, models, environments, and tools to be released publicly.

We present ReXVQA, the largest and most comprehensive benchmark for visual question answering (VQA) in chest radiology, comprising approximately 696,000 questions paired with 160,000 chest X-rays studies across training, validation, and test sets. Unlike prior efforts that rely heavily on template based queries, ReXVQA introduces a diverse and clinically authentic task suite reflecting five core radiological reasoning skills: presence assessment, location analysis, negation detection, differential diagnosis, and geometric reasoning. We evaluate eight state-of-the-art multimodal large language models, including MedGemma-4B-it, Qwen2.5-VL, Janus-Pro-7B, and Eagle2-9B. The best-performing model (MedGemma) achieves 83.24% overall accuracy. To bridge the gap between AI performance and clinical expertise, we conducted a comprehensive human reader study involving 3 radiology residents on 200 randomly sampled cases. Our evaluation demonstrates that MedGemma achieved superior performance (83.84% accuracy) compared to human readers (best radiology resident: 77.27%), representing a significant milestone where AI performance exceeds expert human evaluation on chest X-ray interpretation. The reader study reveals distinct performance patterns between AI models and human experts, with strong inter-reader agreement among radiologists while showing more variable agreement patterns between human readers and AI models. ReXVQA establishes a new standard for evaluating generalist radiological AI systems, offering public leaderboards, fine-grained evaluation splits, structured explanations, and category-level breakdowns. This benchmark lays the foundation for next-generation AI systems capable of mimicking expert-level clinical reasoning beyond narrow pathology classification. Our dataset will be open-sourced at https://huggingface.co/datasets/rajpurkarlab/ReXVQA

Generative models can now produce photorealistic imagery, yet they still struggle with the long, multi-goal prompts that professional designers issue. To expose this gap and better evaluate models' performance in real-world settings, we introduce Long Goal Bench (LGBench), a 2,000-task suite (1,000 T2I and 1,000 I2I) whose average instruction contains 18 to 22 tightly coupled goals spanning global layout, local object placement, typography, and logo fidelity. We find that even state-of-the-art models satisfy fewer than 72 percent of the goals and routinely miss localized edits, confirming the brittleness of current pipelines. To address this, we present VisionDirector, a training-free vision-language supervisor that (i) extracts structured goals from long instructions, (ii) dynamically decides between one-shot generation and staged edits, (iii) runs micro-grid sampling with semantic verification and rollback after every edit, and (iv) logs goal-level rewards. We further fine-tune the planner with Group Relative Policy Optimization, yielding shorter edit trajectories (3.1 versus 4.2 steps) and stronger alignment. VisionDirector achieves new state of the art on GenEval (plus 7 percent overall) and ImgEdit (plus 0.07 absolute) while producing consistent qualitative improvements on typography, multi-object scenes, and pose editing.

While decoder-only Large Language Models (LLMs) have recently dominated the NLP landscape, encoder-only architectures remain a cost-effective and parameter-efficient standard for discriminative tasks. However, classic encoders like BERT are limited by a short context window, which is insufficient for processing long documents. In this paper, we address this limitation for the Polish by introducing a high-quality Polish model capable of processing sequences of up to 8192 tokens. The model was developed by employing a two-stage training procedure that involves positional embedding adaptation and full parameter continuous pre-training. Furthermore, we propose compressed model variants trained via knowledge distillation. The models were evaluated on 25 tasks, including the KLEJ benchmark, a newly introduced financial task suite (FinBench), and other classification and regression tasks, specifically those requiring long-document understanding. The results demonstrate that our model achieves the best average performance among Polish and multilingual models, significantly outperforming competitive solutions in long-context tasks while maintaining comparable quality on short texts.

Graphical user interface (GUI)-based mobile agents automate digital tasks on mobile devices by interpreting natural-language instructions and interacting with the screen. While recent methods apply reinforcement learning (RL) to train vision-language-model(VLM) agents in interactive environments with a primary focus on performance, generalization remains underexplored due to the lack of standardized benchmarks and open-source RL systems. In this work, we formalize the problem as a Contextual Markov Decision Process (CMDP) and introduce AndroidWorld-Generalization, a benchmark with three increasingly challenging regimes for evaluating zero-shot generalization to unseen task instances, templates, and applications. We further propose an RL training system that integrates Group Relative Policy Optimization (GRPO) with a scalable rollout collection system, consisting of containerized infrastructure and asynchronous execution % , and error recovery to support reliable and efficient training. Experiments on AndroidWorld-Generalization show that RL enables a 7B-parameter VLM agent to surpass supervised fine-tuning baselines, yielding a 26.1\% improvement on unseen instances but only limited gains on unseen templates (15.7\%) and apps (8.3\%), underscoring the challenges of generalization. As a preliminary step, we demonstrate that few-shot adaptation at test-time improves performance on unseen apps, motivating future research in this direction. To support reproducibility and fair comparison, we open-source the full RL training system, including the environment, task suite, models, prompt configurations, and the underlying infrastructure https://github.com/zihuanjiang/AndroidWorld-Generalization.

As agentic systems continue to evolve and are widely deployed in real-world scenarios, there is a growing demand to faithfully evaluate their capabilities. However, current benchmarks are typically built on popular applications with relatively simple tasks and focus on a narrow set of capabilities while overlooking broader dimensions, resulting in saturated performance on modern agents and failing to probe their limitations. To this end, we introduce GauntletBench, a web-based benchmark for evaluating agent generalisation in challenging scenarios, focusing on three underexplored capabilities (temporal perception, graphical understanding, and 3D reasoning), across five less-covered professional applications (Video Editor, Workflow Builder, 3D Modeller, Flight Analyser, and Circuit Designer), each with 20 vision-intensive tasks (100 in total). Our benchmark provides a modular pipeline that comprises an environment compatible with both open- and closed-source agent frameworks, a controlled web-based application, a well-structured task suite, and an automated evaluation engine with diverse metrics. Contrary to widespread expectations, our empirical results reveal that frontier agentic systems remain far from achieving human-level performance. Even the state-of-the-art agent achieves only a 19.1% success rate on our GauntletBench, highlighting the limitations in these overlooked capabilities and generalisation. By comparison, non-expert human annotators achieve over 80% success on our challenging yet feasible tasks, revealing the substantial gap between current agent capabilities and those required for complex real-world scenarios.

When a benchmark's accuracy saturates, it is often retired and replaced with a more challenging version. We show that this approach privileges accuracy and misses the opportunity to study six other key dimensions of agent performance: construct validity issues such as shortcuts, out-of-distribution generalizability, efficiency, reliability, the relative importance of the model versus the scaffold, and uplift from human-agent collaboration. We use CORE-Bench Hard, a benchmark for computational reproducibility of scientific code, as a case study to demonstrate that measuring agents along these dimensions yields meaningful insights into agent performance even after accuracy saturates. First, we surface threats to construct validity in CORE-Bench Hard that are difficult to anticipate with less capable agents. We introduce an improved benchmark, CORE-Bench v1.1, and an out-of-distribution task suite, CORE-Bench OOD. Second, we find that despite accuracy saturation, CORE-Bench v1.1 remains useful for measuring efficiency, reliability, model performance, and scaffold performance. Finally, we conduct a small-scale randomized experiment to measure uplift from human-agent collaboration on real-world computational reproducibility tasks. We find a statistically significant speedup by about a factor of two -- likely underestimated due to one-fifth of human-only reproductions reaching the time limit before completing -- and describe various other findings. Together, our contributions present a more rigorous alternative to the dominant accuracy-centric evaluation paradigm.

AI-assisted research compresses ideation, implementation, evaluation, and manuscript writing into a single interactive loop. This compression is useful, but it also creates a publication risk: paper claims can become easier to state than to audit. We present ResearchLoop, an evidence-gated control plane for AI-assisted computational research. ResearchLoop treats research questions, task contracts, evidence objects, claim ledgers, closeouts, and paper bindings as durable project state, realized here as a repository-backed runtime. This technical report provides the complete protocol specification, state model, transition rules, claim-admission algorithm, and insight-compounding mechanism. It also reports the full experimental record spanning nine versions (V0--V9), including a self-hosting case study, a controlled task-suite study with component ablations, a mathematical olympiad evaluation, and a supplementary SciCode boundary experiment evaluated with the official generated-code harness. All artifacts, manifests, and verification reports are preserved in the project repository.

Large multi-modal language models are increasingly deployed in high-stakes domains, making well-calibrated uncertainty essential. Traditional Bayesian methods approximate posteriors over all model weights, which becomes intractable for modern large models. For this reason, recent work instead considers Bayesian low-rank adaptation to enable tractable posterior approximation. Due to a lack of a standardized benchmark to evaluate these approaches, it remains unclear where these methods provide meaningful benefits. To fill this gap, we introduce Bayesian Adaptation Gym (BAG), a benchmark for the Bayesian adaptation of multi-modal language models. BAG provides reference implementations of classic Bayesian baselines and state-of-the-art adaptation methods, along with a multi-modal dataset and task suite designed to probe calibration, robustness under distribution shift, and decision-making under uncertainty via active learning. Using BAG, we conduct and report extensive experiments across model sizes, datasets, and tasks to highlight the successes and failures of current Bayesian adaptation approaches. To enable further research, BAG is fully open source: https://github.com/SRI-CSL/BayesAdapt.

Large language models (LLMs) are advancing rapidly in medical NLP, yet Traditional Chinese Medicine (TCM) with its distinctive ontology, terminology, and reasoning patterns requires domain-faithful evaluation. Existing TCM benchmarks are fragmented in coverage and scale and rely on non-unified or generation-heavy scoring that hinders fair comparison. We present the LingLanMiDian (LingLan) benchmark, a large-scale, expert-curated, multi-task suite that unifies evaluation across knowledge recall, multi-hop reasoning, information extraction, and real-world clinical decision-making. LingLan introduces a consistent metric design, a synonym-tolerant protocol for clinical labels, a per-dataset 400-item Hard subset, and a reframing of diagnosis and treatment recommendation into single-choice decision recognition. We conduct comprehensive, zero-shot evaluations on 14 leading open-source and proprietary LLMs, providing a unified perspective on their strengths and limitations in TCM commonsense knowledge understanding, reasoning, and clinical decision support; critically, the evaluation on Hard subset reveals a substantial gap between current models and human experts in TCM-specialized reasoning. By bridging fundamental knowledge and applied reasoning through standardized evaluation, LingLan establishes a unified, quantitative, and extensible foundation for advancing TCM LLMs and domain-specific medical AI research. All evaluation data and code are available at https://github.com/TCMAI-BJTU/LingLan and http://tcmnlp.com.

As embodied agents operate in increasingly complex environments, the ability to perceive, track, and reason about individual object instances over time becomes essential, especially in tasks requiring sequenced interactions with visually similar objects. In these non-Markovian settings, key decision cues are often hidden in object-specific histories rather than the current scene. Without persistent memory of prior interactions (what has been interacted with, where it has been, or how it has changed) visuomotor policies may fail, repeat past actions, or overlook completed ones. To surface this challenge, we introduce LIBERO-Mem, a non-Markovian task suite for stress-testing robotic manipulation under object-level partial observability. It combines short- and long-horizon object tracking with temporally sequenced subgoals, requiring reasoning beyond the current frame. However, vision-language-action (VLA) models often struggle in such settings, with token scaling quickly becoming intractable even for tasks spanning just a few hundred frames. We propose Embodied-SlotSSM, a slot-centric VLA framework built for temporal scalability. It maintains spatio-temporally consistent slot identities and leverages them through two mechanisms: (1) slot-state-space modeling for reconstructing short-term history, and (2) a relational encoder to align the input tokens with action decoding. Together, these components enable temporally grounded, context-aware action prediction. Experiments show Embodied-SlotSSM's baseline performance on LIBERO-Mem and general tasks, offering a scalable solution for non-Markovian reasoning in object-centric robotic policies.

In this work, we present a scalable reinforcement learning method for training multi-task policies from large offline datasets that can leverage both human demonstrations and autonomously collected data. Our method uses a Transformer to provide a scalable representation for Q-functions trained via offline temporal difference backups. We therefore refer to the method as Q-Transformer. By discretizing each action dimension and representing the Q-value of each action dimension as separate tokens, we can apply effective high-capacity sequence modeling techniques for Q-learning. We present several design decisions that enable good performance with offline RL training, and show that Q-Transformer outperforms prior offline RL algorithms and imitation learning techniques on a large diverse real-world robotic manipulation task suite. The project's website and videos can be found at https://q-transformer.github.io

Recent vision-language-action models (VLAs) build upon pretrained vision-language models and leverage diverse robot datasets to demonstrate strong task execution, language following ability, and semantic generalization. Despite these successes, VLAs struggle with novel robot setups and require fine-tuning to achieve good performance, yet how to most effectively fine-tune them is unclear given many possible strategies. In this work, we study key VLA adaptation design choices such as different action decoding schemes, action representations, and learning objectives for fine-tuning, using OpenVLA as our representative base model. Our empirical analysis informs an Optimized Fine-Tuning (OFT) recipe that integrates parallel decoding, action chunking, a continuous action representation, and a simple L1 regression-based learning objective to altogether improve inference efficiency, policy performance, and flexibility in the model's input-output specifications. We propose OpenVLA-OFT, an instantiation of this recipe, which sets a new state of the art on the LIBERO simulation benchmark, significantly boosting OpenVLA's average success rate across four task suites from 76.5% to 97.1% while increasing action generation throughput by 26times. In real-world evaluations, our fine-tuning recipe enables OpenVLA to successfully execute dexterous, high-frequency control tasks on a bimanual ALOHA robot and outperform other VLAs (pi_0 and RDT-1B) fine-tuned using their default recipes, as well as strong imitation learning policies trained from scratch (Diffusion Policy and ACT) by up to 15% (absolute) in average success rate. We release code for OFT and pretrained model checkpoints at https://openvla-oft.github.io/.

Automating the development of machine learning algorithms has the potential to unlock new breakthroughs. However, our ability to improve and evaluate algorithm discovery systems has thus far been limited by existing task suites. They suffer from many issues, such as: poor evaluation methodologies; data contamination; and containing saturated or very similar problems. Here, we introduce DiscoGen, a procedural generator of algorithm discovery tasks for machine learning, such as developing optimisers for reinforcement learning or loss functions for image classification. Motivated by the success of procedural generation in reinforcement learning, DiscoGen spans millions of tasks of varying difficulty and complexity from a range of machine learning fields. These tasks are specified by a small number of configuration parameters and can be used to optimise algorithm discovery agents (ADAs). We present DiscoBench, a benchmark consisting of a fixed, small subset of DiscoGen tasks for principled evaluation of ADAs. Finally, we propose a number of ambitious, impactful research directions enabled by DiscoGen, in addition to experiments demonstrating its use for prompt optimisation of an ADA. DiscoGen is released open-source at https://github.com/AlexGoldie/discogen.

Although robot-to-robot (R2R) communication improves indoor scene understanding beyond what a single robot can achieve, R2R alone cannot overcome partial observability without substantial exploration overhead or scaling team size. In contrast, many indoor environments already include low-cost Internet of Things (IoT) sensors (e.g., cameras) that provide persistent, building-wide context beyond onboard perception. We therefore introduce IndoorR2X, the first benchmark and simulation framework for Large Language Model (LLM)-driven multi-robot task planning with Robot-to-Everything (R2X) perception and communication in indoor environments. IndoorR2X integrates observations from mobile robots and static IoT devices to construct a global semantic state that supports scalable scene understanding, reduces redundant exploration, and enables high-level coordination through LLM-based planning. IndoorR2X provides configurable simulation environments, sensor layouts, robot teams, and task suites to systematically evaluate high-level semantic coordination strategies. Extensive experiments across diverse settings demonstrate that IoT-augmented world modeling improves multi-robot efficiency and reliability, and we highlight key insights and failure modes for advancing LLM-based collaboration between robot teams and indoor IoT sensors. See our project website: https://fandulu.github.io/IndoorR2X_project_page/.

Active vision -- where a policy controls its own gaze during manipulation -- has emerged as a key capability for imitation learning, with multiple independent systems demonstrating its benefits in the past year. Yet there is no shared benchmark to compare approaches or quantify what active vision contributes, on which task types, and under what conditions. We introduce TAVIS, evaluation infrastructure for active-vision imitation learning, with two complementary task suites -- TAVIS-Head (5 tasks, global search via pan/tilt necks) and TAVIS-Hands (3 tasks, local occlusion via wrist cameras) -- on two humanoid torso embodiments (GR1T2, Reachy2), built on IsaacLab. TAVIS provides three evaluation primitives: a paired headcam-vs-fixedcam protocol on identical demonstrations; GALT (Gaze-Action Lead Time), a novel metric grounded in cognitive science and HRI that quantifies anticipatory gaze in learned policies; and procedural ID/OOD splits. Baseline experiments with Diffusion Policy and π_0 reveal that (i) active-vision generally helps, but benefits are task-conditional rather than uniform; (ii) multi-task policies degrade sharply under controlled distribution shifts on both suites; and (iii) imitation alone yields anticipatory gaze, with median lead times comparable to the human teleoperator reference. Code, evaluation scripts, demonstrations (LeRobot v3.0; ~2200 episodes) and trained baselines are released at https://github.com/spiglerg/tavis and https://huggingface.co/tavis-benchmark.

Learning effective robot manipulation policies requires diverse, high-quality demonstrations, yet existing data pipelines are often difficult to scale because they rely on specialized hardware, centralized operators, or fixed task suites. We present AXIS, a growable community-driven data engine and benchmark for scalable robot learning, which enables browser-based teleoperation for large-scale demonstration collection, automatically generates and validates new manipulation tasks, and transforms community-collected demonstrations into training-ready data through automated success checking, quality filtering, trajectory smoothing, and visual and physics-based augmentation. The AXIS dataset currently contains 207 diverse tasks and 50K+ trajectories. Meanwhile, AXIS organizes data into task snapshots and evaluates policies with a systematic held-out protocol. We compare vision-language-action (VLA) policies under a unified AXIS evaluation suite and analyze scaling behavior across different data volumes. Continual pretraining on AXIS substantially improves the overall success rate of π_{0.5} by 5.8%, outperforms the model pretrained on RoboCasa365 by 37.3%, and exhibits consistent scaling with increasing data volume, with the largest gains observed under layout, sensor-noise, and camera perturbations.

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.

We introduce OpenMobile-2, a near-frontier mobile agent with fully open training environments and recipes. We make three key advances: (1) Diverse environments with simulated commercial apps: We build MobileGym++, featuring 35 realistic, functionally rich commercial-style apps with cross-app workflows, while preserving full controllability for reset and verification. Together with newly configured apps in Android emulators, this yields a diverse playground spanning over 110 apps. (2) Open training data at scale: Building on this foundation, we curate nearly 12K mobile interaction trajectories for supervised fine-tuning and the largest open collection of verifiable RL training data for mobile agents, comprising over 2K executable tasks with automatic rewards. (3) Hybrid GUI and app-native tool use: We explore an experimental mobile-use setting where apps expose selected functionalities as app-native tools alongside their GUIs, allowing agents to interleave GUI actions and tool calls within a task. We implement this setting in MobileGym++ with over 300 carefully scoped tools across 50+ apps. Additionally, we introduce MobileGym++ Bench for complex, long-horizon commercial mobile scenarios, supporting both GUI-only and hybrid GUI-tool evaluation on a shared task suite. OpenMobile-2 performs competitively across established benchmarks, including AndroidWorld (79.9) and MobileWorld (50.4), while showing promising transfer to real-device mobile use, nearly doubling SPA-Bench performance from 31.9 to 59.6. All resources are available at https://os-copilot.github.io/OpenMobile2-Home/.

Large language models (LLMs) have demonstrated remarkable performance on single-turn text-to-SQL tasks, but real-world database applications predominantly require multi-turn interactions to handle ambiguous queries, execution errors, and evolving user requirements. Existing multi-turn benchmarks fall short by treating conversation histories as static context or limiting evaluation to read-only operations, failing to reflect production-grade database assistant challenges. We introduce BIRD-INTERACT, a benchmark that restores this realism through: (1) a comprehensive interaction environment coupling each database with a hierarchical knowledge base, metadata files, and a function-driven user simulator, enabling models to solicit clarifications, retrieve knowledge, and recover from errors without human supervision; (2) two evaluation settings consisting of a pre-defined conversational protocol (c-Interact) and an open-ended agentic setting (a-Interact) where models autonomously decide when to query the user simulator or explore the environment; (3) a challenging task suite covering the full CRUD spectrum for business-intelligence and operational use cases, guarded by executable test cases. Each task features ambiguous and follow-up sub-tasks requiring dynamic interaction. The suite comprises BIRD-INTERACT-FULL (600 tasks, up to 11,796 interactions) for comprehensive performance assessment, and BIRD-INTERACT-LITE (300 tasks with simplified databases) for detailed behavioral analysis and rapid method development. Our empirical results highlight BIRD-INTERACT's difficulty: GPT-5 completes only 8.67% of tasks in c-Interact and 17.00% in a-Interact. Analysis via memory grafting and Interaction Test-time Scaling validates the importance of effective interaction for complex, dynamic text-to-SQL tasks.

Vision-Language-Action (VLA) models are emerging as a next-generation paradigm for robotics. We introduce dVLA, a diffusion-based VLA that leverages a multimodal chain-of-thought to unify visual perception, language reasoning, and robotic control in a single system. dVLA jointly optimizes perception, language understanding, and action under a single diffusion objective, enabling stronger cross-modal reasoning and better generalization to novel instructions and objects. For practical deployment, we mitigate inference latency by incorporating two acceleration strategies, a prefix attention mask and KV caching, yielding up to around times speedup at test-time inference. We evaluate dVLA in both simulation and the real world: on the LIBERO benchmark, it achieves state-of-the-art performance with a 96.4% average success rate, consistently surpassing both discrete and continuous action policies; on a real Franka robot, it succeeds across a diverse task suite, including a challenging bin-picking task that requires multi-step planning, demonstrating robust real-world performance. Together, these results underscore the promise of unified diffusion frameworks for practical, high-performance VLA robotics.

Deformable object manipulation poses challenges beyond task completion: successful execution must also maintain safe physical interaction, holding the object stably without slip or drop while avoiding excessive deformation. However, existing manipulation benchmarks are predominantly success-oriented and rarely evaluate whether a policy remains physically safe throughout execution. We present SoftVTBench, a safety-aware visuo-tactile benchmark for physically constrained deformable object manipulation. Built in Isaac Sim with finite-element-simulated deformable objects, SoftVTBench provides multi-view RGB observations, RGB tactile sensing with marker motion, proprioception, and language instructions, and defines four matched task suites over object type (deformable vs. rigid) and variation axis (object vs. spatial). It separately reports Goal Success and Safety Success; the latter additionally requires no drop and peak deformation below a calibrated object-specific threshold, measured from policy-hidden privileged Finite Element Method (FEM) states. We implement pi0.5-based baselines under this protocol. Experiments show that success-only evaluation substantially overstates policy performance, as a large fraction of goal-completing rollouts still violate physical safety. Furthermore, incorporating tactile sensing improves Safety Success (e.g., from 21.4% to 35.6% on object-centric deformable tasks) and reduces object deformation during execution, while maintaining comparable Goal Success. SoftVTBench provides a reproducible benchmark for studying visuo-tactile deformable manipulation under physical interaction constraints.

To safely navigate intricate real-world scenarios, autonomous vehicles must be able to adapt to diverse road conditions and anticipate future events. World model (WM) based reinforcement learning (RL) has emerged as a promising approach by learning and predicting the complex dynamics of various environments. Nevertheless, to the best of our knowledge, there does not exist an accessible platform for training and testing such algorithms in sophisticated driving environments. To fill this void, we introduce CarDreamer, the first open-source learning platform designed specifically for developing WM based autonomous driving algorithms. It comprises three key components: 1) World model backbone: CarDreamer has integrated some state-of-the-art WMs, which simplifies the reproduction of RL algorithms. The backbone is decoupled from the rest and communicates using the standard Gym interface, so that users can easily integrate and test their own algorithms. 2) Built-in tasks: CarDreamer offers a comprehensive set of highly configurable driving tasks which are compatible with Gym interfaces and are equipped with empirically optimized reward functions. 3) Task development suite: This suite streamlines the creation of driving tasks, enabling easy definition of traffic flows and vehicle routes, along with automatic collection of multi-modal observation data. A visualization server allows users to trace real-time agent driving videos and performance metrics through a browser. Furthermore, we conduct extensive experiments using built-in tasks to evaluate the performance and potential of WMs in autonomous driving. Thanks to the richness and flexibility of CarDreamer, we also systematically study the impact of observation modality, observability, and sharing of vehicle intentions on AV safety and efficiency. All code and documents are accessible on https://github.com/ucd-dare/CarDreamer.

Protein language models are trained primarily with masked language modeling (MLM), which predicts amino-acid identities at masked positions. We ask whether latent-space prediction can complement these token-level objectives under matched wall-clock budget. Across pretrained and random-init protein sequence encoders at 35--150M parameters, we find that the best protein-JEPA design is not all-position latent prediction but a variant: predicting latent targets only at masked positions, and retaining the MLM cross-entropy. We call this recipe masked-position MLM+JEPA. On a 16-task downstream suite (15 frozen linear probes plus SCOPe-40 zero-shot fold retrieval), under matched wall-clock budgets, this recipe wins more tasks than it loses against MLM-only continuation: 10 wins / 3 losses / 3 ties (hereafter W/L/T) on pretrained ESM2-35M, 11/2/3 on ESM2-150M while results in pretraining from scratch are mixed (6/8/2). Gains are seen for multiple models on 11 of 16 tasks, including stability, etaβeta-lactamase fitness, variant effect, intrinsic disorder, remote homology, enzyme classification, and SCOPe-40 fold retrieval. Tasks with more losses than wins are Fluorescence (TAPE) and Peptide-HLA Binding. All-position MLM+JEPA matches MLM-only overall but does not reproduce the masked-position gains. JEPA-only (no MLM) collapses in nearly every experiment. We conclude that JEPA, when combined with MLM, is competitive and can outperform pure MLM in pretraining and continued training, even under matched wall-clock budgets.

We present Qualixar OS, the first application-layer operating system for universal AI agent orchestration. Unlike kernel-level approaches (AIOS) or single-framework tools (AutoGen, CrewAI), Qualixar OS provides a complete runtime for heterogeneous multi-agent systems spanning 10 LLM providers, 8+ agent frameworks, and 7 transports. We contribute: (1) execution semantics for 12 multi-agent topologies including grid, forest, mesh, and maker patterns; (2) Forge, an LLM-driven team design engine with historical strategy memory; (3) three-layer model routing combining Q-learning, five strategies, and Bayesian POMDP with dynamic multi-provider discovery; (4) a consensus-based judge pipeline with Goodhart detection, JSD drift monitoring, and alignment trilemma navigation; (5) four-layer content attribution with HMAC signing and steganographic watermarks; (6) universal compatibility via the Claw Bridge supporting MCP and A2A protocols with a 25-command Universal Command Protocol; (7) a 24-tab production dashboard with visual workflow builder and skill marketplace. Qualixar OS is validated by 2,821 test cases across 217 event types and 8 quality modules. On a custom 20-task evaluation suite, the system achieves 100% accuracy at a mean cost of $0.000039 per task. Source-available under the Elastic License 2.0.

What makes a video task uniquely suited for videos, beyond what can be understood from a single image? Building on recent progress in self-supervised image-language models, we revisit this question in the context of video and language tasks. We propose the atemporal probe (ATP), a new model for video-language analysis which provides a stronger bound on the baseline accuracy of multimodal models constrained by image-level understanding. By applying this model to standard discriminative video and language tasks, such as video question answering and text-to-video retrieval, we characterize the limitations and potential of current video-language benchmarks. We find that understanding of event temporality is often not necessary to achieve strong or state-of-the-art performance, even compared with recent large-scale video-language models and in contexts intended to benchmark deeper video-level understanding. We also demonstrate how ATP can improve both video-language dataset and model design. We describe a technique for leveraging ATP to better disentangle dataset subsets with a higher concentration of temporally challenging data, improving benchmarking efficacy for causal and temporal understanding. Further, we show that effectively integrating ATP into full video-level temporal models can improve efficiency and state-of-the-art accuracy.

Computer-use agents (CUAs) are rapidly improving, yet synthesizing diverse and reliable software environments at scale remains challenging. Native application environments require substantial construction and integration effort, which limits application coverage and increases the risk of application-specific overfitting. Browser-hosted environments provide limited interaction depth and workflow complexity. These limitations motivate SoftwareGym, an automated pipeline for synthesizing GUI software environments with diverse interfaces, executable functionality, and runtime-verified behavior. It generates structurally diverse designs, realizes them as runnable applications, and uses GUI evaluation with iterative repair to ensure functional reliability. Using SoftwareGym, we synthesize 182 applications across twelve categories and achieve a mean GUI task pass rate of 91.5%. We construct SoftwareGym-Bench, a 120-task evaluation suite spanning the same twelve categories, and use trajectories collected from these environments for CUA training. The resulting model improves Pass@3 from 41.39% to 56.97% on the strictly out-of-domain OSWorld benchmark and achieves 40.0% on SoftwareGym-Bench. These results demonstrate the potential of synthesized software environments as scalable and transferable supervision for CUA training.

General-purpose agents can plan, use tools, and revise their behavior from feedback, but it remains unclear whether these capabilities transfer from digital environments to embodied manipulation. To investigate this question, we introduce LIBERO-Agent, an agent-native benchmark for evaluating these agents in robot manipulation tasks. Rather than asking agents to submit task-level Python control programs or operate through high-level robot skills, LIBERO-Agent provides an interactive robotic environment where agents can select which observations to inspect, process them with their own tools, and issue native action commands. LIBERO-Agent integrates 200 tasks into a common interaction framework and provides a 30-task primary suite that separates perception, short-horizon execution, and long-horizon composition. Results reveal a pronounced reliability gap: while agents perform well on perception and easy short-horizon tasks, their performance degrades substantially on hard short-horizon and long-horizon tasks. Richer observations improve short-horizon manipulation, while demonstration benefits depend on the agent and format. Among these agents, GPT-6 Astra achieves the strongest overall performance. Further analysis shows its major advantage lies in mechanism interaction, especially when sustained physical contact is needed, while its remaining failures stem from cross-stage interference and geometric errors.

We present QIMMA, a quality-assured Arabic LLM leaderboard that places systematic benchmark validation at its core. Rather than aggregating existing resources as-is, QIMMA applies a multi-model assessment pipeline combining automated LLM judgment with human review to surface and resolve systematic quality issues in well-established Arabic benchmarks before evaluation. The result is a curated, multi-domain, multi-task evaluation suite of over 52k samples, grounded predominantly in native Arabic content; code evaluation tasks are the sole exception, as they are inherently language-agnostic. Transparent implementation via LightEval, EvalPlus and public release of per-sample inference outputs make QIMMA a reproducible and community-extensible foundation for Arabic NLP evaluation.

Reinforcement learning (RL) has emerged as a promising approach to improve large language model (LLM) reasoning, yet most open efforts focus narrowly on math and code, limiting our understanding of its broader applicability to general reasoning. A key challenge lies in the lack of reliable, scalable RL reward signals across diverse reasoning domains. We introduce Guru, a curated RL reasoning corpus of 92K verifiable examples spanning six reasoning domains--Math, Code, Science, Logic, Simulation, and Tabular--each built through domain-specific reward design, deduplication, and filtering to ensure reliability and effectiveness for RL training. Based on Guru, we systematically revisit established findings in RL for LLM reasoning and observe significant variation across domains. For example, while prior work suggests that RL primarily elicits existing knowledge from pretrained models, our results reveal a more nuanced pattern: domains frequently seen during pretraining (Math, Code, Science) easily benefit from cross-domain RL training, while domains with limited pretraining exposure (Logic, Simulation, and Tabular) require in-domain training to achieve meaningful performance gains, suggesting that RL is likely to facilitate genuine skill acquisition. Finally, we present Guru-7B and Guru-32B, two models that achieve state-of-the-art performance among open models RL-trained with publicly available data, outperforming best baselines by 7.9% and 6.7% on our 17-task evaluation suite across six reasoning domains. We also show that our models effectively improve the Pass@k performance of their base models, particularly on complex tasks less likely to appear in pretraining data. We release data, models, training and evaluation code to facilitate general-purpose reasoning at: https://github.com/LLM360/Reasoning360

Reinforcement learning (RL) with dense rewards and imitation learning (IL) with human-generated trajectories are the most widely used approaches for training modern embodied agents. RL requires extensive reward shaping and auxiliary losses and is often too slow and ineffective for long-horizon tasks. While IL with human supervision is effective, collecting human trajectories at scale is extremely expensive. In this work, we show that imitating shortest-path planners in simulation produces agents that, given a language instruction, can proficiently navigate, explore, and manipulate objects in both simulation and in the real world using only RGB sensors (no depth map or GPS coordinates). This surprising result is enabled by our end-to-end, transformer-based, SPOC architecture, powerful visual encoders paired with extensive image augmentation, and the dramatic scale and diversity of our training data: millions of frames of shortest-path-expert trajectories collected inside approximately 200,000 procedurally generated houses containing 40,000 unique 3D assets. Our models, data, training code, and newly proposed 10-task benchmarking suite CHORES are available in https://spoc-robot.github.io.

Transformers and deep state space models (SSMs) sit at opposite ends of a basic design choice: attention routes each query through a growing key-value (KV) cache by content-based matching at quadratic cost, while deep SSMs compress context into a fixed-size recurrent state that is not directly addressed by query-key matching. We propose Interdomain Attention, which integrates an SSM into an attention module through kernel methods: an attention kernel is approximated by a finite feature map, the resulting key features and values are projected onto a shared set of basis functions maintained by a single SSM recurrence, and each query attends to the compressed coefficients through its own feature map, recovering query-conditioned attention over a fixed-size state. The scalable layer is a learned relaxation of this derivation, and we validate its components through ablations. In a 125M to 1.3B autoregressive language-modeling study on FineWeb-Edu at matched recurrent-state budget, Interdomain Attention improves on an SSM token mixer at every scale, surpasses a same-recipe softmax baseline at 1.3B on validation perplexity and on the eight-task commonsense suite, and inherits the length-flat behavior of its fixed-state core out to 3.5x the training context. Ablations indicate that the query-conditioned projection is the main source of the gain.

Building performant Vision-Language Models (VLMs) requires carefully curating large-scale training datasets, yet the community lacks systematic benchmarks for evaluating such curation strategies. We introduce DataComp for VLMs (DCVLM), a benchmark for controlled data-centric experiments to improve VLM training. As part of DCVLM, we collect 160 datasets spanning four data types -- image-caption pairs, multimodal interleaved documents, text-only, and instruction-tuning data -- into a corpus of 6T multimodal tokens. DCVLM allows participants to test curation strategies (filtering, mixing, formatting, sampling) across 1B-8B models and 6.25B-200B token budgets. Models are then evaluated on a carefully selected suite of up to 52 downstream benchmarks across 9 domains. We conduct extensive experiments on DCVLM and find that data mixing, not filtering, is key to a high-quality training dataset: instruction-heavy mixtures scale better than caption-heavy ones, with gains widening at larger scales. The resulting dataset, DCVLM-Baseline, enables training an 8B VLM to 63.6% accuracy on our 33-task core suite with 200B training tokens. Compared to FineVision, the state-of-the-art open VLM training dataset, this represents an improvement of +5.4pp. DCVLM and all accompanying artifacts will be made publicly available at https://www.datacomp.ai/dcvlm/.

We present N_0-VTLA, a vision-tactile-language-action (VTLA) foundation model capable of (1) fine-grained contact-rich manipulation with tactile perception and tactile-feedback control, and (2) offline policy improvement from stored deployment data. Building on current vision-based backbones, we propose a training recipe for tactile integration consisting of visuo-tactile pre-training, staged tactile-pathway integration, and advantage-conditioned offline policy improvement. During pre-training, the policy learns broad contact priors from NeoData, our large-scale visuo-tactile robot dataset; to our knowledge, N_0-VTLA is the first VTLA model pretrained on tactile data at scale. During post-training, we augment the policy with a predictive tactile pathway that distills the contact patterns learned at scale into the fine motion adjustments required by downstream tactile-centric manipulation. For offline policy improvement, we introduce ALTER, an advantage-conditioned offline reinforcement learning method that converts relative progress and trajectory-event comparisons into binary advantage labels for policy training on a fixed deployment corpus, further improving task-specific learning on contact-rich skills such as deformable object manipulation. Across contact-rich benchmarks, N_0-VTLA outperforms strong baselines by wide margins: it wins all nine real-robot NeoReal tasks and reaches 63.8% mean success on a twenty-task simulation suite, against 44.0% for the strongest baseline. N_0-VTLA policies trained with ALTER reach 75-95% success on three long-horizon real-robot tasks. These results lay a foundation for versatile tactile-driven manipulation policies.

Recursive self-improvement, in which AI systems build stronger AI systems, is one promising route to superintelligence. We study a weaker form that can be measured today: frontier agents post-training a small raw base model with public data, as a test of their research potential. We build Athanor, an open agent-native system for autonomous post-training research (AutoPostTrain) on real GPUs. We give the agents full research autonomy, low-friction engineering tools, and efficient starter code. The agents carry out the research themselves: they read datasets and evaluation results, build their training data, write and modify training algorithms, tune hyperparameters, and analyze the results to decide what to try next. They work in a sandbox and cannot use external teacher models. Athanor's services supply frozen evaluation, managed jobs, decontaminated data, a persistent research record, and a held-out suite the agents are unaware of. In a 104-hour campaign, a team of four Claude Opus 5.5 agents, each on an eight-GPU node and given no research advice, improves Qwen3-4B-Base from 15.9 to 67.3 on a six-task development suite and from 9.2 to 48.2 on a held-out suite, against 66.5 and 47.9 for the official Qwen3-4B-Instruct-2507. To our knowledge, this is the first demonstration that a team of frontier agents can autonomously post-train a raw base model into a cross-domain generalist model that matches or surpasses the official instruct version on average. The campaign's full record shows that the agents reach this level by recombining known methods and each other's models: SFT on public reasoning data, RL with a rollout change one agent introduced, weight merges across agents, and rounds of self-distillation. With execution handled, today's agents already have the engineering and research ability to conquer this task. We hope Athanor provides a foundation for scaling autonomous AI research.

Developing large language models is expensive and involves making decisions with small experiments, typically by evaluating on large, multi-task evaluation suites. In this work, we analyze specific properties which make a benchmark more reliable for such decisions, and interventions to design higher-quality evaluation benchmarks. We introduce two key metrics that show differences in current benchmarks: signal, a benchmark's ability to separate better models from worse models, and noise, a benchmark's sensitivity to random variability between training steps. We demonstrate that benchmarks with a better signal-to-noise ratio are more reliable when making decisions at small scale, and those with less noise have lower scaling law prediction error. These results suggest that improving signal or noise will lead to more useful benchmarks, so we introduce three interventions designed to directly affect signal or noise. For example, we propose that switching to a metric that has better signal and noise (e.g., perplexity rather than accuracy) leads to better reliability and improved scaling law error. We also find that filtering noisy subtasks, to improve an aggregate signal-to-noise ratio, leads to more reliable multi-task evaluations. We also find that averaging the output of a model's intermediate checkpoints to reduce noise leads to consistent improvements. We conclude by recommending that those creating new benchmarks, or selecting which existing benchmarks to use, aim for high signal and low noise. We use 30 benchmarks for these experiments, and 375 open-weight language models from 60M to 32B parameters, resulting in a new, publicly available dataset of 900K evaluation benchmark results, totaling 200M instances.

Frontier LLM companies have repeatedly assured courts and regulators that their models do not store copies of training data. They further rely on safety alignment strategies via RLHF, system prompts, and output filters to block verbatim regurgitation of copyrighted works, and have cited the efficacy of these measures in their legal defenses against copyright infringement claims. We show that finetuning bypasses these protections: by training models to expand plot summaries into full text, a task naturally suited for commercial writing assistants, we cause GPT-4o, Gemini-2.5-Pro, and DeepSeek-V3.1 to reproduce up to 85-90% of held-out copyrighted books, with single verbatim spans exceeding 460 words, using only semantic descriptions as prompts and no actual book text. This extraction generalizes across authors: finetuning exclusively on Haruki Murakami's novels unlocks verbatim recall of copyrighted books from over 30 unrelated authors. The effect is not specific to any training author or corpus: random author pairs and public-domain finetuning data produce comparable extraction, while finetuning on synthetic text yields near-zero extraction, indicating that finetuning on individual authors' works reactivates latent memorization from pretraining. Three models from different providers memorize the same books in the same regions (r ge 0.90), pointing to an industry-wide vulnerability. Our findings offer compelling evidence that model weights store copies of copyrighted works and that the security failures that manifest after finetuning on individual authors' works undermine a key premise of recent fair use rulings, where courts have conditioned favorable outcomes on the adequacy of measures preventing reproduction of protected expression.

Existing foundation models (FMs) in the medical domain often require extensive fine-tuning or rely on training resource-intensive decoders, while many existing encoders are pretrained with objectives biased toward specific tasks. This illustrates a need for a strong, task-agnostic foundation model that requires minimal fine-tuning beyond feature extraction. In this work, we introduce a suite of task-agnostic pretraining of CT foundation models (TAP-CT): a simple yet effective adaptation of Vision Transformers (ViTs) and DINOv2 for volumetric data, enabling scalable self-supervised pretraining directly on 3D CT volumes. Our approach incorporates targeted modifications to patch embeddings, positional encodings, and volumetric augmentations, making the architecture depth-aware while preserving the simplicity of the underlying architectures. We show that large-scale 3D pretraining on an extensive in-house CT dataset (105K volumes) yields stable, robust frozen representations that generalize strongly across downstream tasks. To promote transparency and reproducibility, and to establish a powerful, low-resource baseline for future research in medical imaging, we will release all pretrained models, experimental configurations, and downstream benchmark code at https://huggingface.co/fomofo/tap-ct-b-3d.

Prompt engineering can significantly improve the performance of large language models (LLMs), with automated prompt optimization (APO) gaining significant attention due to the time-consuming and laborious nature of manual prompt design. However, much of the existing work in APO overlooks task-specific characteristics, resulting in prompts that lack domain specificity and are not well-suited for task-specific optimization. In this paper, we introduce TAPO, a multitask-aware prompt optimization framework composed of three key modules. First, a task-aware metric selection module is proposed to enhance task-specific prompt generation capabilities. Second, we present a multi-metrics evaluation module to jointly evaluate prompts from multiple perspectives. Third, an evolution-based optimization framework is introduced for automatic prompt refinement, which improves adaptability across various tasks. Extensive experiments on six datasets demonstrate the effectiveness of our approach, and our code is publicly available.

Spatio-temporal scene graphs provide a principled representation for modeling evolving object interactions, yet existing methods remain fundamentally frame-centric: they reason only about currently visible objects, discard entities upon occlusion, and operate in 2D. To address this, we first introduce ActionGenome4D, a dataset that upgrades Action Genome videos into 4D scenes via feed-forward 3D reconstruction, world-frame oriented bounding boxes for every object involved in actions, and dense relationship annotations including for objects that are temporarily unobserved due to occlusion or camera motion. Building on this data, we formalize World Scene Graph Generation (WSGG), the task of constructing a world scene graph at each timestamp that encompasses all interacting objects in the scene, both observed and unobserved. We then propose three complementary methods, each exploring a different inductive bias for reasoning about unobserved objects: PWG (Persistent World Graph), which implements object permanence via a zero-order feature buffer; MWAE (Masked World Auto-Encoder), which reframes unobserved-object reasoning as masked completion with cross-view associative retrieval; and 4DST (4D Scene Transformer), which replaces the static buffer with differentiable per-object temporal attention enriched by 3D motion and camera-pose features. We further design and evaluate the performance of strong open-source Vision-Language Models on the WSGG task via a suite of Graph RAG-based approaches, establishing baselines for unlocalized relationship prediction. WSGG thus advances video scene understanding toward world-centric, temporally persistent, and interpretable scene reasoning.

We propose a hierarchical entity-centric framework for offline Goal-Conditioned Reinforcement Learning (GCRL) that combines subgoal decomposition with factored structure to solve long-horizon tasks in domains with multiple entities. Achieving long-horizon goals in complex environments remains a core challenge in Reinforcement Learning (RL). Domains with multiple entities are particularly difficult due to their combinatorial complexity. GCRL facilitates generalization across goals and the use of subgoal structure, but struggles with high-dimensional observations and combinatorial state-spaces, especially under sparse reward. We employ a two-level hierarchy composed of a value-based GCRL agent and a factored subgoal-generating conditional diffusion model. The RL agent and subgoal generator are trained independently and composed post hoc through selective subgoal generation based on the value function, making the approach modular and compatible with existing GCRL algorithms. We introduce new variations to benchmark tasks that highlight the challenges of multi-entity domains, and show that our method consistently boosts performance of the underlying RL agent on image-based long-horizon tasks with sparse rewards, achieving over 150% higher success rates on the hardest task in our suite and generalizing to increasing horizons and numbers of entities. Rollout videos are provided at: https://sites.google.com/view/hecrl

We introduce FlexTab, a flexible encoder-decoder architecture for in-context learning on tabular data that pairs a single, task-agnostic encoder with a suite of task-specific decoders. Unlike existing tabular in-context learners, which entangle feature representations with a specific prediction target, our design produces target-agnostic row embeddings that can be leveraged across a wide range of downstream tasks within a table-native in-context learning setup. We demonstrate this flexibility on six distinct problems: classification, regression, anomaly detection, clustering, entity matching, and entity classification in relational databases. Both the encoder and the task-specific decoders are trained on a large corpus of real-world, unlabeled tables. FlexTab achieves state-of-the-art performance on classification, regression, anomaly detection and entity matching, while remaining competitive with specialized models on entity classification in a relational setting. These results demonstrate that a single shared encoder, paired with task-specific decoders, can serve as an effective general-purpose backbone for diverse tabular prediction problems. The inference code and checkpoints will be made publicly available at https://github.com/SAP-samples/flextab.

Omni-modal retrieval promises a single embedding space for text, image, video, document, and audio inputs, but building such a unified retriever is difficult since these modalities differ in data distribution, architecture, and optimization dynamics. In this work, we present Conan-embedding-v3, a decouple--fuse--recover framework for omni-modal retrieval. Conan-embedding-v3 first trains modality specialists independently and fuses their task vectors into a single dense backbone, a strategy we call Decoupled Specialist Fusion. We show that this fusion composes visual, video, and document retrieval capabilities, but also exposes a failure mode for projector-based modalities: when audio is attached through an external encoder and projector, fusing the backbone leaves the projector calibrated to the audio-specialist backbone, causing a large audio retrieval regression despite copying all audio-specific modules unchanged. We call this failure Projector Drift. To repair it, Conan-embedding-v3 applies Projector Recovery (i.e., full-parameter fine-tuning of the projector while keeping the backbone frozen) followed by balanced multi-modal rehearsal. The resulting model supports these retrieval pathways in one backbone, achieving 74.9 scores on MMEB while obtaining 55.61 on the 30-task MAEB audio suite.

We introduce a novel suite of state-of-the-art bilingual text embedding models that are designed to support English and another target language. These models are capable of processing lengthy text inputs with up to 8192 tokens, making them highly versatile for a range of natural language processing tasks such as text retrieval, clustering, and semantic textual similarity (STS) calculations. By focusing on bilingual models and introducing a unique multi-task learning objective, we have significantly improved the model performance on STS tasks, which outperforms the capabilities of existing multilingual models in both target language understanding and cross-lingual evaluation tasks. Moreover, our bilingual models are more efficient, requiring fewer parameters and less memory due to their smaller vocabulary needs. Furthermore, we have expanded the Massive Text Embedding Benchmark (MTEB) to include benchmarks for German and Spanish embedding models. This integration aims to stimulate further research and advancement in text embedding technologies for these languages.

Document intelligence requires accurate text extraction and reliable reasoning over document content. We introduce DISCO, a Document Intelligence Suite for COmparative Evaluation, that evaluates optical character recognition (OCR) pipelines and vision-language models (VLMs) separately on parsing and question answering across diverse document types, including handwritten text, multilingual scripts, medical forms, infographics, and multi-page documents. Our evaluation shows that performance varies substantially across tasks and document characteristics, underscoring the need for complexity-aware approach selection. OCR pipelines are generally more reliable for handwriting and for long or multi-page documents, where explicit text grounding supports text-heavy reasoning, while VLMs perform better on multilingual text and visually rich layouts. Task-aware prompting yields mixed effects, improving performance on some document types while degrading it on others. These findings provide empirical guidance for selecting document processing strategies based on document structure and reasoning demands.

Bayesian predictive inference provides a principled framework for uncertainty quantification, data efficiency, and robust generalization. However, exact inference is often intractable, and scalable approximations may remain computationally expensive or require restrictive modeling assumptions that degrade predictive performance. Prior-Data Fitted and in-context models have recently emerged as an amortized alternative by learning to map datasets directly to predictive distributions, but existing approaches are tightly coupled to the support of the training prior and lack explicit mechanisms for adapting to new priors at test time, resulting in limited robustness under distribution shift. We introduce a multi-task in-context learning framework for amortized hierarchical Bayesian predictive inference that explicitly represents prior information as a prefix of in-context datasets. A transformer trained on sequences of prior and target tasks learns to adapt its predictions across families of priors. On a suite of evaluations with increasing difficulty, including out-of-meta-distribution priors and priors with high-dimensional latent structures, our method matches oracle Bayesian predictors while being orders of magnitude faster. We further demonstrate its practical relevance on a real-world spatiotemporal temperature prediction benchmark. Code is available at https://github.com/martianmartina/multi-task-bayesian-icl/.

Adapting pre-trained models with broad capabilities has become standard practice for learning a wide range of downstream tasks. The typical approach of fine-tuning different models for each task is performant, but incurs a substantial memory cost. To efficiently learn multiple downstream tasks we introduce Task Adaptive Parameter Sharing (TAPS), a general method for tuning a base model to a new task by adaptively modifying a small, task-specific subset of layers. This enables multi-task learning while minimizing resources used and competition between tasks. TAPS solves a joint optimization problem which determines which layers to share with the base model and the value of the task-specific weights. Further, a sparsity penalty on the number of active layers encourages weight sharing with the base model. Compared to other methods, TAPS retains high accuracy on downstream tasks while introducing few task-specific parameters. Moreover, TAPS is agnostic to the model architecture and requires only minor changes to the training scheme. We evaluate our method on a suite of fine-tuning tasks and architectures (ResNet, DenseNet, ViT) and show that it achieves state-of-the-art performance while being simple to implement.

As AI agents become increasingly capable of complex, long-horizon reasoning, rigorous and holistic evaluation is essential for measuring progress toward real-world healthcare applications. We introduce HealthAgentBench, a suite of 54 agentic healthcare tasks across 7 categories each with its unique environment. The benchmark suite spans diverse workflows throughout the patient journey and a broad range of modalities. Each task is designed to replicate an end-to-end clinical workflow: given minimal instructions, an agent must explore raw healthcare data, operate within a complex environment, and execute multi-step solutions that go beyond naive prompting. A final task success rate is reported to provide a single, interpretable metric for HealthAgentBench overall performance for each agent. Evaluating frontier agents on HealthAgentBench, we find that overall task success rate remains low, underscoring the difficulty of the suite. The strongest and the most cost effective agent, Codex GPT-5.5, achieves only approximately 42% success rate. Beyond aggregate performance, HealthAgentBench reveals nuanced strengths and weaknesses across task categories. Frontier agents show promise in automatically developing research modeling pipelines over EHR data, but medical imaging remains especially challenging, particularly for Claude Code models, while Codex GPT-5.5 shows emerging capability. Tasks that combine large search spaces with compositional reasoning requirements remain difficult for all current agents. Together, these results suggest that HealthAgentBench provides a challenging and realistic benchmark with substantial room for future progress. We release our benchmark at https://github.com/microsoft/HealthAgentBench.

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

Webpages have been a rich, scalable resource for vision-language and language only tasks. Yet only pieces of webpages are kept: image-caption pairs, long text articles, or raw HTML, never all in one place. Webpage tasks have resultingly received little attention and structured image-text data left underused. To study multimodal webpage understanding, we introduce the Wikipedia Webpage suite (WikiWeb2M) of 2M pages. We verify its utility on three generative tasks: page description generation, section summarization, and contextual image captioning. We design a novel attention mechanism Prefix Global, which selects the most relevant image and text content as global tokens to attend to the rest of the webpage for context. By using page structure to separate such tokens, it performs better than full attention with lower computational complexity. Experiments show that the new annotations from WikiWeb2M improve task performance compared to data from prior work. We also include ablations on sequence length, input features, and model size.

Large language models trained on massive corpora of data from the web can memorize and reproduce sensitive or private data raising both legal and ethical concerns. Unlearning, or tuning models to forget information present in their training data, provides us with a way to protect private data after training. Although several methods exist for such unlearning, it is unclear to what extent they result in models equivalent to those where the data to be forgotten was never learned in the first place. To address this challenge, we present TOFU, a Task of Fictitious Unlearning, as a benchmark aimed at helping deepen our understanding of unlearning. We offer a dataset of 200 diverse synthetic author profiles, each consisting of 20 question-answer pairs, and a subset of these profiles called the forget set that serves as the target for unlearning. We compile a suite of metrics that work together to provide a holistic picture of unlearning efficacy. Finally, we provide a set of baseline results from existing unlearning algorithms. Importantly, none of the baselines we consider show effective unlearning motivating continued efforts to develop approaches for unlearning that effectively tune models so that they truly behave as if they were never trained on the forget data at all.

Large language models show improved downstream task performance when prompted to generate step-by-step reasoning to justify their final answers. These reasoning steps greatly improve model interpretability and verification, but objectively studying their correctness (independent of the final answer) is difficult without reliable methods for automatic evaluation. We simply do not know how often the stated reasoning steps actually support the final end task predictions. In this work, we present ROSCOE, a suite of interpretable, unsupervised automatic scores that improve and extend previous text generation evaluation metrics. To evaluate ROSCOE against baseline metrics, we design a typology of reasoning errors and collect synthetic and human evaluation scores on commonly used reasoning datasets. In contrast with existing metrics, ROSCOE can measure semantic consistency, logicality, informativeness, fluency, and factuality - among other traits - by leveraging properties of step-by-step rationales. We empirically verify the strength of our metrics on five human annotated and six programmatically perturbed diagnostics datasets - covering a diverse set of tasks that require reasoning skills and show that ROSCOE can consistently outperform baseline metrics.

Log analysis is crucial for ensuring the orderly and stable operation of information systems, particularly in the field of Artificial Intelligence for IT Operations (AIOps). Large Language Models (LLMs) have demonstrated significant potential in natural language processing tasks. In the AIOps domain, they excel in tasks such as anomaly detection, root cause analysis of faults, operations and maintenance script generation, and alert information summarization. However, the performance of current LLMs in log analysis tasks remains inadequately validated. To address this gap, we introduce LogEval, a comprehensive benchmark suite designed to evaluate the capabilities of LLMs in various log analysis tasks for the first time. This benchmark covers tasks such as log parsing, log anomaly detection, log fault diagnosis, and log summarization. LogEval evaluates each task using 4,000 publicly available log data entries and employs 15 different prompts for each task to ensure a thorough and fair assessment. By rigorously evaluating leading LLMs, we demonstrate the impact of various LLM technologies on log analysis performance, focusing on aspects such as self-consistency and few-shot contextual learning. We also discuss findings related to model quantification, Chinese-English question-answering evaluation, and prompt engineering. These findings provide insights into the strengths and weaknesses of LLMs in multilingual environments and the effectiveness of different prompt strategies. Various evaluation methods are employed for different tasks to accurately measure the performance of LLMs in log analysis, ensuring a comprehensive assessment. The insights gained from LogEvals evaluation reveal the strengths and limitations of LLMs in log analysis tasks, providing valuable guidance for researchers and practitioners.

Enzymes are important proteins that catalyze chemical reactions. In recent years, machine learning methods have emerged to predict enzyme function from sequence; however, there are no standardized benchmarks to evaluate these methods. We introduce CARE, a benchmark and dataset suite for the Classification And Retrieval of Enzymes (CARE). CARE centers on two tasks: (1) classification of a protein sequence by its enzyme commission (EC) number and (2) retrieval of an EC number given a chemical reaction. For each task, we design train-test splits to evaluate different kinds of out-of-distribution generalization that are relevant to real use cases. For the classification task, we provide baselines for state-of-the-art methods. Because the retrieval task has not been previously formalized, we propose a method called Contrastive Reaction-EnzymE Pretraining (CREEP) as one of the first baselines for this task and compare it to the recent method, CLIPZyme. CARE is available at https://github.com/jsunn-y/CARE/.

LLM agents hold significant promise for advancing scientific research. To accelerate this progress, we introduce AIRS-Bench (the AI Research Science Benchmark), a suite of 20 tasks sourced from state-of-the-art machine learning papers. These tasks span diverse domains, including language modeling, mathematics, bioinformatics, and time series forecasting. AIRS-Bench tasks assess agentic capabilities over the full research lifecycle -- including idea generation, experiment analysis and iterative refinement -- without providing baseline code. The AIRS-Bench task format is versatile, enabling easy integration of new tasks and rigorous comparison across different agentic frameworks. We establish baselines using frontier models paired with both sequential and parallel scaffolds. Our results show that agents exceed human SOTA in four tasks but fail to match it in sixteen others. Even when agents surpass human benchmarks, they do not reach the theoretical performance ceiling for the underlying tasks. These findings indicate that AIRS-Bench is far from saturated and offers substantial room for improvement. We open-source the AIRS-Bench task definitions and evaluation code to catalyze further development in autonomous scientific research.

Training on verifiable symbolic data is a promising way to expand the reasoning frontier of language models beyond what standard pre-training corpora provide. Yet existing procedural generators often rely on fixed puzzles or templates and do not deliver the distributional breadth needed at scale. We introduce Reasoning Core, a scalable suite that procedurally generates verifiable symbolic reasoning data across core formal domains: PDDL planning over randomized domains, first-order logic with equality, context-free grammar parsing and generation, causal reasoning over random Bayesian networks, and systems of equations. Each task is paired with an external solver for rigorous verification and admits continuous difficulty control for curriculum design. Examples can optionally include solver-derived reasoning traces, enabling supervised training from the earliest pre-training stages, and the same interface provides verifiable reward functions for reinforcement learning. Our experiments show that mixing Reasoning Core data into pre-training improves downstream reasoning while preserving, or slightly improving, language modeling quality. Zero-shot evaluations confirm these tasks challenge frontier models such as GPT-5. The code and data are publicly available under the MIT license.

Native visual reasoning treats visual generation as the medium of reasoning itself: visual states (i.e. images and videos) are not merely inputs to be understood or outputs to be rendered, but first-class substrates for problem solving beyond language. Yet progress remains bottlenecked by the lack of scalable training tasks, reliable feedback, and controlled comparisons across generative substrates. In this work, we introduce VBVR-Pro, a closed-loop testbed that makes native visual reasoning through generation trainable, verifiable, optimizable, and experimentally controllable. 1) Task scaling. VBVR-Pro turns visual reasoning into a controlled task space of 300 procedurally generated tasks. Models trained on VBVR-Pro show strong transfer beyond the proposed suite across seven external visual reasoning benchmarks such as RISE-Video, MME-CoF-Pro, and BabyVision. 2) Verifiable rewards. VBVR-Pro provides verifiable reward scorers for task-grounded evaluation. Through a systematic study of leading MLLMs as judges, we identify recurring failure modes of the prevalent VLM-as-a-judge paradigm. In contrast, the proposed scorers are grounded in deterministic, task-specific rules, achieve fine-grained alignment with human judgments. Importantly, they serve as reliable reward signals for large-scale multi-task reinforcement learning and demonstrate stronger post-RL performance across visual reasoning tasks. 3) Mechanism study. VBVR-Pro enables controlled modality studies across more than 30 image, video, and interleaved generators. Our analysis shows that video generation remains strongest for tasks requiring persistent spatiotemporal state tracking, while interleaved generation provides a compute-efficient alternative. Critically, ablations and probing suggest the presence of vision-native trajectories that are crucial to visual reasoning. We release all data, models, scorers, and code.

Tracking depression from multi-session counseling dialogues requires estimating both current symptom severity and how it changes across sessions. Yet progress on this task is constrained by the scarcity of longitudinal counseling data with standardized session-level depression labels. Existing resources typically provide either multi-session conversations without depression labels or labeled interviews in a single session. Building such a benchmark poses three challenges: maintaining longitudinal consistency and diversity, grounding symptom progression in empirical patterns, and expressing controlled depression states naturally without exposing target labels. To address these challenges, we introduce LongCounsel-8, a benchmark suite of three independently generated datasets totaling 7,749 five-session counseling trajectories, grounded in real-world client profiles, depression trajectories, symptom compositions, and counseling patterns. We combine profile-grounded simulation, empirically informed state construction, and indirect behavioral realization to address these challenges. Across the benchmark, simulated self-reports closely recover the controlled states, supporting label fidelity. Experiments on existing depression tracking methods reveal three key findings: (1) lower single-session score error does not guarantee accurate identification of trend, i.e., improvement or worsening; (2) existing methods are consistently less reliable on worsening trajectories; and (3) additional session history may reduce the accuracy of trend prediction. Together, these findings establish LongCounsel-8 as a foundation for advancing depression assessment from static, single-session prediction toward reliable longitudinal tracking of mental-health change.

Open-world aerial object-goal search is a foundational yet challenging task, requiring aerial agents to autonomously explore large-scale, unstructured three-dimensional environments and reach target objects specified by semantic descriptions or reference images, rather than following route-specific instructions. However, research in this task remains at a nascent stage and relies on small, environment-specific benchmarks with heterogeneous action spaces and data formats. These limitations hinder large-scale training and cross-benchmark evaluation, constraining the scalability and generalizability of aerial agents. To address this problem, we propose AerialDojo-200K, a large-scale benchmark suite for open-world aerial object-goal search, with 3 times as many scenes and 18.7 times as many task instances as the largest existing benchmark for this task. Specifically, we construct 42 simulation scenes spanning four scene families and 21 scene types, including 18 urban, 12 natural, six infrastructure, and six disaster scenes. To ensure data quality, 12 annotators spent two months manually annotating 109 landmarks, 2099 target objects, and 2099 object anchors across these scenes. We further construct 205,732 task instances, comprising over 100K semantic-goal and over 100K image-goal instances across Base, Standard, and Long-Horizon settings. Each task instance includes a collision-free reference trajectory and corresponding multi-view video recordings. We also develop a unified evaluation framework with a scene partition comprising 21 in-distribution scenes and 21 out-of-distribution scenes. Finally, our evaluation of five open-source and four closed-source multimodal large language models reveals that there is still a long way to go toward achieving general-purpose aerial agents. All can be found at https://fengtt42.github.io/AerialDojo/.

Recursive self-improvement requires models not only to acquire useful training experience, but also to decide when their current experience has ceased to be useful. Existing methods often focus on what experience to generate next, while paying less attention to when fresh experience should be introduced. Many also construct training targets through self-consistency or majority voting, making supervision depend on the model's own uncertain predictions. We address these two questions separately: training dynamics determine when to acquire new experience, while the answers to new tasks remain independently checkable. A-RSI monitors reward variation and accuracy in training logs to refresh a task pool when its learning signal declines, without consulting validation or test scores. A static library of model-authored programs produces multimodal questions and derives their reference answers; we specify the task-generation contract, automatically verify chosen properties of the reward and refresh rules, and separately audit the generated answers. The training loop uses no external training examples or human-written answer labels. On the primary eleven-benchmark suite, A-RSI raises the performance average from 57.77 to 61.05 with 800 fresh tasks.

Detecting online hate is a complex task, and low-performing models have harmful consequences when used for sensitive applications such as content moderation. Emoji-based hate is an emerging challenge for automated detection. We present HatemojiCheck, a test suite of 3,930 short-form statements that allows us to evaluate performance on hateful language expressed with emoji. Using the test suite, we expose weaknesses in existing hate detection models. To address these weaknesses, we create the HatemojiBuild dataset using a human-and-model-in-the-loop approach. Models built with these 5,912 adversarial examples perform substantially better at detecting emoji-based hate, while retaining strong performance on text-only hate. Both HatemojiCheck and HatemojiBuild are made publicly available. See our Github Repository (https://github.com/HannahKirk/Hatemoji). HatemojiCheck, HatemojiBuild, and the final Hatemoji Model are also available on HuggingFace (https://huggingface.co/datasets/HannahRoseKirk/).

For natural language understanding (NLU) technology to be maximally useful, both practically and as a scientific object of study, it must be general: it must be able to process language in a way that is not exclusively tailored to any one specific task or dataset. In pursuit of this objective, we introduce the General Language Understanding Evaluation benchmark (GLUE), a tool for evaluating and analyzing the performance of models across a diverse range of existing NLU tasks. GLUE is model-agnostic, but it incentivizes sharing knowledge across tasks because certain tasks have very limited training data. We further provide a hand-crafted diagnostic test suite that enables detailed linguistic analysis of NLU models. We evaluate baselines based on current methods for multi-task and transfer learning and find that they do not immediately give substantial improvements over the aggregate performance of training a separate model per task, indicating room for improvement in developing general and robust NLU systems.

Real-time multi-task multi-model (MTMM) workloads, a new form of deep learning inference workloads, are emerging for applications areas like extended reality (XR) to support metaverse use cases. These workloads combine user interactivity with computationally complex machine learning (ML) activities. Compared to standard ML applications, these ML workloads present unique difficulties and constraints. Real-time MTMM workloads impose heterogeneity and concurrency requirements on future ML systems and devices, necessitating the development of new capabilities. This paper begins with a discussion of the various characteristics of these real-time MTMM ML workloads and presents an ontology for evaluating the performance of future ML hardware for XR systems. Next, we present XRBENCH, a collection of MTMM ML tasks, models, and usage scenarios that execute these models in three representative ways: cascaded, concurrent, and cascaded-concurrent for XR use cases. Finally, we emphasize the need for new metrics that capture the requirements properly. We hope that our work will stimulate research and lead to the development of a new generation of ML systems for XR use cases. XRBench is available as an open-source project: https://github.com/XRBench

Given the ubiquitous nature of numbers in text, reasoning with numbers to perform simple calculations is an important skill of AI systems. While many datasets and models have been developed to this end, state-of-the-art AI systems are brittle; failing to perform the underlying mathematical reasoning when they appear in a slightly different scenario. Drawing inspiration from GLUE that was proposed in the context of natural language understanding, we propose NumGLUE, a multi-task benchmark that evaluates the performance of AI systems on eight different tasks, that at their core require simple arithmetic understanding. We show that this benchmark is far from being solved with neural models including state-of-the-art large-scale language models performing significantly worse than humans (lower by 46.4%). Further, NumGLUE promotes sharing knowledge across tasks, especially those with limited training data as evidenced by the superior performance (average gain of 3.4% on each task) when a model is jointly trained on all the tasks as opposed to task-specific modeling. Finally, we hope that NumGLUE will encourage systems that perform robust and general arithmetic reasoning within language, a first step towards being able to perform more complex mathematical reasoning.

Real-world household tasks present significant challenges for mobile manipulation robots. An analysis of existing robotics benchmarks reveals that successful task performance hinges on three key whole-body control capabilities: bimanual coordination, stable and precise navigation, and extensive end-effector reachability. Achieving these capabilities requires careful hardware design, but the resulting system complexity further complicates visuomotor policy learning. To address these challenges, we introduce the BEHAVIOR Robot Suite (BRS), a comprehensive framework for whole-body manipulation in diverse household tasks. Built on a bimanual, wheeled robot with a 4-DoF torso, BRS integrates a cost-effective whole-body teleoperation interface for data collection and a novel algorithm for learning whole-body visuomotor policies. We evaluate BRS on five challenging household tasks that not only emphasize the three core capabilities but also introduce additional complexities, such as long-range navigation, interaction with articulated and deformable objects, and manipulation in confined spaces. We believe that BRS's integrated robotic embodiment, data collection interface, and learning framework mark a significant step toward enabling real-world whole-body manipulation for everyday household tasks. BRS is open-sourced at https://behavior-robot-suite.github.io/

Recent advances in multimodal large language models (MLLMs) have opened new opportunities for embodied intelligence, enabling multimodal understanding, reasoning, and interaction, as well as continuous spatial decision-making. Nevertheless, current MLLM-based embodied systems face two critical limitations. First, Geometric Adaptability Gap: models trained solely on 2D inputs or with hard-coded 3D geometry injection suffer from either insufficient spatial information or restricted 2D generalization, leading to poor adaptability across tasks with diverse spatial demands. Second, Embodiment Constraint Gap: prior work often neglects the physical constraints and capacities of real robots, resulting in task plans that are theoretically valid but practically infeasible.To address these gaps, we introduce OmniEVA -- an embodied versatile planner that enables advanced embodied reasoning and task planning through two pivotal innovations: (1) a Task-Adaptive 3D Grounding mechanism, which introduces a gated router to perform explicit selective regulation of 3D fusion based on contextual requirements, enabling context-aware 3D grounding for diverse embodied tasks. (2) an Embodiment-Aware Reasoning framework that jointly incorporates task goals and embodiment constraints into the reasoning loop, resulting in planning decisions that are both goal-directed and executable. Extensive experimental results demonstrate that OmniEVA not only achieves state-of-the-art general embodied reasoning performance, but also exhibits a strong ability across a wide range of downstream scenarios. Evaluations of a suite of proposed embodied benchmarks, including both primitive and composite tasks, confirm its robust and versatile planning capabilities. Project page: https://omnieva.github.io

Multi-task learning (MTL) enables a joint model to capture commonalities across multiple tasks, reducing computation costs and improving data efficiency. However, a major challenge in MTL optimization is task conflicts, where the task gradients differ in direction or magnitude, limiting model performance compared to single-task counterparts. Sharpness-aware minimization (SAM) minimizes task loss while simultaneously reducing the sharpness of the loss landscape. Our empirical observations show that SAM effectively mitigates task conflicts in MTL. Motivated by these findings, we explore integrating SAM into MTL but face two key challenges. While both the average loss gradient and individual task gradients-referred to as global and local information-contribute to SAM, how to combine them remains unclear. Moreover, directly computing each task gradient introduces significant computational and memory overheads. To address these challenges, we propose SAMO, a lightweight Sharpness-Aware Multi-task Optimization approach, that leverages a joint global-local perturbation. The local perturbations are approximated using only forward passes and are layerwise normalized to improve efficiency. Extensive experiments on a suite of multi-task benchmarks demonstrate both the effectiveness and efficiency of our method. Code is available at https://github.com/OptMN-Lab/SAMO.

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

Transforming static research papers into dynamic media such as posters, slides, and videos is essential for effective dissemination but remains a labor-intensive challenge. Existing automated approaches often treat these formats in isolation and consequently fail to maintain semantic consistency across the entire presentation suite. We address this fragmentation by formalizing the task of unified presentation suite generation and proposing OmniPresent to orchestrate the creation of coherent deliverables. Our framework adopts a renderable HTML representation to enable centralized content planning and a self-correcting verify-and-repair loop that actively resolves conflicts across modalities. We further facilitate scalable research in this domain by releasing OmniPreBench, a comprehensive dataset comprising over one thousand papers with paired artifacts, and establishing a rigorous VLM-based evaluation protocol. Empirical results confirm that our method generates high-quality and faithful presentation suites that significantly surpass strong baselines in both accuracy and visual appeal.

This paper explores the task of identifying the overall sentiment expressed towards volitional entities (persons and organizations) in a document -- what we refer to as Entity-Level Sentiment Analysis (ELSA). While identifying sentiment conveyed towards an entity is well researched for shorter texts like tweets, we find little to no research on this specific task for longer texts with multiple mentions and opinions towards the same entity. This lack of research would be understandable if ELSA can be derived from existing tasks and models. To assess this, we annotate a set of professional reviews for their overall sentiment towards each volitional entity in the text. We sample from data already annotated for document-level, sentence-level, and target-level sentiment in a multi-domain review corpus, and our results indicate that there is no single proxy task that provides this overall sentiment we seek for the entities at a satisfactory level of performance. We present a suite of experiments aiming to assess the contribution towards ELSA provided by document-, sentence-, and target-level sentiment analysis, and provide a discussion of their shortcomings. We show that sentiment in our dataset is expressed not only with an entity mention as target, but also towards targets with a sentiment-relevant relation to a volitional entity. In our data, these relations extend beyond anaphoric coreference resolution, and our findings call for further research of the topic. Finally, we also present a survey of previous relevant work.

Mobile task automation is an attractive technique that aims to enable voice-based hands-free user interaction with smartphones. However, existing approaches suffer from poor scalability due to the limited language understanding ability and the non-trivial manual efforts required from developers or end-users. The recent advance of large language models (LLMs) in language understanding and reasoning inspires us to rethink the problem from a model-centric perspective, where task preparation, comprehension, and execution are handled by a unified language model. In this work, we introduce AutoDroid, a mobile task automation system that can handle arbitrary tasks on any Android application without manual efforts. The key insight is to combine the commonsense knowledge of LLMs and domain-specific knowledge of apps through automated dynamic analysis. The main components include a functionality-aware UI representation method that bridges the UI with the LLM, exploration-based memory injection techniques that augment the app-specific domain knowledge of LLM, and a multi-granularity query optimization module that reduces the cost of model inference. We integrate AutoDroid with off-the-shelf LLMs including online GPT-4/GPT-3.5 and on-device Vicuna, and evaluate its performance on a new benchmark for memory-augmented Android task automation with 158 common tasks. The results demonstrated that AutoDroid is able to precisely generate actions with an accuracy of 90.9%, and complete tasks with a success rate of 71.3%, outperforming the GPT-4-powered baselines by 36.4% and 39.7%. The demo, benchmark suites, and source code of AutoDroid will be released at url{https://autodroid-sys.github.io/}.

With increasing usage of generative models for text generation and widespread use of machine generated texts in various domains, being able to distinguish between human written and machine generated texts is a significant challenge. While existing models and proprietary systems focus on identifying whether given text is entirely human written or entirely machine generated, only a few systems provide insights at sentence or paragraph level at likelihood of being machine generated at a non reliable accuracy level, working well only for a set of domains and generators. This paper introduces few reliable approaches for the novel task of identifying which part of a given text is machine generated at a word level while comparing results from different approaches and methods. We present a comparison with proprietary systems , performance of our model on unseen domains' and generators' texts. The findings reveal significant improvements in detection accuracy along with comparison on other aspects of detection capabilities. Finally we discuss potential avenues for improvement and implications of our work. The proposed model is also well suited for detecting which parts of a text are machine generated in outputs of Instruct variants of many LLMs.

Generalist visual captioning goes beyond a simple appearance description task, but requires integrating a series of visual cues into a caption and handling various visual domains. In this task, current open-source models present a large performance gap with commercial ones, which limits various applications such as data synthesis. To bridge the gap, this paper proposes CapFlow, a novel multi-agent collaboration workflow. CapFlow demonstrates for the first time that, by capitalizing on open-source models, it is possible to achieve caption quality on par with GPT-4.1 in various domains with an 89.5% reduction in costs. By leveraging CapFlow as the data synthesizer, we produce high-quality visual captions from image and video domains at scale, and obtain a generalist visual captioner via fine-tuning, namely MetaCaptioner. Through extensive experiments, we show that MetaCaptioner not only achieves comparable captioning capabilities with commercial models but also reaches top-tier multimodal performance in the open-source community. We hope CapFlow and MetaCaptioner can benefit future multimodal research by providing a strong and cost-effective visual captioning solution.

Radiological analysis increasingly benefits from pretrained visual representations that can support heterogeneous downstream tasks across imaging modalities. In this work, we introduce OmniRad, a self-supervised radiological foundation model pretrained on 1.2 million medical images, designed with radiology-inspired principles emphasizing representation reuse and cross-task transferability. We evaluate the pretrained encoder under multiple downstream adaptation regimes, including lightweight task-specific adapters with a frozen backbone as well as full end-to-end fine-tuning for classification, allowing us to assess both representation quality and task-specific performance. OmniRad is evaluated on a broad suite of public benchmarks spanning classification and segmentation across multiple modalities. On the MedMNISTv2 collection, OmniRad improves classification F1 by up to 2.05% over competing foundation models. For dense prediction, OmniRad attains mean Dice score improvements across six MedSegBench datasets when using frozen representations. Qualitative analyses and latent-space visualizations suggest improved feature clustering and modality-related separation.

Online form filling is a common yet labor-intensive task involving extensive keyboard and mouse interactions. Despite the long-standing vision of automating this process with "one click", existing tools remain largely rule-based and lack generalizable, generative capabilities. Recent advances in Multimodal Large Language Models (MLLMs) have enabled promising agents for GUI-related tasks in general-purpose scenarios. However, they struggle with the unique challenges of form filling, such as flexible layouts and the difficulty of aligning textual instructions with on-screen fields. To bridge this gap, we formally define the form-filling task and propose FormFactory, an interactive benchmarking suite comprising a web-based interface, backend evaluation module, and carefully constructed dataset. Our benchmark covers diverse real-world scenarios, incorporates various field formats, and simulates high-fidelity form interactions. We conduct a comprehensive evaluation of state-of-the-art MLLMs and observe that no model surpasses 5% accuracy, underscoring the inherent difficulty of the task. These findings also reveal significant limitations in current models' visual layout reasoning and field-value alignment abilities. We hope our benchmark can serve as a stepping stone for further research into robust, practical form-filling agents.

Deep research agents have emerged as LLM-based systems designed to perform multi-step information seeking and reasoning over large, open-domain sources to answer complex questions by synthesizing information from multiple information sources. Given the complexity of the task and despite various recent efforts, evaluation of deep research agents remains fundamentally challenging. This paper identifies a list of requirements and optional properties for evaluating deep research agents. We observe that existing benchmarks do not satisfy all identified requirements. Inspired by prior research on TREC Total Recall Tracks, we introduce the task of Total Recall Question Answering and develop a framework for deep research agents evaluation that satisfies the identified criteria. Our framework constructs single-answer, total recall queries with precise evaluation and relevance judgments derived from a structured knowledge base paired with a text corpus, enabling large-scale data construction. Using this framework, we build TRQA, a deep research benchmark constructed from Wikidata-Wikipedia as a real-world source and a synthetically generated e-commerce knowledge base and corpus to mitigate the effects of data contamination. We benchmark the collection with representative retriever and deep research models and establish baseline retrieval and end-to-end results for future comparative evaluation.

Adapting to a continuously evolving environment is a safety-critical challenge inevitably faced by all autonomous driving systems. Existing image and video driving datasets, however, fall short of capturing the mutable nature of the real world. In this paper, we introduce the largest multi-task synthetic dataset for autonomous driving, SHIFT. It presents discrete and continuous shifts in cloudiness, rain and fog intensity, time of day, and vehicle and pedestrian density. Featuring a comprehensive sensor suite and annotations for several mainstream perception tasks, SHIFT allows investigating the degradation of a perception system performance at increasing levels of domain shift, fostering the development of continuous adaptation strategies to mitigate this problem and assess model robustness and generality. Our dataset and benchmark toolkit are publicly available at www.vis.xyz/shift.

Classification tasks in NLP are typically addressed by selecting a pre-trained language model (PLM) from a model hub, and fine-tuning it for the task at hand. However, given the very large number of PLMs that are currently available, a practical challenge is to determine which of them will perform best for a specific downstream task. With this paper, we introduce TransformerRanker, a lightweight library that efficiently ranks PLMs for classification tasks without the need for computationally costly fine-tuning. Our library implements current approaches for transferability estimation (LogME, H-Score, kNN), in combination with layer aggregation options, which we empirically showed to yield state-of-the-art rankings of PLMs (Garbas et al., 2024). We designed the interface to be lightweight and easy to use, allowing users to directly connect to the HuggingFace Transformers and Dataset libraries. Users need only select a downstream classification task and a list of PLMs to create a ranking of likely best-suited PLMs for their task. We make TransformerRanker available as a pip-installable open-source library https://github.com/flairNLP/transformer-ranker.

With the growth of high-quality data and advancement in visual pre-training paradigms, Video Foundation Models (VFMs) have made significant progress recently, demonstrating their remarkable performance on traditional video understanding benchmarks. However, the existing benchmarks (e.g. Kinetics) and their evaluation protocols are often limited by relatively poor diversity, high evaluation costs, and saturated performance metrics. In this paper, we build a comprehensive benchmark suite to address these issues, namely VideoEval. Specifically, we establish the Video Task Adaption Benchmark (VidTAB) and the Video Embedding Benchmark (VidEB) from two perspectives: evaluating the task adaptability of VFMs under few-shot conditions and assessing their representation power by directly applying to downstream tasks. With VideoEval, we conduct a large-scale study on 20 popular open-source vision foundation models. Our study reveals some insightful findings on VFMs: 1) overall, current VFMs exhibit weak generalization across diverse tasks, 2) increasing video data, whether labeled or weakly-labeled video-text pairs, does not necessarily improve task performance, 3) the effectiveness of some pre-training paradigms may not be fully validated in previous benchmarks, and 4) combining different pre-training paradigms can help improve the generalization capabilities. We believe this study serves as an important complement to the current evaluation for VFMs and offers valuable insights for the future research.

Vector Similarity Search (VSS) in high-dimensional spaces is rapidly emerging as core functionality in next-generation database systems for numerous data-intensive services -- from embedding lookups in large language models (LLMs), to semantic information retrieval and recommendation engines. Current benchmarks, however, evaluate VSS primarily on the recall-latency trade-off against a ground truth defined solely by distance metrics, neglecting how retrieval quality ultimately impacts downstream tasks. This disconnect can mislead both academic research and industrial practice. We present Iceberg, a holistic benchmark suite for end-to-end evaluation of VSS methods in realistic application contexts. From a task-centric view, Iceberg uncovers the Information Loss Funnel, which identifies three principal sources of end-to-end performance degradation: (1) Embedding Loss during feature extraction; (2) Metric Misuse, where distances poorly reflect task relevance; (3) Data Distribution Sensitivity, highlighting index robustness across skews and modalities. For a more comprehensive assessment, Iceberg spans eight diverse datasets across key domains such as image classification, face recognition, text retrieval, and recommendation systems. Each dataset, ranging from 1M to 100M vectors, includes rich, task-specific labels and evaluation metrics, enabling assessment of retrieval algorithms within the full application pipeline rather than in isolation. Iceberg benchmarks 13 state-of-the-art VSS methods and re-ranks them based on application-level metrics, revealing substantial deviations from traditional rankings derived purely from recall-latency evaluations. Building on these insights, we define a set of task-centric meta-features and derive an interpretable decision tree to guide practitioners in selecting and tuning VSS methods for their specific workloads.

The generalization capabilities of robotic manipulation policies are heavily influenced by the choice of visual representations. Existing approaches typically rely on representations extracted from pre-trained encoders, using two dominant types of features: global features, which summarize an entire image via a single pooled vector, and dense features, which preserve a patch-wise embedding from the final encoder layer. While widely used, both feature types mix task-relevant and irrelevant information, leading to poor generalization under distribution shifts, such as changes in lighting, textures, or the presence of distractors. In this work, we explore an intermediate structured alternative: Slot-Based Object-Centric Representations (SBOCR), which group dense features into a finite set of object-like entities. This representation permits to naturally reduce the noise provided to the robotic manipulation policy while keeping enough information to efficiently perform the task. We benchmark a range of global and dense representations against intermediate slot-based representations, across a suite of simulated and real-world manipulation tasks ranging from simple to complex. We evaluate their generalization under diverse visual conditions, including changes in lighting, texture, and the presence of distractors. Our findings reveal that SBOCR-based policies outperform dense and global representation-based policies in generalization settings, even without task-specific pretraining. These insights suggest that SBOCR is a promising direction for designing visual systems that generalize effectively in dynamic, real-world robotic environments.