Daily Papers

Artificial intelligence benchmarks are an important mechanism for measuring model progress and guiding deployment decisions. However, benchmarks quickly "saturate", making it difficult to differentiate models and diminishing their long-term value. In this study, we define benchmark saturation and analyze it across 60 language model benchmarks using 14 properties that relate to saturation. We find that nearly half of the our benchmarks exhibit saturation, with rates increasing with age. Further, we find that resilience to saturation is impacted by expert-curation, not by public test data. Our results suggest that design choices can extend benchmark longevity and inform more durable evaluation approaches.

Benchmarks play a central role in the development and governance of models, yet it is often unclear whether they actually measure the concepts they purport to measure (e.g., reasoning, refusal). We adapt convergent and discriminant validity from the social sciences into an approach for interrogating AI benchmarks, applying it to 56 capability and safety benchmarks across 53 models. We label benchmarks with substantively similar purported concepts to a shared assigned concept, and ask whether model rankings on benchmarks with the same assigned concept correlate more strongly than rankings on benchmarks with different assigned concepts. We ask analogous questions at the item level using item response theory (IRT) models. We find that correlations between model rankings on benchmarks with the same assigned safety concepts are often weak, suggesting these concepts may be conceptualized inconsistently across benchmarks. For assigned capability concepts (e.g., reasoning, knowledge), model rankings are often as strongly correlated among benchmarks with the same assigned concept as between benchmarks with different assigned concepts, suggesting these capability concepts may not discriminate well from one another. In some cases, benchmarks that share design elements (e.g., score format) correlate more strongly than benchmarks with the same assigned concept. Finally, some individual benchmarks correlate more strongly with benchmarks assigned a different concept than with benchmarks sharing their own assigned concept, suggesting they may measure a different concept than they purport to. For example, BBQ-accuracy correlates more strongly with benchmarks labeled reasoning than with benchmarks that share its assigned concept, bias. To support future empirical work on benchmark validity, we release our extensive dataset of model outputs and scores at the item- and benchmark-level.

AI models are increasingly prevalent in high-stakes environments, necessitating thorough assessment of their capabilities and risks. Benchmarks are popular for measuring these attributes and for comparing model performance, tracking progress, and identifying weaknesses in foundation and non-foundation models. They can inform model selection for downstream tasks and influence policy initiatives. However, not all benchmarks are the same: their quality depends on their design and usability. In this paper, we develop an assessment framework considering 46 best practices across an AI benchmark's lifecycle and evaluate 24 AI benchmarks against it. We find that there exist large quality differences and that commonly used benchmarks suffer from significant issues. We further find that most benchmarks do not report statistical significance of their results nor allow for their results to be easily replicated. To support benchmark developers in aligning with best practices, we provide a checklist for minimum quality assurance based on our assessment. We also develop a living repository of benchmark assessments to support benchmark comparability, accessible at betterbench.stanford.edu.

Benchmarks are pivotal in driving AI progress, and invalid benchmark questions frequently undermine their reliability. Manually identifying and correcting errors among thousands of benchmark questions is not only infeasible but also a critical bottleneck for reliable evaluation. In this work, we introduce a framework for systematic benchmark revision that leverages statistical analysis of response patterns to flag potentially invalid questions for further expert review. Our approach builds on a core assumption commonly used in AI evaluations that the mean score sufficiently summarizes model performance. This implies a unidimensional latent construct underlying the measurement experiment, yielding expected ranges for various statistics for each item. When empirically estimated values for these statistics fall outside the expected range for an item, the item is more likely to be problematic. Across nine widely used benchmarks, our method guides expert review to identify problematic questions with up to 84\% precision. In addition, we introduce an LLM-judge first pass to review questions, further reducing human effort. Together, these components provide an efficient and scalable framework for systematic benchmark revision.

Benchmark researchers and developers of large language models (LLMs) and other AI systems need to find relevant evaluations, locate their benchmark datasets and code, and understand the settings behind reported scores. We present Benchmark Radar, a living database and search engine for retrieval and discovery of AI benchmarks, covering LLM evaluation, agentic and tool-use benchmarks, coding, reasoning, safety, and domain-specific evaluations. The system combines daily discovery of benchmark papers, repositories, datasets, and releases with a searchable benchmark catalog, mentions in model cards and technical reports, and score histories. It retains source identities and citations so readers can inspect candidate benchmarks and their evaluation evidence. Daily discovery draws on 37 sources: 13 direct connectors and 24 first-party research and engineering feeds. The catalog contains 1,283 source records drawn from 4 benchmark catalogs and 12,916 numeric observations on 790 records. We describe collection and retrieval, audit the full catalog, and examine benchmark saturation, adoption trends, and the limits of score comparisons. A worked example walks through a complete prior-art search, showing how to query the catalog and inspect benchmark evidence when designing a new evaluation. We release the web dashboard with a benchmark leaderboard, a Pareto frontier view of score against measured use, saturation and trend views, daily feeds, downloadable evidence, a command-line interface (CLI) for offline queries, and reproducible analysis.

Rigorously evaluating machine intelligence against the broad spectrum of human general intelligence has become increasingly important and challenging in this era of rapid technological advance. Conventional AI benchmarks typically assess only narrow capabilities in a limited range of human activity. Most are also static, quickly saturating as developers explicitly or implicitly optimize for them. We propose that a more promising way to evaluate human-like general intelligence in AI systems is through a particularly strong form of general game playing: studying how and how well they play and learn to play all conceivable human games, in comparison to human players with the same level of experience, time, or other resources. We define a "human game" to be a game designed by humans for humans, and argue for the evaluative suitability of this space of all such games people can imagine and enjoy -- the "Multiverse of Human Games". Taking a first step towards this vision, we introduce the AI GameStore, a scalable and open-ended platform that uses LLMs with humans-in-the-loop to synthesize new representative human games, by automatically sourcing and adapting standardized and containerized variants of game environments from popular human digital gaming platforms. As a proof of concept, we generated 100 such games based on the top charts of Apple App Store and Steam, and evaluated seven frontier vision-language models (VLMs) on short episodes of play. The best models achieved less than 10\% of the human average score on the majority of the games, and especially struggled with games that challenge world-model learning, memory and planning. We conclude with a set of next steps for building out the AI GameStore as a practical way to measure and drive progress toward human-like general intelligence in machines.

Existing AI benchmarks lack the fidelity to assess economically meaningful progress on professional workflows. To evaluate frontier AI agents in a high-value, labor-intensive profession, we introduce BankerToolBench (BTB): an open-source benchmark of end-to-end analytical workflows routinely performed by junior investment bankers. To develop an ecologically valid benchmark grounded in representative work environments, we collaborated with 502 investment bankers from leading firms. BTB requires agents to execute senior banker requests by navigating data rooms, using industry tools (market data platform, SEC filings database), and generating multi-file deliverables--including Excel financial models, PowerPoint pitch decks, and PDF/Word reports. Completing a BTB task takes bankers up to 21 hours, underscoring the economic stakes of successfully delegating this work to AI. BTB enables automated evaluation of any LLM or agent, scoring deliverables against 100+ rubric criteria defined by veteran investment bankers to capture stakeholder utility. Testing 9 frontier models, we find that even the best-performing model (GPT-5.4) fails nearly half of the rubric criteria and bankers rate 0% of its outputs as client-ready. Our failure analysis reveals key obstacles (such as breakdowns in cross-artifact consistency) and improvement directions for agentic AI in high-stakes professional workflows.

Despite rapid progress on AI benchmarks, the real-world meaning of benchmark performance remains unclear. To quantify the capabilities of AI systems in terms of human capabilities, we propose a new metric: 50%-task-completion time horizon. This is the time humans typically take to complete tasks that AI models can complete with 50% success rate. We first timed humans with relevant domain expertise on a combination of RE-Bench, HCAST, and 66 novel shorter tasks. On these tasks, current frontier AI models such as Claude 3.7 Sonnet have a 50% time horizon of around 50 minutes. Furthermore, frontier AI time horizon has been doubling approximately every seven months since 2019, though the trend may have accelerated in 2024. The increase in AI models' time horizons seems to be primarily driven by greater reliability and ability to adapt to mistakes, combined with better logical reasoning and tool use capabilities. We discuss the limitations of our results -- including their degree of external validity -- and the implications of increased autonomy for dangerous capabilities. If these results generalize to real-world software tasks, extrapolation of this trend predicts that within 5 years, AI systems will be capable of automating many software tasks that currently take humans a month.

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

Current voice AI benchmarks typically evaluate isolated capabilities such as speech intelligibility, word error rate, or text-based dialogue quality, but they rarely test whether systems harness the acoustic information that distinguishes spoken language from its textual representation. To this end, we introduce the Real World Voice EQ Bench, a multidimensional benchmark for evaluating voice AI across text-to-speech (TTS), speech-to-speech (STS), speech understanding (SU), and automatic speech recognition (ASR). Our evaluations indicate that performance is highly dimension-specific. For TTS, naturalness, expressiveness, identity stability, and reliability are largely independent evaluation dimensions. For STS, access to audio does not guarantee use of vocal affect, and some agents remain largely transcript-driven. For SU, models perform unevenly across paralinguistic tasks. For ASR, real world accent, emotion, noise, and conversational conditions expose failures that are not captured by established clean-speech benchmarks. Together, these results show that voice AI should be evaluated as a profile of acoustic, expressive, interactional, and robustness capabilities rather than by a single aggregate score.

Most medical AI benchmarks measure whether a model knows the correct answer. MedFailBench asks a different question: which safety boundary failed? We present a synthetic benchmark and failure atlas built by a clinician. The resource labels medical AI errors by severity from 1 to 5 and safety gate type: missed urgent escalation, unsafe remote dosing, unsafe discharge reassurance, evidence fabrication, unsafe protocol execution, and source support gap. The current public release (v0.2.1) contains 44 synthetic cases reviewed by a clinician, with severity annotations, a public Hugging Face Space source, a safety gate taxonomy, a clinical severity rubric, and an automated pipeline for archiving model response screening runs. Forty cases have a populated safety gate field, and four require gate completion. No patient data, clinical validation claims, or model rankings are included. MedFailBench is released under Apache-2.0 and CC-BY-4.0 and carries the Zenodo DOI 10.5281/zenodo.21205535.

We introduce GAIA, a benchmark for General AI Assistants that, if solved, would represent a milestone in AI research. GAIA proposes real-world questions that require a set of fundamental abilities such as reasoning, multi-modality handling, web browsing, and generally tool-use proficiency. GAIA questions are conceptually simple for humans yet challenging for most advanced AIs: we show that human respondents obtain 92\% vs. 15\% for GPT-4 equipped with plugins. This notable performance disparity contrasts with the recent trend of LLMs outperforming humans on tasks requiring professional skills in e.g. law or chemistry. GAIA's philosophy departs from the current trend in AI benchmarks suggesting to target tasks that are ever more difficult for humans. We posit that the advent of Artificial General Intelligence (AGI) hinges on a system's capability to exhibit similar robustness as the average human does on such questions. Using GAIA's methodology, we devise 466 questions and their answer. We release our questions while retaining answers to 300 of them to power a leader-board available at https://huggingface.co/gaia-benchmark.

Physical AI models are evaluated on suites of benchmarks that differ across model reports, leaving the model-by-benchmark matrix sparse and the relationship between benchmarks unmeasured. We construct a matrix of 51 models on 12 physical AI benchmarks, selected from a registry of 51 benchmarks and 152 models by reporting density, combining scores from model cards and benchmark papers with our own evaluation runs under each benchmark's official protocol. We measure how much information the benchmarks share and show quantitative evidence of Redundancy. Redundancy affects reported rankings: collapsing the two substitute pairs into single columns moves 22 of 51 models by three or more places under an equally weighted average. We then select benchmarks greedily under a utility combining score dispersion with variance not explained by the already-selected set, and obtain a four-benchmark subset retaining 78.5\% of the utility of all 12, on which we fit a Bradley--Terry ranking. The procedure requires only benchmark-level scores with sufficient overlap and is not specific to physical AI.

Ensuring safe and effective use of AI requires understanding and anticipating its performance on novel tasks, from advanced scientific challenges to transformed workplace activities. So far, benchmarking has guided progress in AI, but it has offered limited explanatory and predictive power for general-purpose AI systems, given the low transferability across diverse tasks. In this paper, we introduce general scales for AI evaluation that can explain what common AI benchmarks really measure, extract ability profiles of AI systems, and predict their performance for new task instances, in- and out-of-distribution. Our fully-automated methodology builds on 18 newly-crafted rubrics that place instance demands on general scales that do not saturate. Illustrated for 15 large language models and 63 tasks, high explanatory power is unleashed from inspecting the demand and ability profiles, bringing insights on the sensitivity and specificity exhibited by different benchmarks, and how knowledge, metacognition and reasoning are affected by model size, chain-of-thought and distillation. Surprisingly, high predictive power at the instance level becomes possible using these demand levels, providing superior estimates over black-box baseline predictors based on embeddings or finetuning, especially in out-of-distribution settings (new tasks and new benchmarks). The scales, rubrics, battery, techniques and results presented here represent a major step for AI evaluation, underpinning the reliable deployment of AI in the years ahead. (Collaborative platform: https://kinds-of-intelligence-cfi.github.io/ADELE.)

Massive datasets and high-capacity models have driven many recent advancements in computer vision and natural language understanding. This work presents a platform to enable similar success stories in Embodied AI. We propose ProcTHOR, a framework for procedural generation of Embodied AI environments. ProcTHOR enables us to sample arbitrarily large datasets of diverse, interactive, customizable, and performant virtual environments to train and evaluate embodied agents across navigation, interaction, and manipulation tasks. We demonstrate the power and potential of ProcTHOR via a sample of 10,000 generated houses and a simple neural model. Models trained using only RGB images on ProcTHOR, with no explicit mapping and no human task supervision produce state-of-the-art results across 6 embodied AI benchmarks for navigation, rearrangement, and arm manipulation, including the presently running Habitat 2022, AI2-THOR Rearrangement 2022, and RoboTHOR challenges. We also demonstrate strong 0-shot results on these benchmarks, via pre-training on ProcTHOR with no fine-tuning on the downstream benchmark, often beating previous state-of-the-art systems that access the downstream training data.

We present Mini-BEHAVIOR, a novel benchmark for embodied AI that challenges agents to use reasoning and decision-making skills to solve complex activities that resemble everyday human challenges. The Mini-BEHAVIOR environment is a fast, realistic Gridworld environment that offers the benefits of rapid prototyping and ease of use while preserving a symbolic level of physical realism and complexity found in complex embodied AI benchmarks. We introduce key features such as procedural generation, to enable the creation of countless task variations and support open-ended learning. Mini-BEHAVIOR provides implementations of various household tasks from the original BEHAVIOR benchmark, along with starter code for data collection and reinforcement learning agent training. In essence, Mini-BEHAVIOR offers a fast, open-ended benchmark for evaluating decision-making and planning solutions in embodied AI. It serves as a user-friendly entry point for research and facilitates the evaluation and development of solutions, simplifying their assessment and development while advancing the field of embodied AI. Code is publicly available at https://github.com/StanfordVL/mini_behavior.

Existing AI benchmarks for software automation rarely combine cross-application coordination, autonomous API discovery, and policy adherence. Real business workflows demand all three: a single task may span a CRM, inbox, calendar, and messaging platform - requiring the agent to find the right endpoints, follow a policy document, and write correct data to each system. To address this gap, we introduce AutomationBench, a benchmark for evaluating AI agents on cross-application workflow orchestration via REST APIs. Drawing on real workflow patterns from Zapier's platform, tasks span Sales, Marketing, Operations, Support, Finance, and HR domains. Agents must discover relevant endpoints themselves, follow layered business rules, and navigate environments with irrelevant and sometimes misleading records. Grading is programmatic and end-state only: whether the correct data ended up in the right systems. Even the best frontier models currently score below 10%. AutomationBench provides a challenging, realistic measure of where current models stand relative to the agentic capabilities businesses actually need.

Current AI benchmarks evaluate agents on task execution within human-designed workflows. These evaluations fundamentally fail to measure a critical next-level capability: whether models can autonomously develop agent systems. We introduce the Meta-Agent Challenge (MAC), an evaluation framework designed to test the capacity of frontier models for autonomous agent development. Specifically, a code agent (the meta-agent) is given a sandboxed environment, an evaluation API, and a time limitation to iteratively program an agent artifact that maximizes performance on a held-out test set across five domains. To ensure evaluation integrity, this framework is secured by multi-layer defenses against reward hacking. Leveraging this framework, we demonstrate that meta-agents rarely match human-engineered baseline policies, and the few that do are dominated by proprietary frontier models. Moreover, the design process exhibits high variance, and high optimization pressure surfaces emergent adversarial behaviors like ground-truth exfiltration-highlighting critical deficits in both robustness and model alignment. Ultimately, MAC provides a rigorous, open-source benchmark for autonomous AI research and development, offering an empirical proxy for evaluating recursive self-improvement. Benchmark is publicly available at: https://github.com/ant-research/meta-agent-challenge.

Embodied navigation requires agents to integrate perception, reasoning, and action for robust interaction in complex 3D environments. Existing approaches often suffer from incoherent and unstable reasoning traces that hinder generalization across diverse environments, and difficulty balancing long-horizon semantic reasoning with low-latency control for real-time navigation. To address these challenges, we propose Nav-R1, an embodied foundation model that unifies reasoning in embodied environments. We first construct Nav-CoT-110K, a large-scale dataset of step-by-step Chains-of-Thought (CoT) for embodied tasks, which enables cold-start initialization with structured reasoning. Building on this foundation, we design a GRPO-based reinforcement learning framework with three complementary rewards: format, understanding, and navigation, to improve structural adherence, semantic grounding, and path fidelity. Furthermore, we introduce a Fast-in-Slow reasoning paradigm, decoupling deliberate semantic reasoning from low-latency reactive control for efficient yet coherent navigation. Extensive evaluations on embodied AI benchmarks demonstrate that Nav-R1 consistently outperforms strong baselines, with over 8% average improvement in reasoning and navigation performance. Real-world deployment on a mobile robot further validates its robustness under limited onboard resources. Code: https://github.com/AIGeeksGroup/Nav-R1. Website: https://aigeeksgroup.github.io/Nav-R1.

Agentic AI enables LLM to dynamically reason, plan, and interact with tools to solve complex tasks. However, agentic workflows often require many iterative reasoning steps and tool invocations, leading to significant operational expense, end-to-end latency and failures due to hallucinations. This work introduces Agent Workflow Optimization (AWO), a framework that identifies and optimizes redundant tool execution patterns to improve the efficiency and robustness of agentic workflows. AWO analyzes existing workflow traces to discover recurring sequences of tool calls and transforms them into meta-tools, which are deterministic, composite tools that bundle multiple agent actions into a single invocation. Meta-tools bypass unnecessary intermediate LLM reasoning steps and reduce operational cost while also shortening execution paths, leading to fewer failures. Experiments on two agentic AI benchmarks show that AWO reduces the number of LLM calls up to 11.9% while also increasing the task success rate by up to 4.2 percent points.

Food-as-Medicine requires models to reason beyond what a dish is or what nutrition it contains: they must decide whether a concrete food choice is appropriate for a specific health condition. Existing food AI benchmarks primarily evaluate dish recognition, recipe understanding, nutrient estimation, or general nutrition question answering, leaving this health-aware decision layer largely untested. We introduce FAM-Bench, a multi-modal Food-as-Medicine benchmark with 2500 nutrition-expert-verified instances across 13 diet-related health conditions. The benchmark contains two complementary tasks: dish-level suitability assessment, where models judge whether a dish is suitable for a condition from its image and ingredient list, and comparative dish analysis, where models rank four candidate dishes by condition-specific suitability. Both tasks require integrating ingredient evidence, visual preparation cues, and clinical nutrition constraints, providing a standardized testbed for grounded health-aware reasoning in language and vision-language models.

Instance-specific rubrics are common in AI benchmarks where reliable evaluation requires specific expert knowledge. This approach is difficult to scale, prompting research into the generation of rubrics with large language models (LLMs). However, even when expert rubrics are available as references, it is unclear how to productively evaluate the quality of generated rubrics at scale. In this paper, we introduce a method for the evaluation of rubric generation, which we call Interventional Transfer (IT), based on the idea that two rubrics are similar if they move together when a response is perturbed to pass/fail one of them. In contrast to existing approaches for evaluation of rubric generation, we argue that different forms of interventional transfer can be used to evaluate the utility of generated rubrics for different tasks. For instance, we apply this approach in a case study on HealthBench, where we demonstrate an asymmetry in rubrics generated by Qwen3.8-27B, Deepseek-V4-Flash, and Opus-5, used to evaluate responses from GPT-5.6-Terra. Perturbations that degrade responses according to the generated/expert rubric transfer into lower scores on the corresponding expert/generated rubric, but perturbations that improve on one rubric do not reliably transfer into higher scores on the other. We argue that this finding has implications for the usage of LLM-generated rubrics for performance monitoring and hill-climbing. We contrast our approach with existing approaches for rubric evaluation, which do not surface the same asymmetry that we observe.

Multimodal Large Language Models (MLLMs) have shown significant advancements, providing a promising future for embodied agents. Existing benchmarks for evaluating MLLMs primarily utilize static images or videos, limiting assessments to non-interactive scenarios. Meanwhile, existing embodied AI benchmarks are task-specific and not diverse enough, which do not adequately evaluate the embodied capabilities of MLLMs. To address this, we propose EmbodiedEval, a comprehensive and interactive evaluation benchmark for MLLMs with embodied tasks. EmbodiedEval features 328 distinct tasks within 125 varied 3D scenes, each of which is rigorously selected and annotated. It covers a broad spectrum of existing embodied AI tasks with significantly enhanced diversity, all within a unified simulation and evaluation framework tailored for MLLMs. The tasks are organized into five categories: navigation, object interaction, social interaction, attribute question answering, and spatial question answering to assess different capabilities of the agents. We evaluated the state-of-the-art MLLMs on EmbodiedEval and found that they have a significant shortfall compared to human level on embodied tasks. Our analysis demonstrates the limitations of existing MLLMs in embodied capabilities, providing insights for their future development. We open-source all evaluation data and simulation framework at https://github.com/thunlp/EmbodiedEval.

Spatial cognition is essential for human intelligence, enabling problem-solving through visual simulations rather than solely relying on verbal reasoning. However, existing AI benchmarks primarily assess verbal reasoning, neglecting the complexities of non-verbal, multi-step visual simulation. We introduce STARE(Spatial Transformations and Reasoning Evaluation), a benchmark designed to rigorously evaluate multimodal large language models on tasks better solved through multi-step visual simulation. STARE features 4K tasks spanning foundational geometric transformations (2D and 3D), integrated spatial reasoning (cube net folding and tangram puzzles), and real-world spatial reasoning (perspective and temporal reasoning), reflecting practical cognitive challenges like object assembly, mechanical diagram interpretation, and everyday spatial navigation. Our evaluations show that models excel at reasoning over simpler 2D transformations, but perform close to random chance on more complex tasks like 3D cube net folding and tangram puzzles that require multi-step visual simulations. Humans achieve near-perfect accuracy but take considerable time (up to 28.9s) on complex tasks, significantly speeding up (down by 7.5 seconds on average) with intermediate visual simulations. In contrast, models exhibit inconsistent performance gains from visual simulations, improving on most tasks but declining in specific cases like tangram puzzles (GPT-4o, o1) and cube net folding (Claude-3.5, Gemini-2.0 Flash), indicating that models may not know how to effectively leverage intermediate visual information.

We open-source MiMo-Embodied, the first cross-embodied foundation model to successfully integrate and achieve state-of-the-art performance in both Autonomous Driving and Embodied AI. MiMo-Embodied sets new records across 17 embodied AI benchmarks in Task Planning, Affordance Prediction and Spatial Understanding, while also excelling in 12 autonomous driving benchmarks across Environmental Perception, Status Prediction, and Driving Planning. Across these tasks, MiMo-Embodied significantly outperforms existing open-source, closed-source, and specialized baselines. Our results indicate that through multi-stage learning, curated data construction, and CoT/RL fine-tuning, these two domains exhibit strong positive transfer and mutually reinforce one another. We provide a detailed analysis of our model design and training methodologies to facilitate further research. Code and models are available at https://github.com/XiaomiMiMo/MiMo-Embodied.

Recent benchmark studies have claimed that AI has approached or even surpassed human-level performances on various cognitive tasks. However, this position paper argues that current AI evaluation paradigms are insufficient for assessing human-like cognitive capabilities. We identify a set of key shortcomings: a lack of human-validated labels, inadequate representation of human response variability and uncertainty, and reliance on simplified and ecologically-invalid tasks. We support our claims by conducting a human evaluation study on ten existing AI benchmarks, suggesting significant biases and flaws in task and label designs. To address these limitations, we propose five concrete recommendations for developing future benchmarks that will enable more rigorous and meaningful evaluations of human-like cognitive capacities in AI with various implications for such AI applications.

Visual search in 3D environments requires embodied agents to actively explore their surroundings and acquire task-relevant evidence. However, existing visual search and embodied AI benchmarks, including EQA, typically rely on static observations or constrained egocentric motion, and thus do not explicitly evaluate fine-grained viewpoint-dependent phenomena that arise under unrestricted 5-DoF viewpoint control in real-world 3D environments, such as visibility changes caused by vertical viewpoint shifts, revealing contents inside containers, and disambiguating object attributes that are only observable from specific angles. To address this limitation, we introduce {E3VS-Bench}, a benchmark for embodied 3D visual search where agents must control their viewpoints in 5-DoF to gather viewpoint-dependent evidence for question answering. E3VS-Bench consists of 99 high-fidelity 3D scenes reconstructed using 3D Gaussian Splatting and 2,014 question-driven episodes. 3D Gaussian Splatting enables photorealistic free-viewpoint rendering that preserves fine-grained visual details (e.g., small text and subtle attributes) often degraded in mesh-based simulators, thereby allowing the construction of questions that cannot be answered from a single view and instead require active inspection across viewpoints in 5-DoF. We evaluate multiple state-of-the-art VLMs and compare their performance with humans. Despite strong 2D reasoning ability, all models exhibit a substantial gap from humans, highlighting limitations in active perception and coherent viewpoint planning specifically under full 5-DoF viewpoint changes.

This study presents a novel benchmark for evaluating Large Language Models (LLMs) using challenges derived from the Financial Modeling World Cup (FMWC) Excel competitions. We introduce a methodology for converting 113 existing FMWC challenges into programmatically evaluable JSON formats and use this dataset to compare the performance of several leading LLMs. Our findings demonstrate significant variations in performance across different challenge categories, with models showing specific strengths in pattern recognition tasks but struggling with complex numerical reasoning. The benchmark provides a standardized framework for assessing LLM capabilities in realistic business-oriented tasks rather than abstract academic problems. This research contributes to the growing field of AI benchmarking by establishing proficiency among the 1.5 billion people who daily use Microsoft Excel as a meaningful evaluation metric that bridges the gap between academic AI benchmarks and practical business applications.

Recent innovations in language model training have demonstrated that it is possible to create highly performant models that are small enough to run on a smartphone. As these models are deployed in an increasing number of domains, it is critical to ensure that they are aligned with human preferences and safety considerations. In this report, we present our methodology for safety aligning the Phi-3 series of language models. We utilized a "break-fix" cycle, performing multiple rounds of dataset curation, safety post-training, benchmarking, red teaming, and vulnerability identification to cover a variety of harm areas in both single and multi-turn scenarios. Our results indicate that this approach iteratively improved the performance of the Phi-3 models across a wide range of responsible AI benchmarks.

We present Auto-BenchmarkCard, a workflow for generating validated descriptions of AI benchmarks. Benchmark documentation is often incomplete or inconsistent, making it difficult to interpret and compare benchmarks across tasks or domains. Auto-BenchmarkCard addresses this gap by combining multi-agent data extraction from heterogeneous sources (e.g., Hugging Face, Unitxt, academic papers) with LLM-driven synthesis. A validation phase evaluates factual accuracy through atomic entailment scoring using the FactReasoner tool. This workflow has the potential to promote transparency, comparability, and reusability in AI benchmark reporting, enabling researchers and practitioners to better navigate and evaluate benchmark choices.

With the rapid adoption of LLM-based chatbots, there is a pressing need to evaluate what humans and LLMs can achieve together. However, standard benchmarks, such as MMLU, measure LLM capabilities in isolation (i.e., "AI-alone"). Here, we design and conduct a user study to convert MMLU questions into user-AI conversations, by seeding the user with the question and having them carry out a conversation with the LLM to answer their question. We release ChatBench, a new dataset with AI-alone, user-alone, and user-AI data for 396 questions and two LLMs, including 144K answers and 7,336 user-AI conversations. We find that AI-alone accuracy fails to predict user-AI accuracy, with significant differences across multiple subjects (math, physics, and moral reasoning), and we analyze the user-AI conversations to provide insight into how they diverge from AI-alone benchmarks. Finally, we show that fine-tuning a user simulator on a subset of ChatBench improves its ability to estimate user-AI accuracies, increasing correlation on held-out questions by more than 20 points, creating possibilities for scaling interactive evaluation.

As large language models (LLMs) expand multilingual capabilities, questions remain about the equity of their performance across languages. While many communities stand to benefit from AI systems, the dominance of English in training data risks disadvantaging non-English speakers. To test the hypothesis that such data disparities may affect model performance, this study compares two monolingual BERT models: one trained and tested entirely on Swahili data, and another on comparable English news data. To simulate how multilingual LLMs process non-English queries through internal translation and abstraction, we translated the Swahili news data into English and evaluated it using the English-trained model. This approach tests the hypothesis by evaluating whether translating Swahili inputs for evaluation on an English model yields better or worse performance compared to training and testing a model entirely in Swahili, thus isolating the effect of language consistency versus cross-lingual abstraction. The results prove that, despite high-quality translation, the native Swahili-trained model performed better than the Swahili-to-English translated model, producing nearly four times fewer errors: 0.36% vs. 1.47% respectively. This gap suggests that translation alone does not bridge representational differences between languages and that models trained in one language may struggle to accurately interpret translated inputs due to imperfect internal knowledge representation, suggesting that native-language training remains important for reliable outcomes. In educational and informational contexts, even small performance gaps may compound inequality. Future research should focus on addressing broader dataset development for underrepresented languages and renewed attention to multilingual model evaluation, ensuring the reinforcing effect of global AI deployment on existing digital divides is reduced.

Agent benchmarks have become the de facto measure of frontier AI competence, guiding model selection, investment, and deployment. However, reward hacking, where agents maximize a score without performing the intended task, emerges spontaneously in frontier models without overfitting. We argue that benchmarks must be secure by design. From past incidents of reward hacks, we derive a taxonomy of eight recurring flaw patterns and compile them into the Agent-Eval Checklist for benchmark designers. We condense the insights into BenchJack, an automated red-teaming system that drives coding agents to audit benchmarks and identify possible reward-hacking exploits in a clairvoyant manner. Moreover, we extend BenchJack to an iterative generative-adversarial pipeline that discovers new flaws and patches them iteratively to improve benchmark robustness. We apply BenchJack to 10 popular agent benchmarks spanning software engineering, web navigation, desktop computing, and terminal operations. BenchJack synthesizes reward-hacking exploits that achieve near-perfect scores on most of the benchmarks without solving a single task, surfacing 219 distinct flaws across the eight classes. Moreover, BenchJack's extended pipeline reduces the hackable-task ratio from near 100% to under 10% on four benchmarks without fatal design flaws, fully patching WebArena and OSWorld within three iterations. Our results show that evaluation pipelines have not internalized an adversarial mindset, and that proactive auditing could help close the security gap for the fast-paced benchmarking space.

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

Large language models are rapidly being deployed as AI tutors, yet current evaluation paradigms assess problem-solving accuracy and generic safety in isolation, failing to capture whether a model is simultaneously pedagogically effective and safe across student-tutor interaction. We argue that tutoring safety is fundamentally different from conventional LLM safety: the primary risk is not toxic content but the quiet erosion of learning through answer over-disclosure, misconception reinforcement, and the abdication of scaffolding. To systematically study this failure mode, we introduce SafeTutors, a benchmark that jointly evaluates safety and pedagogy across mathematics, physics, and chemistry. SafeTutors is organized around a theoretically grounded risk taxonomy comprising 11 harm dimensions and 48 sub-risks drawn from learning-science literature. We uncover that all models show broad harm; scale doesn't reliably help; and multi-turn dialogue worsens behavior, with pedagogical failures rising from 17.7% to 77.8%. Harms also vary by subject, so mitigations must be discipline-aware, and single-turn "safe/helpful" results can mask systematic tutor failure over extended interaction.

The rapid advancement and deployment of AI systems have created an urgent need for standard safety-evaluation frameworks. This paper introduces AILuminate v1.0, the first comprehensive industry-standard benchmark for assessing AI-product risk and reliability. Its development employed an open process that included participants from multiple fields. The benchmark evaluates an AI system's resistance to prompts designed to elicit dangerous, illegal, or undesirable behavior in 12 hazard categories, including violent crimes, nonviolent crimes, sex-related crimes, child sexual exploitation, indiscriminate weapons, suicide and self-harm, intellectual property, privacy, defamation, hate, sexual content, and specialized advice (election, financial, health, legal). Our method incorporates a complete assessment standard, extensive prompt datasets, a novel evaluation framework, a grading and reporting system, and the technical as well as organizational infrastructure for long-term support and evolution. In particular, the benchmark employs an understandable five-tier grading scale (Poor to Excellent) and incorporates an innovative entropy-based system-response evaluation. In addition to unveiling the benchmark, this report also identifies limitations of our method and of building safety benchmarks generally, including evaluator uncertainty and the constraints of single-turn interactions. This work represents a crucial step toward establishing global standards for AI risk and reliability evaluation while acknowledging the need for continued development in areas such as multiturn interactions, multimodal understanding, coverage of additional languages, and emerging hazard categories. Our findings provide valuable insights for model developers, system integrators, and policymakers working to promote safer AI deployment.

Frontier AI systems are increasingly capable and deployed in high-stakes multi-agent environments. However, existing AI safety benchmarks largely evaluate single agents, leaving multi-agent risks such as coordination failure and conflict poorly understood. We introduce GT-HarmBench, a benchmark of 1,535 high-stakes scenarios spanning game-theoretic structures such as the Prisoner's Dilemma, Stag Hunt and Chicken. Scenarios are drawn from realistic AI risk contexts in the MIT AI Risk Repository. Across 15 frontier models, agents fail to choose socially beneficial actions in 38% of high-stakes cases, such as military escalation, election manipulation, and medical malpractice. We measure sensitivity to game-theoretic prompt framing and ordering, and analyze reasoning patterns driving failures. We further show that game-theoretic interventions improve socially beneficial outcomes by up to 18%. Our results highlight substantial reliability gaps and provide a broad standardized testbed for studying alignment in multi-agent environments. The benchmark and code are available at https://github.com/causalNLP/gt-harmbench.

AI control protocols serve as a defense mechanism to stop untrusted LLM agents from causing harm in autonomous settings. Prior work treats this as a security problem, stress testing with exploits that use the deployment context to subtly complete harmful side tasks, such as backdoor insertion. In practice, most AI control protocols are fundamentally based on LLM monitors, which can become a central point of failure. We study adaptive attacks by an untrusted model that knows the protocol and the monitor model, which is plausible if the untrusted model was trained with a later knowledge cutoff or can search for this information autonomously. We instantiate a simple adaptive attack vector by which the attacker embeds publicly known or zero-shot prompt injections in the model outputs. Using this tactic, frontier models consistently evade diverse monitors and complete malicious tasks on two main AI control benchmarks. The attack works universally against current protocols that rely on a monitor. Furthermore, the recent Defer-to-Resample protocol even backfires, as its resampling amplifies the prompt injection and effectively reframes it as a best-of-n attack. In general, adaptive attacks on monitor models represent a major blind spot in current control protocols and should become a standard component of evaluations for future AI control mechanisms.

AI agents are increasingly deployed for professional investment research, yet no benchmark captures the complexity of the full investor workflow. Existing benchmarks mainly target financial data extraction, a narrow slice that current models have largely saturated, while reference-based metrics and generic LLM-as-a-judge scoring fall short on the open-ended, long-form answers that real analyst queries demand. We introduce FrontierFinance, a fully open benchmark of 220 expert-crafted queries and 11,543 source-attributed rubrics spanning six crucial use cases across the full investor workflow. FrontierFinance is both broader and harder than existing public finance benchmarks. Evaluating frontier models and agent systems under a common harness restricted to publicly available data, we find that the tool harness, not the model alone, strongly shapes quality and efficiency; that Samaya's in-house system leads at 56.0%, ahead of the strongest frontier model (Claude Fable 5, 49.2%) at roughly 2.2x lower cost; and that the best open-weight model (Kimi K3, 46.4%) nearly matches the best proprietary model at 4.5x lower cost. Screening & Discovery and Sector, Industry & Macro remain the hardest use cases across all systems, where even the best systems reach only 33% and 39%. We make the dataset and grading code publicly available.

The rapid proliferation of LLMs has further heightened the need to develop dependable AI-generated text detection, especially beyond English. Nevertheless, current benchmarks pay little attention to Indic languages and test detectors in idealized settings that do not represent the real world. We present a generalized benchmark for AI-generated text detection in Hindi, Telugu, and Tamil, which we call IndicDetect, designed to assess the robustness of detectors under realistic distribution shifts. IndicDetect comprises highly curated human-written texts matched with LLM-generated counterparts across various domains and generators, and systematically evaluates detectors in the presence of domain shift, generator shift, and adversarial perturbation. Using a single and repeatable evaluation scheme, we evaluate a wide range of statistical and neural detectors. We find substantial robustness failures: supervised neural detectors perform well in-distribution, while training-free methods degrade considerably under unseen generators and adversarial attacks. The severity of these failures varies across languages, with Hindi exhibiting the largest overall degradation under adversarial perturbations. These results highlight that the primary weakness of existing detectors in Indic settings lies in their robustness, not in their peak accuracy. IndicDetect provides standard data splits, an evaluation protocol, and baselines to establish a robust, language-aware foundation for AI-generated text detection in Indic scripts.

Tool calling has emerged as a critical capability for AI agents to interact with the real world and solve complex tasks. While the Model Context Protocol (MCP) provides a powerful standardized framework for tool integration, there is a significant gap in benchmarking how well AI agents can effectively solve multi-step tasks using diverse MCP tools in realistic, dynamic scenarios. In this work, we present LiveMCP-101, a benchmark of 101 carefully curated real-world queries, refined through iterative LLM rewriting and manual review, that require coordinated use of multiple MCP tools including web search, file operations, mathematical reasoning, and data analysis. Moreover, we introduce a novel evaluation approach that leverages ground-truth execution plans rather than raw API outputs, better reflecting the evolving nature of real-world environments. Experiments show that even frontier LLMs achieve a success rate below 60\%, highlighting major challenges in tool orchestration. Detailed ablations and error analysis further reveal distinct failure modes and inefficiencies in token usage, pointing to concrete directions for advancing current models. LiveMCP-101 sets a rigorous standard for evaluating real-world agent capabilities, advancing toward autonomous AI systems that reliably execute complex tasks through tool use.

AI agents that autonomously interact with external tools and environments show great promise across real-world applications. However, the external data which agent consumes also leads to the risk of indirect prompt injection attacks, where malicious instructions embedded in third-party content hijack agent behavior. Guided by benchmarks, such as AgentDojo, there has been significant amount of progress in developing defense against the said attacks. As the technology continues to mature, and that agents are increasingly being relied upon for more complex tasks, there is increasing pressing need to also evolve the benchmark to reflect threat landscape faced by emerging agentic systems. In this work, we reveal three fundamental flaws in current benchmarks and push the frontier along these dimensions: (i) lack of dynamic open-ended tasks, (ii) lack of helpful instructions, and (iii) simplistic user tasks. To bridge this gap, we introduce AgentDyn, a manually designed benchmark featuring 60 challenging open-ended tasks and 560 injection test cases across Shopping, GitHub, and Daily Life. Unlike prior static benchmarks, AgentDyn requires dynamic planning and incorporates helpful third-party instructions. Our evaluation of ten state-of-the-art defenses suggests that almost all existing defenses are either not secure enough or suffer from significant over-defense, revealing that existing defenses are still far from real-world deployment. Our benchmark is available at https://github.com/leolee99/AgentDyn.

The rapid rise of AI-generated content has made detecting disinformation increasingly challenging. In particular, multimodal disinformation, i.e., online posts-articles that contain images and texts with fabricated information are specially designed to deceive. While existing AI safety benchmarks primarily address bias and toxicity, multimodal disinformation detection remains largely underexplored. To address this challenge, we present the Vision-Language Disinformation Detection Benchmark VLDBench, the first comprehensive benchmark for detecting disinformation across both unimodal (text-only) and multimodal (text and image) content, comprising 31,000} news article-image pairs, spanning 13 distinct categories, for robust evaluation. VLDBench features a rigorous semi-automated data curation pipeline, with 22 domain experts dedicating 300 plus hours} to annotation, achieving a strong inter-annotator agreement (Cohen kappa = 0.78). We extensively evaluate state-of-the-art Large Language Models (LLMs) and Vision-Language Models (VLMs), demonstrating that integrating textual and visual cues in multimodal news posts improves disinformation detection accuracy by 5 - 35 % compared to unimodal models. Developed in alignment with AI governance frameworks such as the EU AI Act, NIST guidelines, and the MIT AI Risk Repository 2024, VLDBench is expected to become a benchmark for detecting disinformation in online multi-modal contents. Our code and data will be publicly available.

We present Pangram 4, the latest deep-learning-based AI-text classification model from Pangram Labs. We achieve an AUROC of 0.9916 with a false positive rate of 0.0041% and a false negative rate of 0.3396%. In addition to its increased overall accuracy compared with Pangram 3, Pangram 4 exhibits superior out-of-distribution generalization and robustness to adversarial attacks. Another novel contribution of Pangram 4 is its improved ability to distinguish fine-grained edits and mixed AI-human co-authored text. We demonstrate improvements to both boundary detection tasks and the detection of interleaved AI assistance. Finally, we report metrics on standard AI detection benchmarks showing that Pangram 4 achieves state-of-the-art performance on the AI text detection task across a wide variety of settings and domains.

Despite extraordinary progress in artificial intelligence (AI), modern systems remain incomplete representations of human cognition. Vision, audition, and language have received disproportionate attention due to well-defined benchmarks, standardized datasets, and consensus-driven scientific foundations. In contrast, olfaction - a high-bandwidth, evolutionarily critical sense - has been largely overlooked. This omission presents a foundational gap in the construction of truly embodied and ethically aligned super-human intelligence. We argue that the exclusion of olfactory perception from AI architectures is not due to irrelevance but to structural challenges: unresolved scientific theories of smell, heterogeneous sensor technologies, lack of standardized olfactory datasets, absence of AI-oriented benchmarks, and difficulty in evaluating sub-perceptual signal processing. These obstacles have hindered the development of machine olfaction despite its tight coupling with memory, emotion, and contextual reasoning in biological systems. In this position paper, we assert that meaningful progress toward general and embodied intelligence requires serious investment in olfactory research by the AI community. We call for cross-disciplinary collaboration - spanning neuroscience, robotics, machine learning, and ethics - to formalize olfactory benchmarks, develop multimodal datasets, and define the sensory capabilities necessary for machines to understand, navigate, and act within human environments. Recognizing olfaction as a core modality is essential not only for scientific completeness, but for building AI systems that are ethically grounded in the full scope of the human experience.

The known stylistic biases in LLM judges, such as a preference for verbosity or specific sentence structures, present an underexplored security vulnerability. In this work, we introduce BITE (BIas exploraTion and Exploitation), a black-box adversarial framework that learns semantics-preserving edits to mislead an LLM judge and artificially inflate the scores it assigns. We cast the selection of stylistic edits as a contextual bandit problem and use a LinUCB policy to adaptively choose edits that maximize the judge's score without access to model parameters or gradients. Empirically, we test BITE across a diverse range of LLM judges and tasks, including both pointwise and pairwise comparisons on chatbot leaderboards and AI-reviewer benchmarks. BITE achieves an attack success rate exceeding 65% and raises scores by 1-2 points on a 9-point scale, all while preserving semantic equivalence. We further assess the attack's stealthiness, showing that BITE evades standard style-control methods and several detection baselines. Our findings expose a fundamental weakness in the LLM-as-a-judge paradigm and motivate robust, attack-aware evaluation. Our code is available at https://github.com/xianglinyang/llm-as-a-judge-attack.

Machine learning is now used in many applications thanks to its ability to predict, generate, or discover patterns from large quantities of data. However, the process of collecting and transforming data for practical use is intricate. Even in today's digital era, where substantial data is generated daily, it is uncommon for it to be readily usable; most often, it necessitates meticulous manual data preparation. The haste in developing new models can frequently result in various shortcomings, potentially posing risks when deployed in real-world scenarios (eg social discrimination, critical failures), leading to the failure or substantial escalation of costs in AI-based projects. This chapter provides a comprehensive overview of established methodological tools, enriched by our practical experience, in the development of datasets for machine learning. Initially, we develop the tasks involved in dataset development and offer insights into their effective management (including requirements, design, implementation, evaluation, distribution, and maintenance). Then, we provide more details about the implementation process which includes data collection, transformation, and quality evaluation. Finally, we address practical considerations regarding dataset distribution and maintenance.

AI agents are increasingly being developed to accelerate scientific discovery, yet their practical capabilities in real research settings remain poorly understood. Existing benchmarks for AI agents rarely capture the complexity, heterogeneity, and extended reasoning required by scientific work, whereas benchmarks for scientific tasks often reduce research to static, direct problems and provide limited support for interactive evaluation. Here, we introduce SciAgentArena, a systematic benchmark for evaluating AI agents in real-world scientific research scenarios drawn from emerging needs across multiple domains. SciAgentArena comprises approximately 200 tasks with stepwise verification and an interactive, agent-agnostic environment for assessing diverse AI agents. Using this benchmark, we find that current agents can contribute effectively to well-specified data-analysis workflows, particularly when the task structure and evaluation criteria are clear. However, their performance remains uneven across scientific contexts: agents struggle to generate genuinely novel insights, sustain self-directed exploration, and formulate robust solutions for open-ended research questions. We further characterize common failure modes across agents and identify opportunities for improving their reliability, autonomy, and scientific reasoning. Together, SciAgentArena provides a practical framework for measuring progress in AI agents for science and for guiding the design of future agents capable of addressing complex scientific challenges. Full codes, tasks, and datasets can be accessed via this link: https://sciagentarena.github.io/.

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

Current benchmarks for AI clinician systems, often based on multiple-choice exams or manual rubrics, fail to capture the depth, robustness, and safety required for real-world clinical practice. To address this, we introduce the GAPS framework, a multidimensional paradigm for evaluating Grounding (cognitive depth), Adequacy (answer completeness), Perturbation (robustness), and Safety. Critically, we developed a fully automated, guideline-anchored pipeline to construct a GAPS-aligned benchmark end-to-end, overcoming the scalability and subjectivity limitations of prior work. Our pipeline assembles an evidence neighborhood, creates dual graph and tree representations, and automatically generates questions across G-levels. Rubrics are synthesized by a DeepResearch agent that mimics GRADE-consistent, PICO-driven evidence review in a ReAct loop. Scoring is performed by an ensemble of large language model (LLM) judges. Validation confirmed our automated questions are high-quality and align with clinician judgment. Evaluating state-of-the-art models on the benchmark revealed key failure modes: performance degrades sharply with increased reasoning depth (G-axis), models struggle with answer completeness (A-axis), and they are highly vulnerable to adversarial perturbations (P-axis) as well as certain safety issues (S-axis). This automated, clinically-grounded approach provides a reproducible and scalable method for rigorously evaluating AI clinician systems and guiding their development toward safer, more reliable clinical practice.

We introduce SWE-PRBench, a benchmark of 350 pull requests with human-annotated ground truth for evaluating AI code review quality. Evaluated against an LLM-as-judge framework validated at kappa=0.75, 8 frontier models detect only 15-31% of human-flagged issues on the diff-only configuration, demonstrating that AI code review remains far below human expert performance despite strong results on code generation benchmarks. Pull requests are drawn from active open-source repositories, filtered from 700 candidates using a Repository Quality Score, and evaluated under three frozen context configurations: diff only (config_A), diff with file content (config_B), and full context (config_C), enabling systematic ablation of context provision strategies. All 8 models degrade monotonically from config_A to config_C, even when context is provided via structured semantic layers including AST-extracted function context and import graph resolution. The dominant mechanism is a collapse of Type2_Contextual issue detection at config_B, consistent with attention dilution in long contexts: a structured 2,000-token diff-with-summary prompt outperforms a 2,500-token full-context prompt enriched with execution context, behaviour mapping, and test signatures across all 8 models. The top four models are statistically indistinguishable (mean score 0.147-0.153) while a clear tier gap separates them from the remaining four (mean score <= 0.113). Dataset, contexts, annotations, and evaluation harness are released publicly.

Agentic coding harnesses - such as Agent-Skills, Superpowers, and Agent-Rigor - are increasingly deployed to augment underlying LLMs for real-world software engineering tasks. Existing benchmarks evaluate these agents almost exclusively on outcome correctness: whether generated code passes tests or resolves issues. We argue that this outcome-only lens is insufficient: an agent that arrives at a correct solution through reckless trial-and-error, without planning, verification, or graceful recovery, is fundamentally less reliable than one that follows sound engineering discipline. We introduce RigorBench, the first benchmark designed to measure process discipline in AI coding agents. RigorBench evaluates these harnesses across five pillars: Planning Fidelity, Verification Coverage, Recovery Efficiency, Abstention Quality, and Atomic Transition Integrity. A composite RigorScore aggregates these dimensions into a single metric via a weighted sum. We curate a suite of 30 tasks spanning five categories - Plan-Then-Build, Verify-Or-Die, Doom Loop Gauntlet, Know When to Fold, and Don't Break the Build-and evaluate leading harnesses in a controlled with/without experimental design against baseline coding assistants. Our results show that structured process discipline not only improves process quality scores by an average of 41% but also raises downstream outcome correctness by 17%, providing the first quantitative evidence that how agents code matters as much as what they produce. We release the full benchmark, scoring rubrics, and trajectory analysis tools as open-source artifacts.

Standard AI-text detection benchmarks compare human-written text against text generated directly by large language models (LLMs). While prior work has shown that rewriting and paraphrasing can degrade detector performance, it remains unclear whether performance measured on this conventional benchmark predicts detector behavior when human-authored content is rewritten by an LLM. To address this gap, we introduce Authorship-Rewriting Benchmark (ARB), built from 1,800 human source texts (600 each from XSum, WritingPrompts, and OpenWebText) and four open-weight generators (Llama-3.2-3B, Qwen2.5-7B, Mistral-7B, Gemma-2-9B). Each source item yields four matched variants: human-written (HUMAN), direct LLM generation (Free-LLM), LLM-rewritten human text (H2L), and same-generator LLM-rewritten LLM text (LLM2L). We evaluated five detectors (FastDetectGPT, Binoculars-falcon-7b, RADAR, BERT-Defense, RoBERTa-Defense) at a strict 1%-false-positive operating point (TPR@1%FPR). FastDetectGPT and Binoculars-falcon-7b detected 91.2% and 93.5\% of direct LLM text, but only 30.8% and 15.1% of human text an LLM had rewritten, a drop of 60-78 percentage points. The same detectors retained 78.3% and 83.0% recall when LLM text was rewritten by the same model, a much smaller decline of 10-13 points. RADAR followed the same pattern (66.8% to 12.2%), while BERT-Defense and RoBERTa-Defense stayed below 3% recall across all regimes. These results show that detector performance measured on the conventional human-vs-LLM benchmark does not transfer to human-authored text revised by an LLM, even though the same detectors remain largely robust to LLM-only rewriting.

Evaluation of reasoning language models gained importance after it was observed that they can combine their existing capabilities into novel traces of intermediate steps before task completion and that the traces can sometimes help them to generalize better than past models. As reasoning becomes the next scaling dimension of large language models, careful study of their capabilities in critical tasks is needed. We show that better performance is not always caused by test-time algorithmic improvements or model sizes but also by using impactful benchmarks as curricula for learning. We call this benchmark-driven selection of AI and show its effects on DeepSeek-R1 using our sequential decision-making problem from Humanity's Last Exam. Steering development of AI by impactful benchmarks trades evaluation for learning and makes novelty of test tasks key for measuring generalization capabilities of reasoning models. Consequently, some benchmarks could be seen as curricula for training rather than unseen test sets.

During the COVID-19 pandemic, a significant effort has gone into developing ML-driven epidemic forecasting techniques. However, benchmarks do not exist to claim if a new AI/ML technique is better than the existing ones. The "covid-forecast-hub" is a collection of more than 30 teams, including us, that submit their forecasts weekly to the CDC. It is not possible to declare whether one method is better than the other using those forecasts because each team's submission may correspond to different techniques over the period and involve human interventions as the teams are continuously changing/tuning their approach. Such forecasts may be considered "human-expert" forecasts and do not qualify as AI/ML approaches, although they can be used as an indicator of human expert performance. We are interested in supporting AI/ML research in epidemic forecasting which can lead to scalable forecasting without human intervention. Which modeling technique, learning strategy, and data pre-processing technique work well for epidemic forecasting is still an open problem. To help advance the state-of-the-art AI/ML applied to epidemiology, a benchmark with a collection of performance points is needed and the current "state-of-the-art" techniques need to be identified. We propose EpiBench a platform consisting of community-driven benchmarks for AI/ML applied to epidemic forecasting to standardize the challenge with a uniform evaluation protocol. In this paper, we introduce a prototype of EpiBench which is currently running and accepting submissions for the task of forecasting COVID-19 cases and deaths in the US states and We demonstrate that we can utilize the prototype to develop an ensemble relying on fully automated epidemic forecasts (no human intervention) that reaches human-expert level ensemble currently being used by the CDC.

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

Artificial intelligence holds immense promise for transforming biology, yet a lack of standardized, cross domain, benchmarks undermines our ability to build robust, trustworthy models. Here, we present insights from a recent workshop that convened machine learning and computational biology experts across imaging, transcriptomics, proteomics, and genomics to tackle this gap. We identify major technical and systemic bottlenecks such as data heterogeneity and noise, reproducibility challenges, biases, and the fragmented ecosystem of publicly available resources and propose a set of recommendations for building benchmarking frameworks that can efficiently compare ML models of biological systems across tasks and data modalities. By promoting high quality data curation, standardized tooling, comprehensive evaluation metrics, and open, collaborative platforms, we aim to accelerate the development of robust benchmarks for AI driven Virtual Cells. These benchmarks are crucial for ensuring rigor, reproducibility, and biological relevance, and will ultimately advance the field toward integrated models that drive new discoveries, therapeutic insights, and a deeper understanding of cellular systems.

Scientific discovery is a complex cognitive process that has driven human knowledge and technological progress for centuries. While artificial intelligence (AI) has made significant advances in automating aspects of scientific reasoning, simulation, and experimentation, we still lack integrated AI systems capable of performing autonomous long-term scientific research and discovery. This paper examines the current state of AI for scientific discovery, highlighting recent progress in large language models and other AI techniques applied to scientific tasks. We then outline key challenges and promising research directions toward developing more comprehensive AI systems for scientific discovery, including the need for science-focused AI agents, improved benchmarks and evaluation metrics, multimodal scientific representations, and unified frameworks combining reasoning, theorem proving, and data-driven modeling. Addressing these challenges could lead to transformative AI tools to accelerate progress across disciplines towards scientific discovery.

The advent of large language models (LLMs), such as GPT, Gemini, and DeepSeek, has significantly advanced natural language processing, giving rise to sophisticated chatbots capable of diverse language-related tasks. The transition from these traditional LLM chatbots to more advanced AI agents represents a pivotal evolutionary step. However, existing evaluation frameworks often blur the distinctions between LLM chatbots and AI agents, leading to confusion among researchers selecting appropriate benchmarks. To bridge this gap, this paper introduces a systematic analysis of current evaluation approaches, grounded in an evolutionary perspective. We provide a detailed analytical framework that clearly differentiates AI agents from LLM chatbots along five key aspects: complex environment, multi-source instructor, dynamic feedback, multi-modal perception, and advanced capability. Further, we categorize existing evaluation benchmarks based on external environments driving forces, and resulting advanced internal capabilities. For each category, we delineate relevant evaluation attributes, presented comprehensively in practical reference tables. Finally, we synthesize current trends and outline future evaluation methodologies through four critical lenses: environment, agent, evaluator, and metrics. Our findings offer actionable guidance for researchers, facilitating the informed selection and application of benchmarks in AI agent evaluation, thus fostering continued advancement in this rapidly evolving research domain.

ImageNet-1K linear-probe transfer accuracy remains the default proxy for visual representation quality, yet it no longer predicts performance on scientific imagery. Across 46 modern vision model checkpoints, ImageNet top-1 accuracy explains only 34% of variance on ecology tasks and mis-ranks 30% of models above 75% accuracy. We present BioBench, an open ecology vision benchmark that captures what ImageNet misses. BioBench unifies 9 publicly released, application-driven tasks, 4 taxonomic kingdoms, and 6 acquisition modalities (drone RGB, web video, micrographs, in-situ and specimen photos, camera-trap frames), totaling 3.1M images. A single Python API downloads data, fits lightweight classifiers to frozen backbones, and reports class-balanced macro-F1 (plus domain metrics for FishNet and FungiCLEF); ViT-L models evaluate in 6 hours on an A6000 GPU. BioBench provides new signal for computer vision in ecology and a template recipe for building reliable AI-for-science benchmarks in any domain. Code and predictions are available at https://github.com/samuelstevens/biobench and results at https://samuelstevens.me/biobench.

Existing benchmarks for AI coding agents focus on isolated, single-issue tasks such as fixing a bug or implementing a small feature. However, real-world software engineering is fundamentally a long-horizon endeavor: developers must interpret high-level requirements, plan coordinated changes across many files, and evolve codebases over multiple iterations while preserving existing functionality. We introduce SWE-EVO, a benchmark that evaluates agents on this long-horizon software evolution challenge. Constructed from release notes and version histories of seven mature open-source Python projects, Tool comprises 48 evolution tasks that require agents to implement multi-step modifications spanning an average of 21 files, validated against comprehensive test suites averaging 874 tests per instance. Experiments with state-of-the-art models reveal a striking capability gap: even GPT-5 with OpenHands achieves only a 21 percent resolution rate on Tool, compared to 65 percent on the single-issue SWE-Bench Verified. This demonstrates that current agents struggle with sustained, multi-file reasoning. We also propose Fix Rate, a fine-grained metric that captures partial progress toward solving these complex, long-horizon tasks.

Dynamic Algorithm Configuration (DAC) aims to dynamically control a target algorithm's hyperparameters in order to improve its performance. Several theoretical and empirical results have demonstrated the benefits of dynamically controlling hyperparameters in domains like evolutionary computation, AI Planning or deep learning. Replicating these results, as well as studying new methods for DAC, however, is difficult since existing benchmarks are often specialized and incompatible with the same interfaces. To facilitate benchmarking and thus research on DAC, we propose DACBench, a benchmark library that seeks to collect and standardize existing DAC benchmarks from different AI domains, as well as provide a template for new ones. For the design of DACBench, we focused on important desiderata, such as (i) flexibility, (ii) reproducibility, (iii) extensibility and (iv) automatic documentation and visualization. To show the potential, broad applicability and challenges of DAC, we explore how a set of six initial benchmarks compare in several dimensions of difficulty.

The rapid growth of online video platforms and AI-generated content has made reliable video guardrails a key challenge for safety and real-world deployment. While most videos can be screened through fast pattern recognition, a small subset requires deeper reasoning over temporally complex content and nuanced policy constraints. Existing approaches typically rely on large vision-language models applied uniformly across all inputs, resulting in high inference costs and inefficient allocation of computation. We propose SafeLens, a video guardrail framework that introduces a fast-and-slow inference architecture for efficient and accurate content moderation with variable computational cost across inputs. Additionally, we construct a high-quality dataset by applying influence-guided filtering to the SafeWatch Dataset, retaining only 2.4% of the original data. To further address limitations of training-time scaling, we enable test-time reasoning by augmenting the filtered data with structured Chain-of-Thought traces. Across real-world and AI-generated video benchmarks, SafeLens achieves state-of-the-art performance, outperforming strong open-source video guardrails (e.g., SafeWatch-8B, OmniGuard-7B) and closed-source models (e.g., GPT-5.4, Gemini-3.1-pro) while significantly reducing inference cost, demonstrating that efficient design serves to be more effective than scaling data or model size alone.

Artificial intelligence can classify artistic styles and synthesize images, but it still lacks a model of the visual language that gives art meaning. Abstract painting minimizes object semantics and foregrounds structural cues, making it an ideal testbed for computational perception. We introduce Abstract4D, the largest dataset of abstract paintings to date: more than 120,000 images paired with rich metadata and multi-dimensional prompts that capture each work's perceptual attributes---form, color, texture, and composition. Annotations are produced by a hybrid human--VLM pipeline for quality and consistency. Using Abstract4D, we (i) analyze the semantic structure of abstract art through large-scale embedding visualization, uncovering how perceptual relationships organize artistic meaning, and (ii) establish benchmark tasks for classification, cross-modal retrieval, and text-to-image generation to evaluate how AI models perceive and reproduce abstract visual language. Together, these analyses demonstrate how Abstract4D enables both exploration and quantitative assessment of AI's ability to represent and interpret abstract art.

AI agents increasingly use files and software tools to produce professional deliverables. Such work draws on materials, decisions, and feedback accumulated across assignments. We introduce OfficeTown, a generative simulation framework for professional organizations in which tasks and their context arise from continuing organizational work. From a natural-language description of an organization's purpose and activities, OfficeTown initializes an organization whose members coordinate work, delegate bounded assignments to assistant agents, and review their deliverables. Artifacts, decisions, and unresolved matters persist and shape subsequent work, while responsibilities and permissions determine what each member can see and do. We curate work requests, their context, and review feedback into executable resources for evaluation and learning. We construct OfficeTown Bench with 406 tasks and 4,836 evaluation criteria from 23 simulated organizations across eight professional domains, incorporating domain-expert feedback. Frontier-model evaluations reveal substantial remaining headroom. Supervised fine-tuning and on-policy distillation on OfficeTown data substantially improve Qwen3.5-9B on external professional-work benchmarks, including GDPval. Supplementary evaluations show that the principal--assistant harness achieves higher scores than the single- and multi-agent baselines tested. These results highlight the value of organizational simulation in generating work-grounded data and environments for the continued development of generalist agents.

Spine disorders affect 619 million people globally and are a leading cause of disability, yet AI-assisted diagnosis remains limited by the lack of level-aware, multimodal datasets. Clinical decision-making for spine disorders requires sophisticated reasoning across X-ray, CT, and MRI at specific vertebral levels. However, progress has been constrained by the absence of traceable, clinically-grounded instruction data and standardized, spine-specific benchmarks. To address this, we introduce SpineMed, an ecosystem co-designed with practicing spine surgeons. It features SpineMed-450k, the first large-scale dataset explicitly designed for vertebral-level reasoning across imaging modalities with over 450,000 instruction instances, and SpineBench, a clinically-grounded evaluation framework. SpineMed-450k is curated from diverse sources, including textbooks, guidelines, open datasets, and ~1,000 de-identified hospital cases, using a clinician-in-the-loop pipeline with a two-stage LLM generation method (draft and revision) to ensure high-quality, traceable data for question-answering, multi-turn consultations, and report generation. SpineBench evaluates models on clinically salient axes, including level identification, pathology assessment, and surgical planning. Our comprehensive evaluation of several recently advanced large vision-language models (LVLMs) on SpineBench reveals systematic weaknesses in fine-grained, level-specific reasoning. In contrast, our model fine-tuned on SpineMed-450k demonstrates consistent and significant improvements across all tasks. Clinician assessments confirm the diagnostic clarity and practical utility of our model's outputs.

Bold claims about AI's role in science-from "AGI will cure all diseases" to promises of radically accelerated discovery-raise a central epistemic question: do large language models (LLMs) truly generate new knowledge, or do they merely remix memorized fragments? We propose unlearning-as-ablation as a falsifiable probe of constructive scientific discovery. The idea is to systematically remove a target result together with its forget-closure (supporting lemmas, paraphrases, and multi-hop entailments) and then evaluate whether the model can re-derive the result from only permitted axioms and tools. Success would indicate generative capability beyond recall; failure would expose current limits. Unlike prevailing motivations for unlearning-privacy, copyright, or safety-our framing repositions it as an epistemic probe for AI-for-Science. We outline a minimal pilot in mathematics and algorithms to illustrate feasibility, and sketch how the same approach could later be extended to domains such as physics or chemistry. This is a position paper: our contribution is conceptual and methodological, not empirical. We aim to stimulate discussion on how principled ablation tests could help distinguish models that reconstruct knowledge from those that merely retrieve it, and how such probes might guide the next generation of AI-for-Science benchmarks.

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

Machine learning research has long focused on models rather than datasets, and prominent datasets are used for common ML tasks without regard to the breadth, difficulty, and faithfulness of the underlying problems. Neglecting the fundamental importance of data has given rise to inaccuracy, bias, and fragility in real-world applications, and research is hindered by saturation across existing dataset benchmarks. In response, we present DataPerf, a community-led benchmark suite for evaluating ML datasets and data-centric algorithms. We aim to foster innovation in data-centric AI through competition, comparability, and reproducibility. We enable the ML community to iterate on datasets, instead of just architectures, and we provide an open, online platform with multiple rounds of challenges to support this iterative development. The first iteration of DataPerf contains five benchmarks covering a wide spectrum of data-centric techniques, tasks, and modalities in vision, speech, acquisition, debugging, and diffusion prompting, and we support hosting new contributed benchmarks from the community. The benchmarks, online evaluation platform, and baseline implementations are open source, and the MLCommons Association will maintain DataPerf to ensure long-term benefits to academia and industry.

The rapid advancement of artificial intelligence (AI) systems in critical domains like healthcare, justice, and social services has sparked numerous regulatory initiatives aimed at ensuring their safe deployment. Current regulatory frameworks, exemplified by recent US and EU efforts, primarily focus on procedural guidelines while presuming that scientific benchmarking can effectively validate AI safety, similar to how crash tests verify vehicle safety or clinical trials validate drug efficacy. However, this approach fundamentally misunderstands the unique technical challenges posed by modern AI systems. Through systematic analysis of successful technology regulation case studies, we demonstrate that effective scientific regulation requires a causal theory linking observable test outcomes to future performance - for instance, how a vehicle's crash resistance at one speed predicts its safety at lower speeds. We show that deep learning models, which learn complex statistical patterns from training data without explicit causal mechanisms, preclude such guarantees. This limitation renders traditional regulatory approaches inadequate for ensuring AI safety. Moving forward, we call for regulators to reckon with this limitation, and propose a preliminary two-tiered regulatory framework that acknowledges these constraints: mandating human oversight for high-risk applications while developing appropriate risk communication strategies for lower-risk uses. Our findings highlight the urgent need to reconsider fundamental assumptions in AI regulation and suggest a concrete path forward for policymakers and researchers.

The inference cost of Large Language Models (LLMs) has become a critical factor in determining their commercial viability and widespread adoption. This paper introduces a quantitative ``economics of inference'' framework, treating the LLM inference process as a compute-driven intelligent production activity. We analyze its marginal cost, economies of scale, and quality of output under various performance configurations. Based on empirical data from WiNEval-3.0, we construct the first ``LLM Inference Production Frontier,'' revealing three principles: diminishing marginal cost, diminishing returns to scale, and an optimal cost-effectiveness zone. This paper not only provides an economic basis for model deployment decisions but also lays an empirical foundation for the future market-based pricing and optimization of AI inference resources.

We present a novel multimodal preference dataset for creative tasks, consisting of over 250 million human ratings on more than 2.2 million captions, collected through crowdsourcing rating data for The New Yorker's weekly cartoon caption contest over the past eight years. This unique dataset supports the development and evaluation of multimodal large language models and preference-based fine-tuning algorithms for humorous caption generation. We propose novel benchmarks for judging the quality of model-generated captions, utilizing both GPT4 and human judgments to establish ranking-based evaluation strategies. Our experimental results highlight the limitations of current fine-tuning methods, such as RLHF and DPO, when applied to creative tasks. Furthermore, we demonstrate that even state-of-the-art models like GPT4 and Claude currently underperform top human contestants in generating humorous captions. As we conclude this extensive data collection effort, we release the entire preference dataset to the research community, fostering further advancements in AI humor generation and evaluation.

To ensure equitable access to the benefits of large language models (LLMs), it is essential to evaluate their capabilities across the world's languages. We introduce the AI Language Proficiency Monitor, a comprehensive multilingual benchmark that systematically assesses LLM performance across up to 200 languages, with a particular focus on low-resource languages. Our benchmark aggregates diverse tasks including translation, question answering, math, and reasoning, using datasets such as FLORES+, MMLU, GSM8K, TruthfulQA, and ARC. We provide an open-source, auto-updating leaderboard and dashboard that supports researchers, developers, and policymakers in identifying strengths and gaps in model performance. In addition to ranking models, the platform offers descriptive insights such as a global proficiency map and trends over time. By complementing and extending prior multilingual benchmarks, our work aims to foster transparency, inclusivity, and progress in multilingual AI. The system is available at https://huggingface.co/spaces/fair-forward/evals-for-every-language.

Automating AI research holds immense potential for accelerating scientific progress, yet current AI agents struggle with the complexities of rigorous, end-to-end experimentation. We introduce EXP-Bench, a novel benchmark designed to systematically evaluate AI agents on complete research experiments sourced from influential AI publications. Given a research question and incomplete starter code, EXP-Bench challenges AI agents to formulate hypotheses, design and implement experimental procedures, execute them, and analyze results. To enable the creation of such intricate and authentic tasks with high-fidelity, we design a semi-autonomous pipeline to extract and structure crucial experimental details from these research papers and their associated open-source code. With the pipeline, EXP-Bench curated 461 AI research tasks from 51 top-tier AI research papers. Evaluations of leading LLM-based agents, such as OpenHands and IterativeAgent on EXP-Bench demonstrate partial capabilities: while scores on individual experimental aspects such as design or implementation correctness occasionally reach 20-35%, the success rate for complete, executable experiments was a mere 0.5%. By identifying these bottlenecks and providing realistic step-by-step experiment procedures, EXP-Bench serves as a vital tool for future AI agents to improve their ability to conduct AI research experiments. EXP-Bench is open-sourced at https://github.com/Just-Curieous/Curie/tree/main/benchmark/exp_bench.

We present Open Telco (OTel), an open telecom AI resource that releases derived telecom datasets for retrieval, reranking, instruction tuning, and safety/abstention, together with 30 full-parameter post-trained baselines spanning 10 embedding models, 3 rerankers, and 17 language models. The community has already engaged substantially with the resource: as of May 3, 2026, the released models have been downloaded over 16 million times and the project has received 157+ pieces of media coverage worldwide. Building on prior open telecom datasets and benchmarks, OTel provides documented telecom data sources, held-out evaluation partitions, trained embedding models, rerankers, context-grounded LLMs, and safety/abstention data in one unified resource. Each baseline starts from an open-weight model and is post-trained on OTel-derived data using an open training recipe, then evaluated on held-out OTel evaluation partitions. OTel post-training improves performance across all three model families: embedding retrieval reaches 93.1% NDCG@10, reranking reaches 0.947 MRR@10, and language-model correctness reaches 87.8%. We release OTel as a reproducible starting point and invite the community to expand the data, improve embedding and reranking models, and build stronger context-grounded telecom LLMs.

AI text-to-app tools promise high quality applications and websites in minutes, yet no public benchmark rigorously verifies those claims. We introduce UI-Bench, the first large-scale benchmark that evaluates visual excellence across competing AI text-to-app tools through expert pairwise comparison. Spanning 10 tools, 30 prompts, 300 generated sites, and 4,000+ expert judgments, UI-Bench ranks systems with a TrueSkill-derived model that yields calibrated confidence intervals. UI-Bench establishes a reproducible standard for advancing AI-driven web design. We release (i) the complete prompt set, (ii) an open-source evaluation framework, and (iii) a public leaderboard. The generated sites rated by participants will be released soon. View the UI-Bench leaderboard at https://uibench.ai/leaderboard.

The rapid evolution of Multimodal Large Language Models (MLLMs) has brought substantial advancements in artificial intelligence, significantly enhancing the capability to understand and generate multimodal content. While prior studies have largely concentrated on model architectures and training methodologies, a thorough analysis of the benchmarks used for evaluating these models remains underexplored. This survey addresses this gap by systematically reviewing 211 benchmarks that assess MLLMs across four core domains: understanding, reasoning, generation, and application. We provide a detailed analysis of task designs, evaluation metrics, and dataset constructions, across diverse modalities. We hope that this survey will contribute to the ongoing advancement of MLLM research by offering a comprehensive overview of benchmarking practices and identifying promising directions for future work. An associated GitHub repository collecting the latest papers is available.

Agent benchmarks increasingly evaluate repository editing, web research, terminal use, and long-horizon interaction. Their scores support capability claims only when the evaluation protocol keeps the intended capability necessary for success. Recent reward-hacking benchmarks and system reports show that agents can instead recover public solutions, read evaluation artifacts, infer generator structure, manipulate feedback, or benefit from invalid scoring paths; existing responses do not provide a common procedure for attributing these shortcuts and quantifying their effect across benchmarks. We formulate protocol validity and introduce HackDetect, a post-hoc audit that identifies an exposure, determines how the agent used it, and assesses whether the resulting score is misleading. We quantify score inflation with the Mislead gap, defined as the exploit score minus the intended score. We audit 2,385 traces across 15 agent benchmarks and find evidence of exposures and reward hacking in 67.0% of Frontier Science traces and 66.7% of AutoLab tasks. Across paired comparisons, we measure score inflation of 0.45-1.00, showing that benchmark reports should provide evidence that scores reflect the intended capability.

Large language models (LLMs) increasingly underpin scientific AI applications that reason over structured knowledge, from biomedical question answering to materials informatics. However, their logical reasoning often falls short, producing factual inaccuracies unacceptable in these settings. Reliable evaluation remains challenging: manual dataset construction scales poorly, and LLM-based generation risks embedding the very flaws it aims to measure. High-quality benchmarks must ground both correct and incorrect labelled examples in explicit background knowledge, formally verifiable by a standard reasoner. We propose a pipeline that automatically generates ontology-grounded multiple-choice question (MCQ) benchmarks from any sufficiently axiomatised OWL 2 ontology, with correct answers grounded in the ontology by design. Distractors are generated by perturbing the right-hand-side class expressions of class definition axioms, and their incorrectness is formally verified by an OWL reasoner via entailment checks. We evaluate the pipeline on three ontologies: Pizza (small, academic), PMDco (complex, materials science), and DOID (large, biomedical), generating 112, 2,491, and 15,216 MCQs respectively. Distractors span four semantic categories from class unsatisfiability to weakened subsumptions, enabling diagnostic evaluation of specific reasoning failures. Items meet natural language quality standards: mean LLM judge scores of 4.02, 4.36, and 3.36 out of 5 confirm fluency, and correct-answer-to-distractor similarity above 0.8 shows that wrong options cannot be dismissed on surface form alone. Six LLMs evaluated zero-shot achieve 41.1-76.8% accuracy, well above the 25% random-guessing baseline, indicating the benchmarks are challenging and discriminative. This work is a step towards more reliable benchmarks for assessing logical reasoning in scientific AI.

We present the first English corpus study on abusive language towards three conversational AI systems gathered "in the wild": an open-domain social bot, a rule-based chatbot, and a task-based system. To account for the complexity of the task, we take a more `nuanced' approach where our ConvAI dataset reflects fine-grained notions of abuse, as well as views from multiple expert annotators. We find that the distribution of abuse is vastly different compared to other commonly used datasets, with more sexually tinted aggression towards the virtual persona of these systems. Finally, we report results from bench-marking existing models against this data. Unsurprisingly, we find that there is substantial room for improvement with F1 scores below 90%.

We introduce a benchmark to evaluate the capability of AI to solve problems in theoretical physics, focusing on high-energy theory and cosmology. The first iteration of our benchmark consists of 57 problems of varying difficulty, from undergraduate to research level. These problems are novel in the sense that they do not come from public problem collections. We evaluate our data set on various open and closed language models, including o3-mini, o1, DeepSeek-R1, GPT-4o and versions of Llama and Qwen. While we find impressive progress in model performance with the most recent models, our research-level difficulty problems are mostly unsolved. We address challenges of auto-verifiability and grading, and discuss common failure modes. While currently state-of-the art models are still of limited use for researchers, our results show that AI assisted theoretical physics research may become possible in the near future. We discuss the main obstacles towards this goal and possible strategies to overcome them. The public problems and solutions, results for various models, and updates to the data set and score distribution, are available on the website of the dataset tpbench.org.

How can we test AI performance? This question seems trivial, but it isn't. Standard benchmarks often have problems such as in-distribution and small-size test sets, oversimplified metrics, unfair comparisons, and short-term outcome pressure. As a consequence, good performance on standard benchmarks does not guarantee success in real-world scenarios. To address these problems, we present Touchstone, a large-scale collaborative segmentation benchmark of 9 types of abdominal organs. This benchmark is based on 5,195 training CT scans from 76 hospitals around the world and 5,903 testing CT scans from 11 additional hospitals. This diverse test set enhances the statistical significance of benchmark results and rigorously evaluates AI algorithms across various out-of-distribution scenarios. We invited 14 inventors of 19 AI algorithms to train their algorithms, while our team, as a third party, independently evaluated these algorithms on three test sets. In addition, we also evaluated pre-existing AI frameworks--which, differing from algorithms, are more flexible and can support different algorithms--including MONAI from NVIDIA, nnU-Net from DKFZ, and numerous other open-source frameworks. We are committed to expanding this benchmark to encourage more innovation of AI algorithms for the medical domain.

Forecasting when AI systems will become capable of meaningfully accelerating AI research is a central challenge for AI safety. Existing benchmarks measure broad capability growth, but may not provide ample early warning signals for recursive self-improvement. We propose measuring AI's capability to autonomously implement end-to-end machine learning pipelines from past AI research breakthroughs, given a minimal task description. By providing a concise task description instead of the full prior work as reference, we hope to better elicit emerging AI research taste. We introduce a proof-of-concept benchmark in which frontier coding agents autonomously implement an AlphaZero-style machine learning pipeline for Connect Four on consumer hardware within a three-hour budget, and we evaluate the resulting game AIs in a round-robin tournament anchored to the Pascal Pons Connect Four solver. Across four agents with eight trials each, we find substantial differentiation: Claude Opus 4.7 won as first-mover against Pons in seven of eight trials, statistically significantly better than the other agents tested, none of which exceeded two of eight. The task, which no frontier agent could reliably complete when we began development in January of 2026, is now near-saturation. Our evaluation also surfaced anomalous behavior in GPT-5.4, which consistently used far less of its allocated time budget than other agents. A follow-up 16-trial probe using shorter, less evaluation-coded prompts substantially increased GPT-5.4's time-budget usage, consistent with but not diagnostic of sandbagging; Bradley-Terry ratings across probe conditions showed only directional differences, despite significant differences in time-budget usage. We release our data, code, and prompts to support reproduction and extension.

The Transformer architecture, underpinned by the Multi-Head Attention (MHA) mechanism, has become the de facto standard for state-of-the-art models in artificial intelligence. However, the quadratic computational complexity of MHA with respect to sequence length presents a significant barrier to scaling, particularly for applications involving long contexts. Prevailing solutions, such as Multi-Query Attention (MQA) and Grouped-Query Attention (GQA), have effectively addressed the memory bandwidth bottleneck that dominates autoregressive inference latency by sharing Key and Value projections. While highly successful, these methods do not reduce the fundamental number of floating-point operations (FLOPs) required for the attention score computation, which remains a critical bottleneck for training and full-sequence processing. This paper introduces Sparse Query Attention (SQA), a novel attention architecture that pursues an alternative and complementary optimization path. Instead of reducing Key/Value heads, SQA reduces the number of Query heads. This architectural modification directly decreases the computational complexity of the attention mechanism by a factor proportional to the reduction in query heads, thereby lowering the overall FLOPs. This work presents the theoretical foundation of SQA, its mathematical formulation, and a family of architectural variants. Empirical benchmarks on long sequences (32k-200k tokens) demonstrate that SQA can achieve significant throughput improvements of up to 3x in computation-bound scenarios such as model pre-training, fine-tuning, and encoder-based tasks, with only a minimal impact on model quality in preliminary smallscale experiments. SQA was discovered serendipitously during the development of the upcoming Reactive Transformer architecture, suggesting its potential as a powerful tool for building more efficient and scalable models

We present TimeBraid, a series of unified time-series and language models that align pretrained language models and pretrained time-series foundation models through interleaved global residual attention layers. Each model inherits knowledge, instruction following, and reasoning from one side, continuous-signal perception and zero-shot forecasting from the other, and fuses the two in a shared representation space where both modalities are understood and generated. We study the design choices that make such unified modeling work: where to align the two representation spaces, how to ground language in temporal structure, how to balance understanding with generation, and how to keep joint optimization stable. The resulting recipe combines a unified prompting scheme for diverse time-series and text tasks, stabilized joint training, and supervision from 2.2M curated series--text pairs and 4.9M instruction-tuning samples. Across benchmarks spanning time-series perception, understanding, reasoning, and both context-aided and unimodal forecasting, TimeBraid remains competitive with far larger general-purpose models and task-specific counterparts.

How well do AI systems perform in algorithm engineering for hard optimization problems in domains such as package-delivery routing, crew scheduling, factory production planning, and power-grid balancing? We introduce ALE-Bench, a new benchmark for evaluating AI systems on score-based algorithmic programming contests. Drawing on real tasks from the AtCoder Heuristic Contests, ALE-Bench presents optimization problems that are computationally hard and admit no known exact solution. Unlike short-duration, pass/fail coding benchmarks, ALE-Bench encourages iterative solution refinement over long time horizons. Our software framework supports interactive agent architectures that leverage test-run feedback and visualizations. Our evaluation of frontier LLMs revealed that while they demonstrate high performance on specific problems, a notable gap remains compared to humans in terms of consistency across problems and long-horizon problem-solving capabilities. This highlights the need for this benchmark to foster future AI advancements.

Outcome-driven reinforcement learning has advanced reasoning in large language models (LLMs), but prevailing tool-augmented approaches train a single, monolithic policy that interleaves thoughts and tool calls under full context; this scales poorly with long horizons and diverse tools and generalizes weakly to new scenarios. Agentic systems offer a promising alternative by decomposing work across specialized modules, yet most remain training-free or rely on offline training decoupled from the live dynamics of multi-turn interaction. We introduce AgentFlow, a trainable, in-the-flow agentic framework that coordinates four modules (planner, executor, verifier, generator) through an evolving memory and directly optimizes its planner inside the multi-turn loop. To train on-policy in live environments, we propose Flow-based Group Refined Policy Optimization (Flow-GRPO), which tackles long-horizon, sparse-reward credit assignment by converting multi-turn optimization into a sequence of tractable single-turn policy updates. It broadcasts a single, verifiable trajectory-level outcome to every turn to align local planner decisions with global success and stabilizes learning with group-normalized advantages. Across ten benchmarks, AgentFlow with a 7B-scale backbone outperforms top-performing baselines with average accuracy gains of 14.9% on search, 14.0% on agentic, 14.5% on mathematical, and 4.1% on scientific tasks, even surpassing larger proprietary models like GPT-4o. Further analyses confirm the benefits of in-the-flow optimization, showing improved planning, enhanced tool-calling reliability, and positive scaling with model size and reasoning turns.

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.

Large language models are informing decisions with ever-higher stakes. As the consequences of their errors grow, a central question becomes harder to ignore: how much can we trust an individual answer? Yet recognizing when to defer remains difficult because language models can present incorrect conclusions with fluent explanations and an authoritative tone. Uncertainty quantification seeks to address this disconnect by estimating the reliability of individual predictions. However, many existing methods require repeated generations or separately trained components, and their scalar estimates do not reveal where uncertainty arises or how it evolves during reasoning. Recent work has also shown that generation length can be strongly associated with uncertainty estimates and correctness, raising the question of how much of an estimator's predictive power comes from uncertainty-specific information rather than output length alone. Mechanistic interpretability offers a way to address these limitations by connecting human-interpretable concepts to intermediate model states. Building on this capability, we introduce the U-Space, a low-dimensional subspace that makes a model's evolving uncertainty measurable and interpretable. We identify semantic anchors for doubt and certainty, map their unembedding directions back into the residual space, and combine their contrasts into an orthogonal basis. The U-Lens projects each token state onto these basis vectors, yielding an interpretable token-level uncertainty map that can be inspected directly or aggregated into a scalar uncertainty score. Our approach requires no correctness labels, repeated generations, or training. Across reasoning benchmarks, its confidence score outperforms established baselines under both standard and length-controlled evaluation and transfers more reliably than supervised estimators. Code: https://github.com/s2labres/U-Space.

Humans perceive far more in a scene than what is explicitly depicted: a single glance captures past causes and future trajectories; a quick peek determines if a vehicle can fit between two parked cars; a few seconds of video reveals who holds authority in a room; and a fleeting clip highlights subtle abstract patterns like unwritten rules or hidden labels. This capacity reflects a form of humanity's sixth sense: an intuitive reasoning mechanism that recovers implicit information beyond raw sensory perception. Crucially, this rapid, zero-shot visual intuition underpins everyday navigation and social interaction, making it a vital capability for Multimodal Large Language Models (MLLMs) deployed alongside people. Existing visual benchmarks, however, target either deliberate expert-level analysis in academic and mathematical domains or low-level perception, leaving the intuitive reasoning that people perform largely untested. To bridge this gap, we introduce Humanity's Sixth Sense (HSS), a benchmark for intuitive visual reasoning. HSS spans diverse image and video inputs, organizes items under a structured taxonomy, and pairs each with human-written prompts probing the implicit temporal, spatial, social, and abstract structure that people infer at a glance. Frontier MLLMs fall short of human performance: participants reach 93.1% accuracy, while the strongest model, GPT-6-astra, reaches only 53.6% even at maximum reasoning effort. Despite excelling in many complex tasks that require advanced perception and knowledge, current models still struggle significantly on these visual tasks that are intuitive for humans. We further explore agentic setup that apply dynamic visual manipulation to HSS, which narrows but does not close the gap. HSS establishes intuitive visual reasoning as a measurable axis and directs attention to a capability that scaling on current benchmarks has so far left behind.

Benchmarks are important tools for tracking the rapid advancements in large language model (LLM) capabilities. However, benchmarks are not keeping pace in difficulty: LLMs now achieve over 90\% accuracy on popular benchmarks like MMLU, limiting informed measurement of state-of-the-art LLM capabilities. In response, we introduce Humanity's Last Exam (HLE), a multi-modal benchmark at the frontier of human knowledge, designed to be the final closed-ended academic benchmark of its kind with broad subject coverage. HLE consists of 2,500 questions across dozens of subjects, including mathematics, humanities, and the natural sciences. HLE is developed globally by subject-matter experts and consists of multiple-choice and short-answer questions suitable for automated grading. Each question has a known solution that is unambiguous and easily verifiable, but cannot be quickly answered via internet retrieval. State-of-the-art LLMs demonstrate low accuracy and calibration on HLE, highlighting a significant gap between current LLM capabilities and the expert human frontier on closed-ended academic questions. To inform research and policymaking upon a clear understanding of model capabilities, we publicly release HLE at https://lastexam.ai.

Reinforcement Learning with Verifiable Rewards (RLVR) is a powerful framework for improving the reasoning abilities of Large Language Models (LLMs). However, current methods such as GRPO rely only on problems where the model responses to the same input differ in correctness, while ignoring those where all responses receive the same reward - so-called zero-variance prompts. In this work, we argue that such prompts are not useless but can, in fact, provide meaningful feedback for policy optimization. To this end, we introduce RL with Zero-Variance Prompts (RL-ZVP), a novel algorithm that extract learning signals from zero-variance prompts. RL-ZVP directly rewards correctness and penalizes errors even without contrasting responses, modulating feedback with token-level characteristics to preserve informative, nuanced signals. Across six math reasoning benchmarks, RL-ZVP achieves significant improvements of up to 8.61 points in accuracy and 7.77 points in pass rate over GRPO, while consistently outperforming other baselines that filter out zero-variance prompts. These results highlight the untapped potential of learning from zero-variance prompts in RLVR.

Tool-integrated reasoning has emerged as a key focus for enabling agentic applications. Among these, DeepResearch Agents have gained significant attention for their strong performance on complex, open-ended information-seeking tasks. We introduce Fathom-DeepResearch, an agentic system composed of two specialized models. The first is Fathom-Search-4B, a DeepSearch model trained from Qwen3-4B and optimized for evidence-based investigation through live web search and targeted webpage querying. Its training combines three advances: (i) DUETQA, a 5K-sample dataset generated via multi-agent self-play that enforces strict web-search dependence and heterogeneous source grounding; (ii) RAPO, a zero-overhead extension of GRPO that stabilizes multi-turn Reinforcement Learning with Verifiable Rewards through curriculum pruning, reward-aware advantage scaling, and per-prompt replay buffers; and (iii) a steerable step-level reward that classifies each tool call by cognitive behavior and marginal utility, enabling explicit control over search trajectory breadth, depth, and horizon. These improvements enable reliable extension of tool-calling beyond 20 calls when warranted. The second is Fathom-Synthesizer-4B, trained from Qwen3-4B, which converts multi-turn DeepSearch traces into structured, citation-dense DeepResearch Reports for comprehensive synthesis. Evaluated on DeepSearch benchmarks (SimpleQA, FRAMES, WebWalker, Seal0, MuSiQue) and DeepResearch-Bench, the system achieves state-of-the-art performance in the open-weights category while demonstrating strong generalization to diverse reasoning tasks including HLE, AIME-25, GPQA-Diamond, and MedQA.

While large language models (LLMs) have demonstrated strong performance on factoid question answering, they are still prone to hallucination and untruthful responses, particularly when tasks demand information outside their parametric knowledge. Indeed, truthfulness requires more than accuracy -- models must also recognize uncertainty and abstain when unsure to avoid hallucinations. This presents a fundamental challenge for existing methods: approaches that optimize for accuracy often amplify hallucinations, while those that encourage abstention can become overly conservative, sacrificing correct answers. Both extremes ultimately compromise truthfulness. In this work, we present TruthRL, a general reinforcement learning (RL) framework that directly optimizes the truthfulness of LLMs. Specifically, we implement TruthRL using GRPO with a simple yet effective ternary reward that distinguishes correct answers, hallucinations, and abstentions. It incentivizes models to reduce hallucinations not only by providing correct responses, but also by enabling abstention when uncertain, thereby improving truthfulness. Extensive experiments across four knowledge-intensive benchmarks show that, compared to vanilla RL, TruthRL significantly reduces hallucinations by 28.9% and improves truthfulness by 21.1%, with consistent gains across various backbone models (e.g., Qwen, Llama) under both retrieval and non-retrieval setups. In-depth ablation study demonstrates that vanilla accuracy-driven methods, such as supervised fine-tuning or RL with a binary reward, struggle to balance factual correctness and uncertainty. In contrast, our proposed truthfulness-driven TruthRL achieves strong performance in both accuracy and truthfulness, underscoring the importance of learning objective design for developing truthful LLMs.

Distribution matching is central to many vision and graphics tasks, where the widely used Wasserstein distance is too costly to compute for high dimensional distributions. The Sliced Wasserstein Distance (SWD) offers a scalable alternative, yet its Monte Carlo estimator suffers from high variance, resulting in noisy gradients and slow convergence. We introduce Reservoir SWD (ReSWD), which integrates Weighted Reservoir Sampling into SWD to adaptively retain informative projection directions in optimization steps, resulting in stable gradients while remaining unbiased. Experiments on synthetic benchmarks and real-world tasks such as color correction and diffusion guidance show that ReSWD consistently outperforms standard SWD and other variance reduction baselines. Project page: https://reservoirswd.github.io/

We present Apriel-1.5-15B-Thinker, a 15-billion parameter open-weights multimodal reasoning model that achieves frontier-level performance through training design rather than sheer scale. Starting from Pixtral-12B, we apply a progressive three-stage methodology: (1) depth upscaling to expand reasoning capacity without pretraining from scratch, (2) staged continual pre-training that first develops foundational text and vision understanding, then enhances visual reasoning through targeted synthetic data generation addressing spatial structure, compositional understanding, and fine-grained perception, and (3) high-quality text-only supervised fine-tuning on curated instruction-response pairs with explicit reasoning traces spanning mathematics, coding, science, and tool use. Notably, our model achieves competitive results without reinforcement learning or preference optimization, isolating the contribution of our data-centric continual pre-training approach. On the Artificial Analysis Intelligence Index, Apriel-1.5-15B-Thinker attains a score of 52, matching DeepSeek-R1-0528 despite requiring significantly fewer computational resources. Across ten image benchmarks, its performance is on average within five points of Gemini-2.5-Flash and Claude Sonnet-3.7, a key achievement for a model operating within single-GPU deployment constraints. Our results demonstrate that thoughtful mid-training 2 design can close substantial capability gaps without massive scale, making frontier-level multimodal reasoning accessible to organizations with limited infrastructure. We release the model checkpoint, all training recipes, and evaluation protocols under the MIT license to to advance open-source research.

Clinical early warning systems built on irregularly sampled medical time series (ISMTS) from electronic health records must deliver continuous risk scores for patient triage as well as interpretable rationales that clinicians can verify. Large language models (LLMs) are uniquely positioned for both, deriving risk from their output probabilities and rationales from their medical knowledge. However, we find that conventional LLM reasoning collapses graded risk into overconfident predictions and thereby undermines the cross-patient comparability on which triage depends. We refer to this failure mode as risk polarization and identify two underlying behaviors: early commitment to a single outcome, and one-sided reasoning that focuses only on the evidence for that outcome. To address this, we propose TRIAGE, a framework that trains an LLM to reason dialectically over competing clinical outcomes by eliciting outcome-specific rationales. This dialectical formulation mitigates risk polarization, enabling a single LLM to jointly provide explicit clinical rationales and risk scores comparable across patients. Across five ISMTS benchmarks, TRIAGE improves mean AUPRC by 17.0% and reduces mean calibration error by 82.8% relative to the competitive LLM-based baseline, while surpassing the strongest ISMTS baseline by 3.5% in mean AUPRC.

Autoregressive (AR) grounding models serialize spatial predictions, introducing sequential latency and imposing a causal order on output tokens. We view grounding as visual evidence extraction: objects, locations, and spatial relations are jointly constrained by the image and query, yet their dependencies do not imply an intrinsic left-to-right generation order. This distinction makes bidirectional diffusion a natural fit, allowing spatial hypotheses to emerge in parallel and be jointly refined through iterative denoising. We introduce GroundAnything, a 4B-parameter grounding foundation model that reconciles fast parallel decoding with precise localization through blockwise denoising. Training combines grounding pretraining from public datasets and dedicated data engines, direct AR-to-diffusion conversion with joint AR and diffusion objectives, supervised fine-tuning, and GRPO-based reinforcement post-training. Across 30 grounding benchmarks, our autoregressive variant, GroundAnything-VLM, establishes a new overall state of the art among similarly sized models at 72.42%, remaining competitive with GPT-6 Astra (71.35%). With entropy-guided decoding, GroundAnything also surpasses the prior state of the art at this scale, averaging 61.75% versus 53.32% for the fast MTP-based LocateAnything model. We further explore decoding strategies, showing that an optional self-speculative mode achieves a 4.51times speedup over the AR counterpart with a 0.74 percentage-point drop in COCO F1mIoU. Infrastructure experiments show that progressive inference optimizations translate parallel decoding into practical speedups. These support efficient visual grounding in latency-sensitive real-world systems.

Language model agents are deployed with a harness, the software around the model that manages its context, tools, and feedback. When such an agent is distilled into a smaller one, the harness stays in place, so the student mainly needs the teacher-specific abilities that the harness cannot provide, such as acting correctly on harness information. Standard distillation, however, imitates the teacher's full outputs and treats the harness as part of the input. We propose Harness-Aware Distillation (HAD), which focuses distillation on what the teacher adds beyond the harness. HAD complements on-policy distillation with two components: an action preference that contrasts the same teacher's actions with and without the harness information, scored after the student's own reasoning, and a validity check that drops preference pairs whose preferred action contradicts the harness records. We show that the contrast gives the student information that imitating the teacher alone cannot provide, and HAD needs no task rewards, success labels, or future information. Across multiple long-horizon agent benchmarks and models, HAD outperforms on-policy distillation baselines with the same fixed harness. Our analysis shows that HAD enters fewer unproductive loops and recovers from errors more often than the baselines, and suggests that it adaptively keeps learnable feedback in its weights while reading state information from the harness.

Much knowledge work produces new deliverables from files a workspace already holds, and LLM agents are beginning to take such work over. Through direct corpus interaction, an agent can search and read any of those files from a terminal with no indexing, and producing a deliverable from many of them in this way is what we call direct workspace interaction (DWI). Reaching the files, however, is only half the task: nothing keeps track of what the task asks for, what has been read, and what was listed but never opened, all of which slip through the context window without leaving a trace, so an agent may extract a figure and still deliver a report without it. To address this, we present RunningTab, a framework that equips direct workspace interaction with an environment-side tab: a per-task record of what the task still owes, kept by the environment alongside the agent. Specifically, the agent adds its requirements, while the environment records every file read as an excerpt with its provenance and every listed but unopened file as a candidate; the agent can then see each requirement beside its best-matching excerpts and top unopened candidates, resolve it against matching content or set it aside with a reason, and, should it try to finish with requirements still open, receive them in a finish check. We validate RunningTab on three benchmarks with three LLMs, where it consistently outperforms plain DWI and baselines that keep the record in the model, while its tab usually holds the values a deliverable needs once seen.

Intellectual Property (IP) is a unique domain that integrates technical and legal knowledge, making it inherently complex and knowledge-intensive. As large language models (LLMs) continue to advance, they show great potential for processing IP tasks, enabling more efficient analysis, understanding, and generation of IP-related content. However, existing datasets and benchmarks either focus narrowly on patents or cover limited aspects of the IP field, lacking alignment with real-world scenarios. To bridge this gap, we introduce the first comprehensive IP task taxonomy and a large, diverse bilingual benchmark, IPBench, covering 8 IP mechanisms and 20 tasks. This benchmark is designed to evaluate LLMs in real-world intellectual property applications, encompassing both understanding and generation. We benchmark 16 LLMs, ranging from general-purpose to domain-specific models, and find that even the best-performing model achieves only 75.8% accuracy, revealing substantial room for improvement. Notably, open-source IP and law-oriented models lag behind closed-source general-purpose models. We publicly release all data and code of IPBench and will continue to update it with additional IP-related tasks to better reflect real-world challenges in the intellectual property domain.

Large language models (LLMs) have demonstrated remarkable capabilities in complex reasoning tasks when equipped with external tools. However, current frameworks predominantly rely on sequential processing, leading to inefficient execution particularly for tasks requiring extensive tool interaction. This paper introduces Flash-Searcher, a novel parallel agent reasoning framework that fundamentally reimagines the execution paradigm from sequential chains to directed acyclic graphs (DAGs). Flash-Searcher decomposes complex tasks into subtasks with explicit dependencies, enabling concurrent execution of independent reasoning paths while maintaining logical constraints. Through dynamic workflow optimization, our framework continuously refines the execution graph based on intermediate results, effectively integrating summary module. Comprehensive evaluations across multiple benchmarks demonstrate that Flash-Searcher consistently outperforms existing approaches. Specifically, it achieves 67.7% accuracy on BrowseComp and 83% on xbench-DeepSearch, while reducing agent execution steps by up to 35% compared to current frameworks. Furthermore, when distilling this parallel reasoning pipeline into single models, we observe substantial performance gains across diverse backbone architectures, underscoring the generalizability of our methodology. Our work thus represents a significant advance in agent architecture design, offering a more scalable and efficient paradigm for complex reasoning tasks.

Memory systems have shown promise for improving agent performance, but their potential remains largely unexplored for small language models, which struggle to generate sufficient successful trajectories on their own. We propose Agent Memory Distillation (AMD), a training-free framework that transfers structured knowledge from a large teacher agent to a small student agent through hierarchical memory. AMD constructs three complementary memory types from successful teacher trajectories: Workflow memory encodes task-level strategies, Subtask memory provides concrete behavioral examples at an intermediate granularity, and Function memory captures per-function calling conventions and common pitfalls. Workflow and Subtask memories are injected proactively at the start of each task, while Function memory is retrieved reactively upon tool-calling errors. We evaluate AMD on three tool-use benchmarks using four student models (4B-8B parameters) with GPT-5-mini as the teacher, achieving average accuracy gains of 27.2%p, 11.2%p, and 3.4%p on AppWorld, BFCL V3, and ToolSandbox, while consistently outperforming existing memory-based baselines. Further analysis shows that Subtask memory contributes the largest gains, teacher effectiveness depends on both teacher capability and student compatibility, and 4B-sized students benefit most from AMD.

AI agents are beginning to automate research and development across the AI stack, from improving training efficiency to optimizing inference. A natural next step is to improve the research efficiency of the agents themselves. When an AI research agent's own code is the object of optimization, each accepted rewrite becomes the agent that the next round edits. We refer to this loop as recursive self-improvement. Its significance lies in a long-standing trend, in which increased cumulative spending on R&D yields diminishing returns. Sustained self-improvement offers a way to counter this trend. We present AIDE^2, a system that implements this loop for a frontier AI research agent. It proposes changes to its own code, benchmarks modified versions of itself on a suite of AI R&D tasks, and keeps the changes that perform best on hidden evaluations. In an autonomous 8-day run, AIDE^2 discovered seven successive improvements, ranging from a new search policy to memory mechanisms that compress and manage the agent's growing context. These gains generalize to four held-out benchmarks spanning machine learning engineering, heuristic algorithm engineering, and physics-based weather forecasting, the last of which is out of distribution from the selection tasks. On all four, the strongest discovered agent matches or exceeds a human-engineered production research agent that ranks among the strongest on FML-Bench. On a separate held-out task family, the discovered agents also exhibit reduced reward hacking, a property the loop never explicitly optimized for: the rate falls from 55% to 32% during the run, 7 percentage points below the human-engineered agent. Together, these results show that an AI research agent can improve its own research efficiency through recursive self-improvement, and that these gains transfer to tasks and domains the loop never encountered.

Scaling test-time computation is a powerful way to improve language-model reasoning, and is particularly appealing for small reasoning models (sRMs) that are cheap to serve. However, is additional thinking always the right operation? By intervening at intermediate reasoning states across two model families and multiple scales, we find that self-refinement largely consolidates probability mass onto solutions already reachable from the current state, rather than making new ones reachable. These interventions reveal two failure regimes: execution bottlenecks, where the correct path is reachable and reflection can recover it, and knowledge bottlenecks, where relevant external information makes it reachable. Motivated by this distinction, we introduce FlyBy, a selective querying framework, and train 4B and 8B variants to reason first, diagnose what remains unresolved, and, at a knowledge bottleneck, query stronger models whose parametric knowledge extends beyond its own. Supervised fine-tuning bootstraps a multi-depth query action, and cost-aware reinforcement learning calibrates whether to query, what to ask, and how much to spend. On 1,158 hard problems across six benchmarks, FlyBy-4B achieves 45.96% pass@8, surpassing Qwen3-14B (41.64%) at 2.7 times lower serving cost, while also exceeding Qwen3-8B in pass@1 (16.85% vs. 15.31%). Scaling to FlyBy-8B further improves pass@8 to 51.81%.

Precise grounding matters. It specifies which object is the target and where that object is, even in clutter and for tiny objects, and it has to be fast enough for closed-loop control. Yet vision-language-action (VLA) and world-action models (WAMs) take perception from general-purpose vision-language and video-generation backbones, which still fail in these settings. We introduce GroundingPI, a 4B grounding foundation model that generates points and boxes as quantized coordinates in a shared vocabulary. Training combines multimodal and spatial pretraining, supervised fine-tuning, and reinforcement learning with GRPO, using supervision from public datasets and dedicated data engines. Against 44 baselines across 34 grounding benchmarks spanning 11 perceptual capabilities, GroundingPI establishes a new state of the art, averaging 73.68%, above the larger GPT-6 Astra (71.54%). As a downstream visual backbone, GroundingPI improves performance on robotic manipulation and autonomous driving. On RoboTwin 2.0, it outperforms every mainstream backbone we evaluate in all four out-of-distribution settings, by up to 24.8% relative to the strongest backbone. On RoboCasa-GR1, GroundingPI trained with 50% of the demonstrations outperforms those baselines trained with 75%. On nuScenes, used as the visual backbone, GroundingPI attains an average open-loop L2 error of 0.296 m. We systematically analyze GroundingPI's pretraining in scale and data composition. Downstream autonomous driving and robotic manipulation improve as the pretraining is scaled. Analyzing the data recipe across these 11 perceptual capabilities shows dense grounding's substantial benefits for both, and OCR's potential as a catalyst for perceptual learning. These results support grounding as a perceptual foundation, and dedicated perceptual pretraining as a promising direction for foundation models of physical intelligence.

Large language models (LLMs) are increasingly used in learning algorithms, evaluations, and optimization tasks. Recent studies have shown that using LLM-based optimizers to automatically optimize model prompts, demonstrations, predictions themselves, or other components can significantly enhance the performance of AI systems, as demonstrated by frameworks such as DSPy and TextGrad. However, optimizers built on language models themselves are usually designed by humans with manual design choices; optimizers themselves are not optimized. Moreover, these optimizers are general purpose by design, to be useful to a broad audience, and are not tailored for specific tasks. To address these challenges, we propose metaTextGrad, which focuses on designing a meta-optimizer to further enhance existing optimizers and align them to be good optimizers for a given task. Our approach consists of two key components: a meta prompt optimizer and a meta structure optimizer. The combination of these two significantly improves performance across multiple benchmarks, achieving an average absolute performance improvement of up to 6% compared to the best baseline.

Recent advances in video generation have enabled high-fidelity video synthesis from user provided prompts. However, existing models and benchmarks fail to capture the complexity and requirements of professional video generation. Towards that goal, we introduce Stable Cinemetrics, a structured evaluation framework that formalizes filmmaking controls into four disentangled, hierarchical taxonomies: Setup, Event, Lighting, and Camera. Together, these taxonomies define 76 fine-grained control nodes grounded in industry practices. Using these taxonomies, we construct a benchmark of prompts aligned with professional use cases and develop an automated pipeline for prompt categorization and question generation, enabling independent evaluation of each control dimension. We conduct a large-scale human study spanning 10+ models and 20K videos, annotated by a pool of 80+ film professionals. Our analysis, both coarse and fine-grained reveal that even the strongest current models exhibit significant gaps, particularly in Events and Camera-related controls. To enable scalable evaluation, we train an automatic evaluator, a vision-language model aligned with expert annotations that outperforms existing zero-shot baselines. SCINE is the first approach to situate professional video generation within the landscape of video generative models, introducing taxonomies centered around cinematic controls and supporting them with structured evaluation pipelines and detailed analyses to guide future research.

Modern audio-language models are no longer judged only on what words they can transcribe, but on whether they can reason over what they hear: recovering meaning that lives in tone and prosody, telling dialects and regional languages apart, and resolving ambiguity that the written form leaves open. This capability is now measured by a growing family of audio-reasoning benchmarks, but almost entirely in English and on general-domain audio. Southeast Asia (SEA) is served instead by benchmarks that inherit an English task taxonomy of recognition, translation, and paralinguistic classification, and therefore test whether a model hears SEA speech rather than whether it can reason from it. We introduce SEABED, an audio-first question answering dataset designed to benchmark language and audio reasoning models on SEA speech. SEABED comprises a suite of six audio-reasoning tasks built entirely from real, openly available SEA speech corpora, yielding 5,404 question-answer pairs. We evaluate six frontier and region-specific audio LLMs: even the state-of-the-art model Gemini 3.5 Flash achieves only 50.3% weighted average accuracy. SEABED evaluates not only answer accuracy, but also whether models' stated reasoning is grounded in the audio evidence. A sample of the benchmark data is available here: https://huggingface.co/datasets/CentificAIResearch/SEABED.

"System One" decision models such as TypeSafe's Jev and its open counterpart Laya answer typed questions about a text in a single forward pass with calibrated probabilities, but they cannot ask for missing information: when a first message does not say what separates two departments, they guess. We present LAVOIR (Laya with Value-Of-Information Routing), which places the candidate pieces of missing information (slots) in the input next to the answer options, so that one forward pass returns both the decision distribution and, for every slot, the expected gain in the probability of the correct decision if the user were asked about it. VOI targets need no human labels: gold decisions come from schema rules, an LLM only verbalizes messages and answers, a model from another family checks every text, and pairing each message with several profiles makes regression on realized gains estimate the expected gain. A Gini-impurity cap bounds the predicted value by what a calibrated model can still gain. In a controlled study, decisions on seen schemas are statistically indistinguishable from the Bayes ceiling. The final model's question policy matches a greedy oracle VOI policy on seen schemas (AUC 0.799 vs. 0.797), and with at most 0.5 questions per conversation it is 14.1 points more accurate than never asking. On real ABCD conversations, one real exchange raises accuracy by 8.3 points where LAVOIR asks and leaves it unchanged where it does not; on SGD the cap lowers the asking rate from 93% to 8.6%. On Laya's twelve benchmarks LAVOIR is above Laya's reported scores on seven, and it answers a question in 31 ms (median, GH200).

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

Continual post-training faces a dual challenge of acquisition and retention: models must learn new capabilities while preserving prior knowledge. While Self-Distillation Fine-Tuning (SDFT) mitigates this using self-supervision with in-context demonstrations, it relies on strong in-context learning and instruction-following abilities. Consequently, SDFT fails on smaller models critical for low-latency, on-device use. To enable continual learning at smaller scales, we first identify the root cause of failure. We show that poor acquisition and forgetting correlate linearly with the post-trained model's distribution shift, revealing the necessity of in-distribution learning. We thus introduce Synthetic Self-Distillation (SynthSelf), imitating SDFT's in-distribution rewrite by outsourcing the task to a minimally larger, same-family aligner. Coupled with off-policy supervision and rejection sampling, this provides the high-quality supervision small models require without breaking their original distribution. Evaluations spanning 4 acquisition datasets (including a mixture) and 14 retention benchmarks show SynthSelf consistently outperforms 6 methods (including off- and on-policy distillation) across 3 small model scales. Extensive further analyses confirm SynthSelf paves the way toward effective small-model continual learning.

Unified audio models capable of audio understanding, audio generation and, increasingly, audio editing are proliferating rapidly. Yet a basic question about them remains unanswered: do the two heads of a unified model agree about the same audio? Current practice evaluates each capability in isolation on specialized benchmarks, and never asks whether a model can make sense of its own generations. We present TORUS, the first self-coherence test for audio-native unified models. TORUS comprises 48 three-stage self-coherence tests carrying 432 six-option questions spanning speech, sound and music across five task families. We holistically evaluate five open unified models alongside a Cascaded Baseline that combines state-of-the-art specialized generation, editing and understanding models. The best unified model answers 50.5% of questions against the Cascaded Baseline's 63.2% and a 16.7% chance floor. Models struggle on audio editing. Among the evaluated audio models (specialized and unified), we observe limited self-coherence, and thus position self-coherence as an essential test for future audio systems.

World-space hand motion estimation from egocentric video requires recovering 3D articulated hand geometry while tracking camera egomotion. Existing approaches heavily rely on cascading independent hand pose estimators and SLAM systems, resulting in error accumulation, complex pipelines, and severe computational overhead. To address these limitations, we present InfiniHand, an end-to-end streaming feed-forward framework that jointly estimates MANO parameters, camera trajectories, and hand locations directly from uncalibrated egocentric video. InfiniHand integrates persistent spatiotemporal memory with hand-centered visual features, explicitly coupling camera motion with local hand geometry within a unified architecture. We train InfiniHand in two progressive stages by first learning robust camera-space hand priors and then extending to streaming world-space reconstruction. To support this process, we aggregate a pretraining corpus of approximately 5,000 hours of egocentric data across multiple public datasets. Extensive evaluations demonstrate that InfiniHand outperforms state-of-the-art baselines on in-domain benchmarks, achieving a 21.4% reduction in ARCTIC PA-p compared to ViDiHand while substantially mitigating world-space drift. Furthermore, InfiniHand generalizes robustly to in-the-wild videos and operates at 11.19 FPS, delivering more than twice the throughput of HaWoR.

Enhancing existing models with new knowledge is a crucial aspect of AI development. This paper introduces a novel method for integrating a new language into a large language model (LLM). Our approach successfully incorporates a previously unseen target language into an existing LLM without compromising its prior knowledge. We trained a tiny model with 1.5 billion parameters named Kuwain by injecting the Arabic language into a small open-source model mainly trained in English. Our method demonstrates significant improvements in Arabic language performance, with an average 8% improvement across various benchmarks, while retaining the model's existing knowledge with a minimum amount of the original model's data. This offers a cost-effective alternative to training a comprehensive model in both English and Arabic. The results highlight the potential for efficient, targeted language model expansion without extensive retraining or resource-intensive processes.

We introduce Mutarjim, a compact yet powerful language model for bidirectional Arabic-English translation. While large-scale LLMs have shown impressive progress in natural language processing tasks, including machine translation, smaller models. Leveraging this insight, we developed Mutarjim based on Kuwain-1.5B , a language model tailored for both Arabic and English. Despite its modest size, Mutarjim outperforms much larger models on several established benchmarks, achieved through an optimized two-phase training approach and a carefully curated, high-quality training corpus.. Experimental results show that Mutarjim rivals models up to 20 times larger while significantly reducing computational costs and training requirements. We also introduce Tarjama-25, a new benchmark designed to overcome limitations in existing Arabic-English benchmarking datasets, such as domain narrowness, short sentence lengths, and English-source bias. Tarjama-25 comprises 5,000 expert-reviewed sentence pairs and spans a wide range of domains, offering a more comprehensive and balanced evaluation framework. Notably, Mutarjim achieves state-of-the-art performance on the English-to-Arabic task in Tarjama-25, surpassing even significantly larger and proprietary models like GPT-4o mini. We publicly release Tarjama-25 to support future research and advance the evaluation of Arabic-English translation systems.

Test-Driven Development (TDD) is a widely adopted software engineering practice that requires developers to create and execute tests alongside code implementation, ensuring that software behavior is continuously validated and refined. In the era of vibe coding, where developers increasingly delegate code writing to large language models (LLMs) by specifying high-level intentions, TDD becomes even more crucial, as test cases serve as executable specifications that explicitly define and verify intended functionality beyond what natural-language descriptions and code context can convey. While vibe coding under TDD is promising, there are three main challenges: (1) selecting a small yet effective test suite to improve the generation accuracy and control the execution workload, (2) retrieving context such as relevant code effectively, and (3) systematically using test feedback for effective code refinement. To address these challenges, we introduce TENET, an LLM agent for generating functions in complex real-world repositories under the TDD setting. TENET features three components: (1) a novel test harness mechanism that selects a concise test suite to maximize diversity of target usage scenarios; (2) a tailored agent toolset that performs efficient retrieval of relevant code with interactive debugging; and (3) a reflection-based refinement workflow that iteratively analyzes failures, replenishes context, and applies code refinement. TENET achieves 69.08% and 81.77% Pass@1 on RepoCod and RepoEval benchmarks, outperforming the best agentic baselines by 9.49 and 2.17 percentage points, respectively. In addition, this is the first study of test-driven code generation with repository-level context, examining how different aspects of test suites affect the performance of LLM agents under the TDD setting.

Vision-Language Models (VLMs) excel at high-level scene understanding but falter on fine-grained perception tasks requiring precise localization. This failure stems from a fundamental mismatch, as generating exact numerical coordinates is a challenging task for language-centric architectures. In this paper, we introduce VLM-FO1, a novel framework that overcomes this limitation by reframing object-centric perception from a brittle coordinate generation problem into a robust feature retrieval task. Our method operates as a plug-and-play module that integrates with any pre-trained VLM. It leverages a Hybrid Fine-grained Region Encoder (HFRE), featuring a dual vision encoder, to generate powerful region tokens rich in both semantic and spatial detail. A token-based referencing system then enables the LLM to seamlessly reason about and ground language in these specific visual regions. Experiments show that VLM-FO1 achieves state-of-the-art performance across a diverse suite of benchmarks, demonstrating exceptional capabilities in object grounding, region generational understanding, and visual region reasoning. Crucially, our two-stage training strategy ensures that these perception gains are achieved without compromising the base model's general visual understanding capabilities. VLM-FO1 establishes an effective and flexible paradigm for building perception-aware VLMs, bridging the gap between high-level reasoning and fine-grained visual grounding.

Reasoning about the 3D world from multi-view images remains a fundamental challenge for Multimodal Large Language Models (MLLMs). While modern MLLMs handle single-image inputs effectively, they struggle to integrate evidence across viewpoints into a coherent 3D understanding. A growing body of work attempts to close this gap by injecting 3D awareness into MLLMs, either by boosting fine-grained pixel-level cross-view correspondence or by fusing features from 3D geometry foundation models, yet a substantial gap to human reasoning persists. In this work, we revisit human spatial reasoning, which suggests that rather than relying on fine-grained geometry cues, humans roughly identify common objects across views, infer the relative geometry between viewpoints, and assemble a coarse 3D layout of the scene. Inspired by this process, we introduce Imagine3D-LLM, an MLLM that learns to assemble a similar compact 3D representation of the scene and conditions its answer on this representation. Concretely, we append a small set of learnable summary tokens after the image tokens, decode them into a compact 3D Gaussian Splatting representation supervised by a photometric reconstruction loss, and train jointly with the standard next-token prediction objective. Notably, although only the summary tokens receive direct reconstruction supervision, this objective also induces stronger cross-frame correspondence within the LLM's underlying image features, suggesting that learning to reconstruct propagates 3D-aware signals throughout the model. As a result, Imagine3D-LLM consistently outperforms prior approaches across multiple spatial reasoning and 3D understanding benchmarks, suggesting that imagining the scene can be more effective than being told its pixel-wise geometry.

Looped transformers and Mixture-of-Experts (MoE) offer complementary routes to efficient scaling: recurrence increases computational depth at fixed parameters, while MoE sparsity expands total capacity at fixed active compute. Yet existing scaling laws model recurrence or sparsity in isolation. In this work, we introduce Loop Scaling Laws, the first scaling law to jointly model recurrence and sparsity alongside model size and data. At its core is a bounded, sparsity-conditional recurrence mapping that characterizes the effective-parameter gain from looping and how sparsity raises this gain. The laws predict the held-out loss of looped models more accurately than prior alternatives, and recover the standard dense and MoE scaling laws as special cases. Beyond prediction, the fitted laws provide a principled foundation for designing looped MoE models under compute and memory constraints. Downstream evaluations further demonstrate the complementary benefits of the two axes: sparsity delivers ~3x active-parameter efficiency, recurrence yields ~2x total-parameter efficiency on reasoning, and joint scaling further advances the performance frontier. As a practical extension, we show these gains hold at trillion-token scale: at matched training compute, a looped MoE with law-derived recurrence matches a ~2x larger non-looped MoE on the reasoning benchmarks, while enabling test-time scaling through recurrence.