Daily Papers

We examine how closely pretrained LeWorldModel (LeWM) reproduces the effects of small action changes. Using matched simulator and model responses, we measure action sensitivity with Gram matrices at three perturbation scales. Across nearly a thousand states from Reacher and OGBench Cube, the most sensitive action directions differ by median angles of about 36^circ and 42^circ, respectively, at the smallest scale. Increasing the perturbation tenfold reduces median normalized Gram discrepancy by roughly 70\% on Reacher and 47\% on Cube. Geometry measured in pixels and embeddings varies more across scales than the chosen physical-state geometry. On held-out directions, environment Grams achieve median embedding response errors of 17\% and 28\%, compared with 26\% and 38\% for an isotropic reference with the same trace. LeWM's errors are higher at this scale. These findings support scale-specific evaluation of action sensitivity; its effect on control remains an open question.

Accurate fish detection in underwater imagery is essential for ecological monitoring, aquaculture automation, and robotic perception. However, practical deployment remains limited by fragmented datasets, heterogeneous imaging conditions, and inconsistent evaluation protocols. To address these gaps, we present FishDet-M, the largest unified benchmark for fish detection, comprising 13 publicly available datasets spanning diverse aquatic environments including marine, brackish, occluded, and aquarium scenes. All data are harmonized using COCO-style annotations with both bounding boxes and segmentation masks, enabling consistent and scalable cross-domain evaluation. We systematically benchmark 28 contemporary object detection models, covering the YOLOv8 to YOLOv12 series, R-CNN based detectors, and DETR based models. Evaluations are conducted using standard metrics including mAP, mAP@50, and mAP@75, along with scale-specific analyses (AP_S, AP_M, AP_L) and inference profiling in terms of latency and parameter count. The results highlight the varying detection performance across models trained on FishDet-M, as well as the trade-off between accuracy and efficiency across models of different architectures. To support adaptive deployment, we introduce a CLIP-based model selection framework that leverages vision-language alignment to dynamically identify the most semantically appropriate detector for each input image. This zero-shot selection strategy achieves high performance without requiring ensemble computation, offering a scalable solution for real-time applications. FishDet-M establishes a standardized and reproducible platform for evaluating object detection in complex aquatic scenes. All datasets, pretrained models, and evaluation tools are publicly available to facilitate future research in underwater computer vision and intelligent marine systems.

Machine translation (MT) has become indispensable for cross-border communication in globalized industries like e-commerce, finance, and legal services, with recent advancements in large language models (LLMs) significantly enhancing translation quality. However, applying general-purpose MT models to industrial scenarios reveals critical limitations due to domain-specific terminology, cultural nuances, and stylistic conventions absent in generic benchmarks. Existing evaluation frameworks inadequately assess performance in specialized contexts, creating a gap between academic benchmarks and real-world efficacy. To address this, we propose a three-level translation capability framework: (1) Basic Linguistic Competence, (2) Domain-Specific Proficiency, and (3) Cultural Adaptation, emphasizing the need for holistic evaluation across these dimensions. We introduce TransBench, a benchmark tailored for industrial MT, initially targeting international e-commerce with 17,000 professionally translated sentences spanning 4 main scenarios and 33 language pairs. TransBench integrates traditional metrics (BLEU, TER) with Marco-MOS, a domain-specific evaluation model, and provides guidelines for reproducible benchmark construction. Our contributions include: (1) a structured framework for industrial MT evaluation, (2) the first publicly available benchmark for e-commerce translation, (3) novel metrics probing multi-level translation quality, and (4) open-sourced evaluation tools. This work bridges the evaluation gap, enabling researchers and practitioners to systematically assess and enhance MT systems for industry-specific needs.

With the rapid growth of video generative models (VGMs), it is essential to develop reliable and comprehensive automatic metrics for AI-generated videos (AIGVs). Existing methods either use off-the-shelf models optimized for other tasks or rely on human assessment data to train specialized evaluators. These approaches are constrained to specific evaluation aspects and are difficult to scale with the increasing demands for finer-grained and more comprehensive evaluations. To address this issue, this work investigates the feasibility of using multimodal large language models (MLLMs) as a unified evaluator for AIGVs, leveraging their strong visual perception and language understanding capabilities. To evaluate the performance of automatic metrics in unified AIGV evaluation, we introduce a benchmark called UVE-Bench. UVE-Bench collects videos generated by state-of-the-art VGMs and provides pairwise human preference annotations across 15 evaluation aspects. Using UVE-Bench, we extensively evaluate 16 MLLMs. Our empirical results suggest that while advanced MLLMs (e.g., Qwen2VL-72B and InternVL2.5-78B) still lag behind human evaluators, they demonstrate promising ability in unified AIGV evaluation, significantly surpassing existing specialized evaluation methods. Additionally, we conduct an in-depth analysis of key design choices that impact the performance of MLLM-driven evaluators, offering valuable insights for future research on AIGV evaluation. The code is available at https://github.com/bytedance/UVE.

Debates about whether development projects improve living conditions persist, partly because observational estimates can be biased by incomplete adjustment and because reliable outcome data are scarce at the neighborhood level. We address both issues in a continent-scale, sector-specific evaluation of Chinese and World Bank projects across 9,899 neighborhoods in 36 African countries (2002 to 2013), representative of 88% of the population. First, we use a recent dataset that measures living conditions with a machine-learned wealth index derived from contemporaneous satellite imagery, yielding a consistent panel of 6.7 km square mosaics. Second, to strengthen identification, we proxy officials' map-based placement criteria using pre-treatment daytime satellite images and fuse these with rich tabular covariates to estimate funder- and sector-specific ATEs via inverse-probability weighting. Incorporating imagery systematically shrinks effects relative to tabular-only models, indicating prior work likely overstated benefits. On average, both donors raise wealth, with larger gains for China; sector extremes in our sample include Trade and Tourism for the World Bank (+6.27 IWI points), and Emergency Response for China (+14.32). Assignment-mechanism analyses show World Bank placement is generally more predictable from imagery alone, as well as from tabular covariates. This suggests that Chinese project placements are more driven by non-visible, political, or event-driven factors than World Bank placements. To probe residual concerns about selection on observables, we also estimate within-neighborhood (unit) fixed-effects models at a spatial resolution about 450 times finer than prior fixed effects analyses, leveraging the computer-vision-imputed IWI panels; these deliver smaller but directionally consistent effects.

The rapid advancement and expanding applications of Audio Large Language Models (ALLMs) demand a rigorous understanding of their trustworthiness. However, systematic research on evaluating these models, particularly concerning risks unique to the audio modality, remains largely unexplored. Existing evaluation frameworks primarily focus on the text modality or address only a restricted set of safety dimensions, failing to adequately account for the unique characteristics and application scenarios inherent to the audio modality. We introduce AudioTrust-the first multifaceted trustworthiness evaluation framework and benchmark specifically designed for ALLMs. AudioTrust facilitates assessments across six key dimensions: fairness, hallucination, safety, privacy, robustness, and authentication. To comprehensively evaluate these dimensions, AudioTrust is structured around 18 distinct experimental setups. Its core is a meticulously constructed dataset of over 4,420 audio/text samples, drawn from real-world scenarios (e.g., daily conversations, emergency calls, voice assistant interactions), specifically designed to probe the multifaceted trustworthiness of ALLMs. For assessment, the benchmark carefully designs 9 audio-specific evaluation metrics, and we employ a large-scale automated pipeline for objective and scalable scoring of model outputs. Experimental results reveal the trustworthiness boundaries and limitations of current state-of-the-art open-source and closed-source ALLMs when confronted with various high-risk audio scenarios, offering valuable insights for the secure and trustworthy deployment of future audio models. Our platform and benchmark are available at https://github.com/JusperLee/AudioTrust.

Despite remarkable progress in image generation models, generating realistic hands remains a persistent challenge due to their complex articulation, varying viewpoints, and frequent occlusions. We present FoundHand, a large-scale domain-specific diffusion model for synthesizing single and dual hand images. To train our model, we introduce FoundHand-10M, a large-scale hand dataset with 2D keypoints and segmentation mask annotations. Our insight is to use 2D hand keypoints as a universal representation that encodes both hand articulation and camera viewpoint. FoundHand learns from image pairs to capture physically plausible hand articulations, natively enables precise control through 2D keypoints, and supports appearance control. Our model exhibits core capabilities that include the ability to repose hands, transfer hand appearance, and even synthesize novel views. This leads to zero-shot capabilities for fixing malformed hands in previously generated images, or synthesizing hand video sequences. We present extensive experiments and evaluations that demonstrate state-of-the-art performance of our method.

Evaluating long-form responses to research queries heavily relies on expert annotators, restricting attention to areas like AI where researchers can conveniently enlist colleagues. Yet, research expertise is widespread: survey articles synthesize knowledge distributed across the literature. We introduce ResearchQA, a resource for evaluating LLM systems by distilling survey articles from 75 research fields into 21K queries and 160K rubric items. Each rubric, derived jointly with queries from survey sections, lists query-specific answer evaluation criteria, i.e., citing papers, making explanations, and describing limitations. Assessments by 31 Ph.D. annotators in 8 fields indicate 96% of queries support Ph.D. information needs and 87% of rubric items should be addressed in system responses by a sentence or more. Using our rubrics, we are able to construct an automatic pairwise judge obtaining 74% agreement with expert judgments. We leverage ResearchQA to analyze competency gaps in 18 systems in over 7.6K pairwise evaluations. No parametric or retrieval-augmented system we evaluate exceeds 70% on covering rubric items, and the highest-ranking agentic system shows 75% coverage. Error analysis reveals that the highest-ranking system fully addresses less than 11% of citation rubric items, 48% of limitation items, and 49% of comparison items. We release our data to facilitate more comprehensive multi-field evaluations.

Product-centric advertisement video generation aims to create promotional videos that preserve fine-grained product identity while presenting selling points through coherent multi-shot narratives. However, this emerging task remains underexplored due to the lack of large-scale advertisement-specific datasets and comprehensive evaluation frameworks. To address this gap, we introduce AdSpark, a large-scale dataset and benchmark for product-centric advertisement video generation, based on data from a major e-commerce platform. AdSpark-300K contains approximately 300K reference image--prompt--video triplets, comprising a real-world subset and a synthetic subset. Each sample provides structured advertisement annotations, including product identity annotations, selling-point descriptions, creative plans, and aligned audio scripts, enabling models to learn product preservation and advertisement-oriented visual storytelling. We further propose AdSpark-Bench, a diagnostic benchmark that evaluates generated advertisements across six dimensions, including visual quality, product fidelity, instruction adherence, temporal coherence, audio alignment, and advertisement effectiveness. Based on AdSpark-Bench, we evaluate representative models, revealing key challenges in product preservation, multi-shot storytelling, and selling-point visualization. Experiments with AdSpark-300K-finetuned models further validate the effectiveness of our dataset. AdSpark provides a unified dataset and benchmark for future research, and we will release the dataset upon acceptance.

Since the natural language processing (NLP) community started to make large language models (LLMs), such as GPT-4, act as a critic to evaluate the quality of generated texts, most of them only train a critique generation model of a specific scale on specific datasets. We argue that a comprehensive investigation on the key factor of LLM-based evaluation models, such as scaling properties, is lacking, so that it is still inconclusive whether these models have potential to replace GPT-4's evaluation in practical scenarios. In this paper, we propose a new critique generation model called CritiqueLLM, which includes a dialogue-based prompting method for high-quality referenced / reference-free evaluation data. Experimental results show that our model can achieve comparable evaluation performance to GPT-4 especially in system-level correlations, and even outperform GPT-4 in 3 out of 8 tasks in a challenging reference-free setting. We conduct detailed analysis to show promising scaling properties of our model in the quality of generated critiques. We also demonstrate that our generated critiques can act as scalable feedback to directly improve the generation quality of LLMs.

Self-supervised 3D medical foundation models are increasingly used as general-purpose feature extractors, yet their sensitivity to MRI artifacts remains poorly understood. We present a controlled evaluation of representation robustness across five pretrained 3D encoders spanning different architectures, objectives, pretraining domains, and dataset scales. Using BraTS-Africa cases with four MRI sequences, we generate seven frequency- and image-domain artifacts at five predefined corruption settings. Robustness is assessed using linear centered kernel alignment (CKA), RankMe, and UMAP, complemented by an independent segmentation-consistency analysis. We find that robustness is strongly model- and artifact-dependent. 3DINO exhibits the most consistently stable representations, while BrainIAC is highly sensitive to several corruptions; NeuroVFM, BrainFM, and Neuro-SimCLR show intermediate but distinct artifact-specific profiles. Across many conditions, CKA decreases substantially while RankMe remains comparatively stable, indicating that artifacts often distort representation geometry without causing dimensional collapse. Segmentation consistency also degrades under corruption, particularly for ghosting and Rician noise, but aligns only partially with representation-level robustness. These findings show that larger-scale or domain-specific pretraining alone does not guarantee artifact invariance and motivate explicit robustness evaluation before deploying 3D foundation models in heterogeneous MRI settings.

Empathy is crucial in enabling natural interactions within spoken dialogue systems, allowing machines to recognize and respond appropriately to paralinguistic cues such as age, gender, and emotion. Recent advancements in end-to-end speech language models, which unify speech understanding and generation, provide promising solutions. However, several challenges persist, including an over-reliance on large-scale dialogue datasets, insufficient extraction of paralinguistic cues vital for conveying empathy, and the lack of empathy-specific datasets and evaluation frameworks. To address these issues, we introduce OSUM-EChat, an open-source, end-to-end spoken dialogue system designed to enhance empathetic interactions, particularly in resource-limited settings. OSUM-EChat introduces two key innovations: (1) a three-stage understanding-driven spoken dialogue training strategy that extends the capabilities of a large speech understanding model to spoken dialogue tasks, and (2) a linguistic-paralinguistic dual thinking mechanism that integrates paralinguistic understanding through a chain of thought with dialogue generation, enabling the system to produce more empathetic responses. This approach reduces reliance on large-scale dialogue datasets while maintaining high-quality empathetic interactions. Additionally, we introduce the EChat-200K dataset, a rich corpus of empathetic speech-to-speech dialogues, and the EChat-eval benchmark, a comprehensive framework for evaluating the empathetic capabilities of dialogue systems. Experimental results demonstrate that OSUM-EChat outperforms end-to-end spoken dialogue models regarding empathetic responsiveness, validating its effectiveness.

Designing Optical Character Recognition (OCR) systems for India requires balancing linguistic diversity, document heterogeneity, and deployment constraints. In this paper, we study two training strategies for building multilingual OCR systems with Vision-Language Models through the Chitrapathak series. We first follow a popular multimodal approach, pairing a generic vision encoder with a strong multilingual language model and training the system end-to-end for OCR. Alternatively, we explore fine-tuning an existing OCR model, despite not being trained for the target languages. Through extensive evaluation on multilingual Indic OCR benchmarks and deployment-oriented metrics, we find that the second strategy consistently achieves better accuracy-latency trade-offs. Chitrapathak-2 achieves 3-6x speedup over its predecessor with being state-of-the-art (SOTA) in Telugu (6.69 char ANLS) and second best in the rest. In addition, we present Parichay, an independent OCR model series designed specifically for 9 Indian government documents to extract structured key fields, achieving 89.8% Exact Match score with a faster inference. Together, these systems achieve SOTA performance and provide practical guidance for building production-scale OCR pipelines in the Indian context.

Videos are unique in their ability to capture actions which transcend multiple frames. Accordingly, for many years action recognition was the quintessential task for video understanding. Unfortunately, due to a lack of sufficiently diverse and challenging data, modern vision-language models (VLMs) are no longer evaluated on their action recognition capabilities. To revitalize action recognition in the era of VLMs, we advocate for a returned focus on domain-specific actions. To this end, we introduce VideoNet, a domain-specific action recognition benchmark covering 1,000 distinct actions from 37 domains. We begin with a multiple-choice evaluation setting, where the difference between closed and open models is stark: Gemini 3.1 Pro attains 69.9% accuracy while Qwen3-VL-8B gets a mere 45.0%. To understand why VLMs struggle on VideoNet, we relax the questions into a binary setting, where random chance is 50%. Still, Qwen achieves only 59.2% accuracy. Further relaxing the evaluation setup, we provide kin{1,2,3} in-context examples of the action. Some models excel in the few-shot setting, while others falter; Qwen improves +7.0%, while Gemini declines -4.8%. Notably, these gains fall short of the +13.6% improvement in non-expert humans when given few-shot examples. Finding that VLMs struggle to fully exploit in-context examples, we shift from test-time improvements to the training side. We collect the first large-scale training dataset for domain-specific actions, totaling nearly 500k video question-answer pairs. Fine-tuning a Molmo2-4B model on our data, we surpass all open-weight 8B models on the VideoNet benchmark.

Recent strides in Large Language Models (LLMs) have saturated many NLP benchmarks (even professional domain-specific ones), emphasizing the need for novel, more challenging novel ones to properly assess LLM capabilities. In this paper, we introduce a novel NLP benchmark that poses challenges to current LLMs across four key dimensions: processing long documents (up to 50K tokens), utilizing domain specific knowledge (embodied in legal texts), multilingual understanding (covering five languages), and multitasking (comprising legal document to document Information Retrieval, Court View Generation, Leading Decision Summarization, Citation Extraction, and eight challenging Text Classification tasks). Our benchmark comprises diverse legal NLP datasets from the Swiss legal system, allowing for a comprehensive study of the underlying Non-English, inherently multilingual, federal legal system. Despite recent advances, efficiently processing long documents for intense review/analysis tasks remains an open challenge for language models. Also, comprehensive, domain-specific benchmarks requiring high expertise to develop are rare, as are multilingual benchmarks. This scarcity underscores our contribution's value, considering most public models are trained predominantly on English corpora, while other languages remain understudied, particularly for practical domain-specific NLP tasks. Our benchmark allows for testing and advancing the state-of-the-art LLMs. As part of our study, we evaluate several pre-trained multilingual language models on our benchmark to establish strong baselines as a point of reference. Despite the large size of our datasets (tens to hundreds of thousands of examples), existing publicly available models struggle with most tasks, even after in-domain pretraining. We publish all resources (benchmark suite, pre-trained models, code) under a fully permissive open CC BY-SA license.

Large language models (LLMs), such as ChatGPT, are prone to generate hallucinations, i.e., content that conflicts with the source or cannot be verified by the factual knowledge. To understand what types of content and to which extent LLMs are apt to hallucinate, we introduce the Hallucination Evaluation benchmark for Large Language Models (HaluEval), a large collection of generated and human-annotated hallucinated samples for evaluating the performance of LLMs in recognizing hallucination. To generate these samples, we propose a ChatGPT-based two-step framework, i.e., sampling-then-filtering. Besides, we also hire some human labelers to annotate the hallucinations in ChatGPT responses. The empirical results suggest that ChatGPT is likely to generate hallucinated content in specific topics by fabricating unverifiable information (i.e., about 19.5% responses). Moreover, existing LLMs face great challenges in recognizing the hallucinations in texts. However, our experiments also prove that providing external knowledge or adding reasoning steps can help LLMs recognize hallucinations. Our benchmark can be accessed at https://github.com/RUCAIBox/HaluEval.

Medical question answering (QA) benchmarks often focus on multiple-choice or fact-based tasks, leaving open-ended answers to real patient questions underexplored. This gap is particularly critical in mental health, where patient questions often mix symptoms, treatment concerns, and emotional needs, requiring answers that balance clinical caution with contextual sensitivity. We present CounselBench, a large-scale benchmark developed with 100 mental health professionals to evaluate and stress-test large language models (LLMs) in realistic help-seeking scenarios. The first component, CounselBench-EVAL, contains 2,000 expert evaluations of answers from GPT-4, LLaMA 3, Gemini, and online human therapists on patient questions from the public forum CounselChat. Each answer is rated across six clinically grounded dimensions, with span-level annotations and written rationales. Expert evaluations show that while LLMs achieve high scores on several dimensions, they also exhibit recurring issues, including unconstructive feedback, overgeneralization, and limited personalization or relevance. Responses were frequently flagged for safety risks, most notably unauthorized medical advice. Follow-up experiments show that LLM judges systematically overrate model responses and overlook safety concerns identified by human experts. To probe failure modes more directly, we construct CounselBench-Adv, an adversarial dataset of 120 expert-authored mental health questions designed to trigger specific model issues. Expert evaluation of 1,080 responses from nine LLMs reveals consistent, model-specific failure patterns. Together, CounselBench establishes a clinically grounded framework for benchmarking LLMs in mental health QA.

During the development of large language models (LLMs), the scale and quality of the pre-training data play a crucial role in shaping LLMs' capabilities. To accelerate the research of LLMs, several large-scale datasets, such as C4 [1], Pile [2], RefinedWeb [3] and WanJuan [4], have been released to the public. However, most of the released corpus focus mainly on English, and there is still lack of complete tool-chain for extracting clean texts from web data. Furthermore, fine-grained information of the corpus, e.g. the quality of each text, is missing. To address these challenges, we propose in this paper a new complete tool-chain EvalWeb to extract Chinese clean texts from noisy web data. First, similar to previous work, manually crafted rules are employed to discard explicit noisy texts from the raw crawled web contents. Second, a well-designed evaluation model is leveraged to assess the remaining relatively clean data, and each text is assigned a specific quality score. Finally, we can easily utilize an appropriate threshold to select the high-quality pre-training data for Chinese. Using our proposed approach, we release the largest and latest large-scale high-quality Chinese web text ChineseWebText, which consists of 1.42 TB and each text is associated with a quality score, facilitating the LLM researchers to choose the data according to the desired quality thresholds. We also release a much cleaner subset of 600 GB Chinese data with the quality exceeding 90%.

Automated radiology report generation for multi-sequence brain MRI holds strong promise for alleviating radiologists' workload, yet it remains a demanding problem: brain MRI studies are inherently volumetric, each pulse sequence carries complementary diagnostic information, and annotated data are critically scarce for low-resource clinical languages. Existing brain MRI-report resources are still dominated by English reports, translated clinical text, or datasets adapted from task-specific imaging benchmarks, leaving routine low-resource clinical reporting and heterogeneous multi-sequence MRI protocols insufficiently represented. To bridge this gap, we introduce ViBMR, to the best of our knowledge, the first large-scale multi-sequence brain MRI dataset paired with routine Vietnamese clinical reports, comprising 3,649 patients and 11,875 MRI series across T1, T2, FLAIR, and contrast-enhanced T1 sequences. Building on ViBMR, we establish a comprehensive evaluation setup for brain MRI report generation, covering representative medical vision-language models, multi-sequence inputs, and feature aggregation strategies. The results provide key insights into dataset characteristics and model behavior, while demonstrating the necessity of native low-resource clinical datasets for reliable Vietnamese-origin brain MRI report generation. Our dataset and baseline models will be released for non-commercial research.

We present SODA: the first publicly available, million-scale high-quality social dialogue dataset. Using SODA, we train COSMO: a generalizable conversation agent outperforming previous best-performing agents on both in- and out-of-domain datasets. In contrast to most existing crowdsourced, small-scale dialogue corpora, we distill 1.5M socially-grounded dialogues from a pre-trained language model (InstructGPT; Ouyang et al., 2022). Dialogues are distilled by contextualizing social commonsense knowledge from a knowledge graph (Atomic10x; West et al., 2022). Human evaluation shows that dialogues in SODA are more consistent, specific, and (surprisingly) natural than prior human-authored datasets - e.g., DailyDialog (Li et al., 2017), BlendedSkillTalk (Smith et al., 2020). In addition, extensive evaluations show that COSMO is significantly more natural and consistent on unseen datasets than best-performing dialogue models - e.g., GODEL (Peng et al., 2022), BlenderBot (Roller et al., 2021), DialoGPT (Zhang et al., 2020). Furthermore, it is sometimes even preferred to the original human-written gold responses. We make our data, models, and code public.

Visual information is central to conversation: body gestures and physical behaviour, for example, contribute to meaning that transcends words alone. To date, however, most neural conversational models are limited to just text. We introduce CHAMPAGNE, a generative model of conversations that can account for visual contexts. To train CHAMPAGNE, we collect and release YTD-18M, a large-scale corpus of 18M video-based dialogues. YTD-18M is constructed from web videos: crucial to our data collection pipeline is a pretrained language model that converts error-prone automatic transcripts to a cleaner dialogue format while maintaining meaning. Human evaluation reveals that YTD-18M is more sensible and specific than prior resources (MMDialog, 1M dialogues), while maintaining visual-groundedness. Experiments demonstrate that 1) CHAMPAGNE learns to conduct conversation from YTD-18M; and 2) when fine-tuned, it achieves state-of-the-art results on four vision-language tasks focused on real-world conversations. We release data, models, and code.

Recent advances in multi-modal large language models (MLLMs) have demonstrated strong performance across various domains; however, their ability to comprehend driving scenes remains less proven. The complexity of driving scenarios, which includes multi-view information, poses significant challenges for existing MLLMs. In this paper, we introduce NuPlanQA-Eval, a multi-view, multi-modal evaluation benchmark for driving scene understanding. To further support generalization to multi-view driving scenarios, we also propose NuPlanQA-1M, a large-scale dataset comprising 1M real-world visual question-answering (VQA) pairs. For context-aware analysis of traffic scenes, we categorize our dataset into nine subtasks across three core skills: Road Environment Perception, Spatial Relations Recognition, and Ego-Centric Reasoning. Furthermore, we present BEV-LLM, integrating Bird's-Eye-View (BEV) features from multi-view images into MLLMs. Our evaluation results reveal key challenges that existing MLLMs face in driving scene-specific perception and spatial reasoning from ego-centric perspectives. In contrast, BEV-LLM demonstrates remarkable adaptability to this domain, outperforming other models in six of the nine subtasks. These findings highlight how BEV integration enhances multi-view MLLMs while also identifying key areas that require further refinement for effective adaptation to driving scenes. To facilitate further research, we publicly release NuPlanQA at https://github.com/sungyeonparkk/NuPlanQA.

The advancement of open-source text-to-image (T2I) models has been hindered by the absence of large-scale, reasoning-focused datasets and comprehensive evaluation benchmarks, resulting in a performance gap compared to leading closed-source systems. To address this challenge, We introduce FLUX-Reason-6M and PRISM-Bench (Precise and Robust Image Synthesis Measurement Benchmark). FLUX-Reason-6M is a massive dataset consisting of 6 million high-quality FLUX-generated images and 20 million bilingual (English and Chinese) descriptions specifically designed to teach complex reasoning. The image are organized according to six key characteristics: Imagination, Entity, Text rendering, Style, Affection, and Composition, and design explicit Generation Chain-of-Thought (GCoT) to provide detailed breakdowns of image generation steps. The whole data curation takes 15,000 A100 GPU days, providing the community with a resource previously unattainable outside of large industrial labs. PRISM-Bench offers a novel evaluation standard with seven distinct tracks, including a formidable Long Text challenge using GCoT. Through carefully designed prompts, it utilizes advanced vision-language models for nuanced human-aligned assessment of prompt-image alignment and image aesthetics. Our extensive evaluation of 19 leading models on PRISM-Bench reveals critical performance gaps and highlights specific areas requiring improvement. Our dataset, benchmark, and evaluation code are released to catalyze the next wave of reasoning-oriented T2I generation. Project page: https://flux-reason-6m.github.io/ .

Data-driven approaches like deep learning are rapidly advancing planetary science, particularly in Mars exploration. Despite recent progress, most existing benchmarks remain confined to closed-set supervised visual tasks and do not support text-guided retrieval for geospatial discovery. We introduce MarsRetrieval, a retrieval benchmark for evaluating vision-language models for Martian geospatial discovery. MarsRetrieval includes three tasks: (1) paired image-text retrieval, (2) landform retrieval, and (3) global geo-localization, covering multiple spatial scales and diverse geomorphic origins. We propose a unified retrieval-centric protocol to benchmark multimodal embedding architectures, including contrastive dual-tower encoders and generative vision-language models. Our evaluation shows MarsRetrieval is challenging: even strong foundation models often fail to capture domain-specific geomorphic distinctions. We further show that domain-specific fine-tuning is critical for generalizable geospatial discovery in planetary settings. Our code is available at https://github.com/ml-stat-Sustech/MarsRetrieval

Reference-to-video (R2V) generation is evolving toward increasingly general and versatile reference control, giving rise to the emerging paradigm of omni R2V generation. However, existing benchmarks fall short of these emerging capabilities: their test cases cover limited reference types and compositions, and their evaluation protocols largely assess holistic reference consistency, overlooking whether reference factors are properly preserved, disentangled, and routed. Meanwhile, the high cost of constructing omni R2V training data makes suitable training resources scarce. To address these gaps, we introduce OmniVBench and the Omni-R2V Dataset for evaluating and training omni R2V models. OmniVBench expands R2V evaluation across broader reference types, fine-grained control tasks, and richer reference compositions, covering 7 task families and 18 fine-grained tasks spanning content, motion, style, structure, narrative, and multi-reference settings. We introduce factor-grounded evaluation with 12,172 case-specific checklist items, assessing whether intended reference factors are faithfully preserved, correctly disentangled and bound to their targets, and properly realized according to the instruction. We further introduce the Omni-R2V Dataset, bringing industrial-grade training resources for diverse R2V tasks to the broader research community. Drawing primarily on a large-scale corpus of professional video footage, it comprises 340K processed training samples spanning diverse reference types and multi-reference compositions. We develop task-specific pipelines for reference-target pair construction, offering a practical and scalable recipe for omni R2V data construction. Extensive evaluation of advanced open- and closed-source R2V models reveals clear performance gaps across task families and evaluation dimensions on OmniVBench, highlighting remaining limitations of current R2V models.

Large-scale agentic training and evaluation with large language models (LLMs) rely on isolated, stateful execution environments in which models inspect repositories, invoke tools, execute commands, and interact with task-specific services. These workloads create sandboxes in large bursts, span heterogeneous functionality and isolation requirements, retain state across long interactions, and draw from large image corpora with limited reuse. Supporting them therefore requires an elastic execution platform rather than a single sandbox runtime. This report presents DeepSeek Elastic Compute (DSec), a production sandbox platform that exposes FnCall, container, microVM, and full-VM sandbox backends through a unified SDK. DSec coordinates placement and lifecycle management across the cluster, composes environments from independently versioned layers, combines memory sharing, reclamation, and CPU scheduling for high-density execution, and loads image data on demand from Fire-Flyer File System (3FS), a cluster-wide distributed filesystem. DSec is co-designed with the reinforcement learning (RL) framework, decouples stateful rollout execution from preemptible GPU training, coordinates sandbox lifecycle with training to preserve rollout state while reclaiming idle resources, and mitigates agent misbehavior such as reward hacking. A single production-scale unit of DSec spans around 160 nodes, serving about 3 million sandboxes per day; in production, it supports over 380,000 concurrent sandboxes and sustains over 5,000 sandbox creations per second. Our evaluation and deployment experience show that these mechanisms reduce environment setup and image-distribution overhead, improve memory efficiency, and preserve latency-sensitive performance under high-density overcommit.

The ability to solve problems is a hallmark of intelligence and has been an enduring goal in AI. AI systems that can create programs as solutions to problems or assist developers in writing programs can increase productivity and make programming more accessible. Recently, pre-trained large language models have shown impressive abilities in generating new codes from natural language descriptions, repairing buggy codes, translating codes between languages, and retrieving relevant code segments. However, the evaluation of these models has often been performed in a scattered way on only one or two specific tasks, in a few languages, at a partial granularity (e.g., function) level and in many cases without proper training data. Even more concerning is that in most cases the evaluation of generated codes has been done in terms of mere lexical overlap rather than actual execution whereas semantic similarity (or equivalence) of two code segments depends only on their ``execution similarity'', i.e., being able to get the same output for a given input.

De-identification of clinical text is a prerequisite for the secondary use of electronic health records. Existing public benchmarks such as the i2b2 2006 and 2014 corpora are over a decade old and lack the semantic and demographic diversity of modern clinical narratives. Large Language Models (LLMs) reach state-of-the-art zero-shot extraction, but their use at enterprise scale is limited by computational cost and by hospital data governance that restricts sending Protected Health Information (PHI) to cloud APIs. We introduce SHIELD (Synthetic Human-annotated Identifier-replaced Entries for Learning and De-identification), a diverse clinical note dataset of 1,381 notes with 10,229 gold-standard PHI spans across 9 categories, built with set-cover diversity sampling across demographic and document-type strata and human-in-the-loop adjudication. We evaluate four LLMs (two proprietary, two open-weight) to establish a performance ceiling on SHIELD, then show that a teacher-student distillation framework transfers these capabilities into locally deployable Small Language Models. Our best distilled model reaches micro-averaged span-level precision of 0.89 and recall of 0.88 while running on standard workstation hardware. It trails its cloud teacher on per-category recall (0.90 vs. 0.81 macro-averaged) but remains competitive given its lower cost and on-premise deployability. Cross-dataset evaluation shows that diversity-trained models generalize well on universal structured PHI categories, while institution-specific entities remain hard to transfer in both directions, which suggests pairing broad-coverage models with specialized models for high-volume, semi-structured note types. We publicly release the SHIELD dataset and the distilled DeBERTa v3 model to provide an accurate, cost-effective de-identification pipeline deployable entirely behind institutional firewalls.

Embodied foundation models are gaining increasing attention for their zero-shot generalization, scalability, and adaptability to new tasks through few-shot post-training. However, existing models rely heavily on real-world data, which is costly and labor-intensive to collect. Synthetic data offers a cost-effective alternative, yet its potential remains largely underexplored. To bridge this gap, we explore the feasibility of training Vision-Language-Action models entirely with large-scale synthetic action data. We curate SynGrasp-1B, a billion-frame robotic grasping dataset generated in simulation with photorealistic rendering and extensive domain randomization. Building on this, we present GraspVLA, a VLA model pretrained on large-scale synthetic action data as a foundational model for grasping tasks. GraspVLA integrates autoregressive perception tasks and flow-matching-based action generation into a unified Chain-of-Thought process, enabling joint training on synthetic action data and Internet semantics data. This design helps mitigate sim-to-real gaps and facilitates the transfer of learned actions to a broader range of Internet-covered objects, achieving open-vocabulary generalization in grasping. Extensive evaluations across real-world and simulation benchmarks demonstrate GraspVLA's advanced zero-shot generalizability and few-shot adaptability to specific human preferences. We will release SynGrasp-1B dataset and pre-trained weights to benefit the community.

We present IndoBERTweet, the first large-scale pretrained model for Indonesian Twitter that is trained by extending a monolingually-trained Indonesian BERT model with additive domain-specific vocabulary. We focus in particular on efficient model adaptation under vocabulary mismatch, and benchmark different ways of initializing the BERT embedding layer for new word types. We find that initializing with the average BERT subword embedding makes pretraining five times faster, and is more effective than proposed methods for vocabulary adaptation in terms of extrinsic evaluation over seven Twitter-based datasets.

With tremendous efforts on developing effective e-commerce models, conventional e-commerce models show limited success in generalist e-commerce modeling, and suffer from unsatisfactory performance on new users and new products - a typical out-of-domain generalization challenge. Meanwhile, large language models (LLMs) demonstrate outstanding performance in generalist modeling and out-of-domain generalizability in many fields. Toward fully unleashing their power for e-commerce, in this paper, we construct ECInstruct, the first open-sourced, large-scale, and high-quality benchmark instruction dataset for e-commerce. Leveraging ECInstruct, we develop eCeLLM, a series of e-commerce LLMs, by instruction-tuning general-purpose LLMs. Our comprehensive experiments and evaluation demonstrate that eCeLLM models substantially outperform baseline models, including the most advanced GPT-4, and the state-of-the-art task-specific models in in-domain evaluation. Moreover, eCeLLM exhibits excellent generalizability to out-of-domain settings, including unseen products and unseen instructions, highlighting its superiority as a generalist e-commerce model. Both the ECInstruct dataset and the eCeLLM models show great potential in empowering versatile and effective LLMs for e-commerce. ECInstruct and eCeLLM models are publicly accessible through https://ninglab.github.io/eCeLLM.

End-to-end (E2E) spoken dialogue systems are increasingly replacing cascaded pipelines for voice-based human-AI interaction, processing raw audio directly without intermediate transcription. Existing benchmarks primarily evaluate these models on synthetic speech and single-turn tasks, leaving realistic multi-turn conversational ability underexplored. We introduce Audio MultiChallenge, an open-source benchmark to evaluate E2E spoken dialogue systems under natural multi-turn interaction patterns. Building on the text-based MultiChallenge framework, which evaluates Inference Memory, Instruction Retention, and Self Coherence, we introduce a new axis Voice Editing that tests robustness to mid-utterance speech repairs and backtracking. We further augment each axis to the audio modality, such as introducing Audio-Cue challenges for Inference Memory that require recalling ambient sounds and paralinguistic signals beyond semantic content. We curate 452 conversations from 47 speakers with 1,712 instance-specific rubrics through a hybrid audio-native agentic and human-in-the-loop pipeline that exposes model failures at scale while preserving natural disfluencies found in unscripted human speech. Our evaluation of proprietary and open-source models reveals that even frontier models struggle on our benchmark, with Gemini 3 Pro Preview (Thinking), our highest-performing model achieving a 54.65% pass rate. Error analysis shows that models fail most often on our new axes and that Self Coherence degrades with longer audio context. These failures reflect difficulty of tracking edits, audio cues, and long-range context in natural spoken dialogue. Audio MultiChallenge provides a reproducible testbed to quantify them and drive improvements in audio-native multi-turn interaction capability.

We present L3Cube-IndicQuest v2, a large-scale gold-standard multilingual question-answering benchmark for evaluating the India-specific factual knowledge of Large Language Models (LLMs). The benchmark comprises 3,471 curriculum-grounded English question--answer pairs spanning nine domains, curated from educational curricula, competitive examination materials, and domain-specific reference books. We introduce a practical hybrid construction strategy that combines context-grounded LLM-based question generation and validation with semantic deduplication and human verification, enabling scalable creation of benchmark data while preserving annotation quality. The benchmark is translated into 19 Indic languages, yielding a publicly released multilingual dataset of 69,420 question--answer pairs across 20 languages. We evaluate six LLMs under three protocols: LLM-as-a-judge and two deterministic lexical criteria, exact-substring and word-overlap matching. All three produce almost the same model ranking, showing that the results do not depend on the choice of judge. The frontier commercial model leads by a wide margin, and among open-weight models Gemma4 31B outperforms the Indic-specialised Sarvam 30B in every evaluated Indic language.

Fine-grained and instance-level recognition methods are commonly trained and evaluated on specific domains, in a model per domain scenario. Such an approach, however, is impractical in real large-scale applications. In this work, we address the problem of universal image embedding, where a single universal model is trained and used in multiple domains. First, we leverage existing domain-specific datasets to carefully construct a new large-scale public benchmark for the evaluation of universal image embeddings, with 241k query images, 1.4M index images and 2.8M training images across 8 different domains and 349k classes. We define suitable metrics, training and evaluation protocols to foster future research in this area. Second, we provide a comprehensive experimental evaluation on the new dataset, demonstrating that existing approaches and simplistic extensions lead to worse performance than an assembly of models trained for each domain separately. Finally, we conducted a public research competition on this topic, leveraging industrial datasets, which attracted the participation of more than 1k teams worldwide. This exercise generated many interesting research ideas and findings which we present in detail. Project webpage: https://cmp.felk.cvut.cz/univ_emb/

In recent years, there has been a proliferation of spatiotemporal foundation models in different scientific disciplines. While promising, these models are often domain-specific and are only assessed within the particular applications for which they are designed. Given that many tasks can be represented as video modeling problems, video foundation models (ViFMs) hold considerable promise as general-purpose domain-agnostic approaches. However, it is not known whether the knowledge acquired on large-scale but potentially out-of-domain data can be effectively transferred across diverse scientific disciplines, and if a single, pretrained ViFM can be competitive with domain-specific baselines. To address this, we introduce SciVid, a comprehensive benchmark comprising five *Sci*entific *Vid*eo tasks, across medical computer vision, animal behavior, and weather forecasting. We adapt six leading ViFMs to SciVid using simple trainable readout modules, establishing strong baselines and demonstrating the potential for effective transfer learning. Specifically, we show that state-of-the-art results can be obtained in several applications by leveraging the general-purpose representations from ViFM backbones. Furthermore, our results reveal the limitations of existing ViFMs, and highlight opportunities for the development of generalizable models for high-impact scientific applications. We release our code at https://github.com/google-deepmind/scivid to facilitate further research in the development of ViFMs.

The goal of building a benchmark (suite of datasets) is to provide a unified protocol for fair evaluation and thus facilitate the evolution of a specific area. Nonetheless, we point out that existing protocols of action recognition could yield partial evaluations due to several limitations. To comprehensively probe the effectiveness of spatiotemporal representation learning, we introduce BEAR, a new BEnchmark on video Action Recognition. BEAR is a collection of 18 video datasets grouped into 5 categories (anomaly, gesture, daily, sports, and instructional), which covers a diverse set of real-world applications. With BEAR, we thoroughly evaluate 6 common spatiotemporal models pre-trained by both supervised and self-supervised learning. We also report transfer performance via standard finetuning, few-shot finetuning, and unsupervised domain adaptation. Our observation suggests that current state-of-the-art cannot solidly guarantee high performance on datasets close to real-world applications, and we hope BEAR can serve as a fair and challenging evaluation benchmark to gain insights on building next-generation spatiotemporal learners. Our dataset, code, and models are released at: https://github.com/AndongDeng/BEAR

Existing Indic ASR benchmarks often use scripted, clean speech and leaderboard driven evaluation that encourages dataset specific overfitting. In addition, strict single reference WER penalizes natural spelling variation in Indian languages, including non standardized spellings of code-mixed English origin words. To address these limitations, we introduce Voice of India, a closed source benchmark built from unscripted telephonic conversations covering 15 major Indian languages across 139 regional clusters. The dataset contains 306230 utterances, totaling 536 hours of speech from 36691 speakers with transcripts accounting for spelling variations. We also analyze performance geographically at the district level, revealing disparities. Finally, we provide detailed analysis across factors such as audio quality, speaking rate, gender, and device type, highlighting where current ASR systems struggle and offering insights for improving real world Indic ASR systems.

As Large Language Models (LLMs) saturate elementary benchmarks, the research frontier has shifted from generation to the reliability of automated evaluation. We demonstrate that standard "LLM-as-a-Judge" protocols suffer from a systematic Alignment Gap when applied to upper-undergraduate to early graduate level mathematics. To quantify this, we introduce QEDBench, the first large-scale dual-rubric alignment benchmark to systematically measure alignment with human experts on university-level math proofs by contrasting course-specific rubrics against expert common knowledge criteria. By deploying a dual-evaluation matrix (7 judges x 5 solvers) against 1,000+ hours of human evaluation, we reveal that certain frontier evaluators like Claude Opus 4.5, DeepSeek-V3, Qwen 2.5 Max, and Llama 4 Maverick exhibit significant positive bias (up to +0.18, +0.20, +0.30, +0.36 mean score inflation, respectively). Furthermore, we uncover a critical reasoning gap in the discrete domain: while Gemini 3.0 Pro achieves state-of-the-art performance (0.91 average human evaluation score), other reasoning models like GPT-5 Pro and Claude Sonnet 4.5 see their performance significantly degrade in discrete domains. Specifically, their average human evaluation scores drop to 0.72 and 0.63 in Discrete Math, and to 0.74 and 0.50 in Graph Theory. In addition to these research results, we also release QEDBench as a public benchmark for evaluating and improving AI judges. Our benchmark is publicly published at https://github.com/qqliu/Yale-QEDBench.

Prompt design is a primary control interface for large language models (LLMs), yet standard evaluations largely reduce performance to answer correctness, obscuring why a prompt succeeds or fails and providing little actionable guidance. We propose PEEM (Prompt Engineering Evaluation Metrics), a unified framework for joint and interpretable evaluation of both prompts and responses. PEEM defines a structured rubric with 9 axes: 3 prompt criteria (clarity/structure, linguistic quality, fairness) and 6 response criteria (accuracy, coherence, relevance, objectivity, clarity, conciseness), and uses an LLM-based evaluator to output (i) scalar scores on a 1-5 Likert scale and (ii) criterion-specific natural-language rationales grounded in the rubric. Across 7 benchmarks and 5 task models, PEEM's accuracy axis strongly aligns with conventional accuracy while preserving model rankings (aggregate Spearman rho about 0.97, Pearson r about 0.94, p < 0.001). A multi-evaluator study with four models shows consistent relative judgments (pairwise rho = 0.68-0.85), supporting evaluator-agnostic deployment. Beyond alignment, PEEM captures complementary linguistic failure modes and remains informative under prompt perturbations: prompt-quality trends track downstream accuracy under iterative rewrites, semantic adversarial manipulations induce clear score degradation, and meaning-preserving paraphrases yield high stability (robustness rate about 76.7-80.6%). Finally, using only PEEM scores and rationales as feedback, a zero-shot prompt rewriting loop improves downstream accuracy by up to 11.7 points, outperforming supervised and RL-based prompt-optimization baselines. Overall, PEEM provides a reproducible, criterion-driven protocol that links prompt formulation to response behavior and enables systematic diagnosis and optimization of LLM interactions.

Multilingual benchmarks are central to evaluating large language models (LLMs) across languages, but they suffer from three issues: exhaustive evaluation scales linearly with the number of languages, automatic translation introduces errors that are easily missed at scale, and some items conflate general and culture-specific knowledge. We address all three with a unified statistical framework, Multilingual-IRT, which extends Item Response Theory with per-language difficulty deviations, split discriminability separating content from language effects, and per-language ability residuals. Fitting Multilingual-IRT on 25 LLMs across 29 languages of MMLU-Pro-X, we show that its fitted parameters support three practical applications: predicting unobserved (item, LLM, language) instances with 11-16% lower binary cross-entropy than the strongest accuracy-based baseline, surfacing candidate translation errors distributed across all 28 non-English languages, whereas accuracy-based baselines concentrate detections in a few languages, and recovering culture-specific items that accuracy-based baselines miss.

Evaluating embodied robot foundation models remains a critical bottleneck; unlike large language models efficiently assessed via digital benchmarks, robotic policies require slow, costly real-world rollouts limited by hardware and human supervision, which has driven interest in world models as surrogate policy evaluators, yet the key properties that make a world model reliable for policy assessment remain poorly understood. This work presents a systematic study of world models for robotic policy evaluation and introduces WMBench, a benchmark constructed from real-robot teleoperation data and matched policy rollouts covering diverse manipulation tasks to enable controlled comparisons across model families, action encodings, rollout horizons, and evaluation metrics. Using WMBench, we analyze 7 video world models, 4 action representation schemes, and over 324,000 simulated policy rollouts paired with real robot executions, further enriching our analysis with large-scale community submissions from the CVPR 2026 GigaBrain Challenge, curated synthetic trajectories, and a training videos spanning more than 12,000 hours. Our experiments deliver three core insights: evaluator quality is dominated by long-horizon, action-faithful rollout consistency rather than short-term visual realism; pretraining gains stem not only from data scale but from balancing general world knowledge with robot-specific controllability; and architectural choices including action encoding, memory design, and evaluator-focused post-training strongly determine alignment with real-world robot behavior. Drawing on these results, we derive a practical design roadmap and realize it in GigaWorld-1, a world model specially optimized for policy evaluation, and we fully release our code, models, datasets, and toolkits to advance scalable evaluation research for embodied foundation models.

Differential privacy (DP) provides formal protection for sensitive data but typically incurs substantial losses in diagnostic performance. Model initialization has emerged as a critical factor in mitigating this degradation, yet the role of modern self-supervised learning under full-model DP remains poorly understood. Here, we present a large-scale evaluation of initialization strategies for differentially private medical image analysis, using chest radiograph classification as a representative benchmark with more than 800,000 images. Using state-of-the-art ConvNeXt models trained with DP-SGD across realistic privacy regimes, we compare non-domain-specific supervised ImageNet initialization, non-domain-specific self-supervised DINOv3 initialization, and domain-specific supervised pretraining on MIMIC-CXR, the largest publicly available chest radiograph dataset. Evaluations are conducted across five external datasets spanning diverse institutions and acquisition settings. We show that DINOv3 initialization consistently improves diagnostic utility relative to ImageNet initialization under DP, but remains inferior to domain-specific supervised pretraining, which achieves performance closest to non-private baselines. We further demonstrate that initialization choice strongly influences demographic fairness, cross-dataset generalization, and robustness to data scale and model capacity under privacy constraints. The results establish initialization strategy as a central determinant of utility, fairness, and generalization in differentially private medical imaging.

Recent large language models (LLMs) such as ChatGPT and LLaMA have shown great promise in many AI applications. However, their performance on medical tasks is suboptimal and can be improved by training on extensive domain-specific datasets. This study introduces Me LLaMA, a medical LLM family that includes foundation models - Me LLaMA 13/70B, along with their chat-enhanced versions - Me LLaMA 13/70B-chat, developed through continual pre-training and instruction tuning of LLaMA2 using large medical datasets. Our domain-specific data suite for training and evaluation includes a large-scale, continual pre-training dataset with 129B tokens, an instruction tuning dataset with 214k samples, and a new medical evaluation benchmark (MIBE) across six tasks with 12 datasets. Our extensive evaluation using the MIBE shows that Me LLaMA models achieve overall better performance than existing open-source medical LLMs in zero-shot, few-shot and supervised learning abilities. Their zero-shot performance is comparable with ChatGPT across 7 out of 8 datasets, with a slight variance of within 3%, and yet falls short when compared to GPT-4. In addition, we investigated the catastrophic forgetting problem, and our results show that Me LLaMA models outperform other open-source medical LLMs in mitigating this issue. Me LLaMA is one of the largest open-source medical foundation LLMs that use both biomedical and clinical data. It exhibits superior performance across both general and medical tasks compared to other open-source medical LLMs, rendering it an attractive choice for medical AI applications. We release our models, datasets, and evaluation scripts at: https://github.com/BIDS-Xu-Lab/Me-LLaMA.

Public health reasoning requires population level inference grounded in scientific evidence, expert consensus, and safety constraints. However, it remains underexplored as a structured machine learning problem with limited supervised signals and benchmarks. We introduce GlobalHealthAtlas, a large scale multilingual dataset of 280,210 instances spanning 15 public health domains and 17 languages. We further propose a large language model (LLM) assisted construction and quality control pipeline with retrieval, deduplication, evidence grounding checks, and label validation to improve consistency at scale. Finally, we present a domain aligned evaluator distilled from high confidence judgments of diverse LLMs to assess outputs along six dimensions: Accuracy, Reasoning, Completeness, Consensus Alignment, Terminology Norms, and Insightfulness. Together, these contributions enable reproducible training and evaluation of LLMs for safety critical public health reasoning beyond conventional QA benchmarks. We publicly release project codebase, evaluator, and model at:: https://github.com/Jan8217/GlobalHealthAtlas, https://huggingface.co/aerovane0/GlobalHealthAtlas_Public_Evaluator and https://huggingface.co/aerovane0/GlobalHealthAtlas_Public_Model

The increasing adoption of agentic workflows across diverse domains brings a critical need to scalably and systematically evaluate the complex traces these systems generate. Current evaluation methods depend on manual, domain-specific human analysis of lengthy workflow traces - an approach that does not scale with the growing complexity and volume of agentic outputs. Error analysis in these settings is further complicated by the interplay of external tool outputs and language model reasoning, making it more challenging than traditional software debugging. In this work, we (1) articulate the need for robust and dynamic evaluation methods for agentic workflow traces, (2) introduce a formal taxonomy of error types encountered in agentic systems, and (3) present a set of 148 large human-annotated traces (TRAIL) constructed using this taxonomy and grounded in established agentic benchmarks. To ensure ecological validity, we curate traces from both single and multi-agent systems, focusing on real-world applications such as software engineering and open-world information retrieval. Our evaluations reveal that modern long context LLMs perform poorly at trace debugging, with the best Gemini-2.5-pro model scoring a mere 11% on TRAIL. Our dataset and code are made publicly available to support and accelerate future research in scalable evaluation for agentic workflows.

The rapid advancement of pre-trained language models (PLMs) has demonstrated promising results for various code-related tasks. However, their effectiveness in detecting real-world vulnerabilities remains a critical challenge. % for the security community. While existing empirical studies evaluate PLMs for vulnerability detection (VD), their inadequate consideration in data preparation, evaluation setups, and experimental settings undermines the accuracy and comprehensiveness of evaluations. This paper introduces RevisitVD, an extensive evaluation of 17 PLMs spanning smaller code-specific PLMs and large-scale PLMs using newly constructed datasets. Specifically, we compare the performance of PLMs under both fine-tuning and prompt engineering, assess their effectiveness and generalizability across various training and testing settings, and analyze their robustness against code normalization, abstraction, and semantic-preserving transformations. Our findings reveal that, for VD tasks, PLMs incorporating pre-training tasks designed to capture the syntactic and semantic patterns of code outperform both general-purpose PLMs and those solely pre-trained or fine-tuned on large code corpora. However, these models face notable challenges in real-world scenarios, such as difficulties in detecting vulnerabilities with complex dependencies, handling perturbations introduced by code normalization and abstraction, and identifying semantic-preserving vulnerable code transformations. Also, the truncation caused by the limited context windows of PLMs can lead to a non-negligible amount of labeling errors. This study underscores the importance of thorough evaluations of model performance in practical scenarios and outlines future directions to help enhance the effectiveness of PLMs for realistic VD applications.

Deobfuscating binary code remains a fundamental challenge in reverse engineering, as obfuscation is widely used to hinder analysis and conceal program logic. Although large language models (LLMs) have shown promise in recovering semantics from obfuscated binaries, a systematic evaluation of their effectiveness is still lacking. In this work, we present BinDeObfBench, the first comprehensive benchmark for assessing LLM-based binary deobfuscation across diverse transformations spanning pre-compilation, compile-time, and post-compilation stages. Our evaluation shows that deobfuscation performance depends more on reasoning capability and domain expertise than on model scale, and that task-specific supervised fine-tuning consistently outperforms broad domain pre-training. Reasoning models can maintain robustness under severe obfuscation, generalize across different instruction set architectures (ISAs) and optimization levels. In-context learning benefits standard models but yields limited gains for reasoning models. Overall, our study highlights the importance of task-specific fine-tuning and reasoning-driven strategies, and positions BinDeObfBench as a basis for future work in binary deobfuscation.

A variety of text-guided image editing models have been proposed recently. However, there is no widely-accepted standard evaluation method mainly due to the subjective nature of the task, letting researchers rely on manual user study. To address this, we introduce a novel Human-Aligned benchmark for Text-guided Image Editing (HATIE). Providing a large-scale benchmark set covering a wide range of editing tasks, it allows reliable evaluation, not limited to specific easy-to-evaluate cases. Also, HATIE provides a fully-automated and omnidirectional evaluation pipeline. Particularly, we combine multiple scores measuring various aspects of editing so as to align with human perception. We empirically verify that the evaluation of HATIE is indeed human-aligned in various aspects, and provide benchmark results on several state-of-the-art models to provide deeper insights on their performance.

Despite growing interest in domain-specific benchmarking of large language models (LLMs) and agents, current evaluations remain limited to static, small-scale datasets, especially in high-stakes tasks like network operations that demand reliability for deployments. We present NetPress, an automated benchmark generation framework for evaluating LLM agents in network applications. NetPress introduces a unified abstraction with state and action, enabling dynamic generation of diverse query sets along with corresponding ground truths. At runtime, users can specify benchmark configurations to generate millions of queries on the fly. In addition to dynamic benchmark construction, NetPress integrates with network emulators to provide realistic environment feedback, supporting comprehensive evaluation across correctness, safety, and latency. We instantiate NetPress on three representative applications, revealing interesting fine-grained differences in agent behavior that static, correctness-only benchmarks often miss. NetPress moves LLM evaluation toward realistic, scalable testing in infrastructure-centric domains, helping close the gap between benchmark performance and real-world deployment readiness. Code is available at https://github.com/Froot-NetSys/NetPress.

Context engineering has emerged as a pivotal paradigm for unlocking the potential of Large Language Models (LLMs) in Software Engineering (SE) tasks, enabling performance gains at test time without model fine-tuning. Despite its success, existing research lacks a systematic taxonomy of SE-specific context types and a dedicated benchmark to quantify the heterogeneous effects of different contexts across core SE workflows. To address this gap, we propose CL4SE (Context Learning for Software Engineering), a comprehensive benchmark featuring a fine-grained taxonomy of four SE-oriented context types (interpretable examples, project-specific context, procedural decision-making context, and positive & negative context), each mapped to a representative task (code generation, code summarization, code review, and patch correctness assessment). We construct high-quality datasets comprising over 13,000 samples from more than 30 open-source projects and evaluate five mainstream LLMs across nine metrics. Extensive experiments demonstrate that context learning yields an average performance improvement of 24.7% across all tasks. Specifically, procedural context boosts code review performance by up to 33% (Qwen3-Max), mixed positive-negative context improves patch assessment by 30% (DeepSeek-V3), project-specific context increases code summarization BLEU by 14.78% (GPT-Oss-120B), and interpretable examples enhance code generation PASS@1 by 5.72% (DeepSeek-V3). CL4SE establishes the first standardized evaluation framework for SE context learning, provides actionable empirical insights into task-specific context design, and releases a large-scale dataset to facilitate reproducible research in this domain.

Hallucinations in generative AI, particularly in Large Language Models (LLMs), pose a significant challenge to the reliability of multilingual applications. Existing benchmarks for hallucination detection focus primarily on English and a few widely spoken languages, lacking the breadth to assess inconsistencies in model performance across diverse linguistic contexts. To address this gap, we introduce Poly-FEVER, a large-scale multilingual fact verification benchmark specifically designed for evaluating hallucination detection in LLMs. Poly-FEVER comprises 77,973 labeled factual claims spanning 11 languages, sourced from FEVER, Climate-FEVER, and SciFact. It provides the first large-scale dataset tailored for analyzing hallucination patterns across languages, enabling systematic evaluation of LLMs such as ChatGPT and the LLaMA series. Our analysis reveals how topic distribution and web resource availability influence hallucination frequency, uncovering language-specific biases that impact model accuracy. By offering a multilingual benchmark for fact verification, Poly-FEVER facilitates cross-linguistic comparisons of hallucination detection and contributes to the development of more reliable, language-inclusive AI systems. The dataset is publicly available to advance research in responsible AI, fact-checking methodologies, and multilingual NLP, promoting greater transparency and robustness in LLM performance. The proposed Poly-FEVER is available at: https://huggingface.co/datasets/HanzhiZhang/Poly-FEVER.

In this work, we propose the world's first closed-loop ML-based planning benchmark for autonomous driving. While there is a growing body of ML-based motion planners, the lack of established datasets and metrics has limited the progress in this area. Existing benchmarks for autonomous vehicle motion prediction have focused on short-term motion forecasting, rather than long-term planning. This has led previous works to use open-loop evaluation with L2-based metrics, which are not suitable for fairly evaluating long-term planning. Our benchmark overcomes these limitations by introducing a large-scale driving dataset, lightweight closed-loop simulator, and motion-planning-specific metrics. We provide a high-quality dataset with 1500h of human driving data from 4 cities across the US and Asia with widely varying traffic patterns (Boston, Pittsburgh, Las Vegas and Singapore). We will provide a closed-loop simulation framework with reactive agents and provide a large set of both general and scenario-specific planning metrics. We plan to release the dataset at NeurIPS 2021 and organize benchmark challenges starting in early 2022.

Automated formal verification remains challenging for large language models because data for proof assistants and verification-aware languages is scarce, and correctness depends on satisfying precise machine-checkable specifications rather than producing plausible code. This thesis studies how verifier environments can improve LLM generation of verified programs and proofs through reinforcement learning from verifiable rewards (RLVR) and verifier-guided inference-time search. First, we train open-source models in Dafny with RLVR using Group Relative Policy Optimization (GRPO) and related variants, assembling generated candidates into complete programs and scoring them with compiler and verifier outcomes. Initial experiments on an APPS-derived Dafny dataset increased verified reward from 2.2% to 58.1%, but revealed specification hacking, where models exploit weak formal specifications instead of implementing the intended solutions. After filtering underspecified and vulnerable tasks, multi-turn RLVR on the refined benchmark improves the verified pass rate from 9.7% to 31.1%. Second, we develop a verifier-guided inference scaffold in Lean that treats proof generation as structured search over decomposed subgoals, verifier feedback, diagnostics, and repair. With a fixed base model, the full scaffold with proof reviser improves pass rate on an initial VeriCoding pilot set from 46.2% under direct repair to 69.2%. On the larger VERINA dataset, whole-task decomposition plus proof reviser solves 7 of 42 previously unsolved tasks. We also introduce Dalek-Bench, a repository-scale Lean benchmark derived from the Rust curve25519-dalek verification project; preliminary results remain weak, indicating that stronger progress evaluation and task-specific tool-use policies are still needed.

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

Recent advancements in AI-driven soccer understanding have demonstrated rapid progress, yet existing research predominantly focuses on isolated or narrow tasks. To bridge this gap, we propose a comprehensive framework for holistic soccer understanding. Specifically, we make the following contributions in this paper: (i) we construct SoccerWiki, the first large-scale multimodal soccer knowledge base, integrating rich domain knowledge about players, teams, referees, and venues to enable knowledge-driven reasoning; (ii) we present SoccerBench, the largest and most comprehensive soccer-specific benchmark, featuring around 10K standardized multimodal (text, image, video) multi-choice QA pairs across 13 distinct understanding tasks, curated through automated pipelines and manual verification; (iii) we introduce SoccerAgent, a novel multi-agent system that decomposes complex soccer questions via collaborative reasoning, leveraging domain expertise from SoccerWiki and achieving robust performance; (iv) extensive evaluations and ablations that benchmark state-of-the-art MLLMs on SoccerBench, highlighting the superiority of our proposed agentic system. All data and code are publicly available at: https://jyrao.github.io/SoccerAgent/.

Style-content dual-reference generation aims to synthesize an image that preserves the structure and semantics of a content reference while adopting the style of a separate style reference.Despite recent progress, this setting remains challenging because models must balance content fidelity, style alignment, and instruction following avoiding semantic leakage from the style reference.A key bottleneck is the lack of large-scale triplet data with clean content-style separation and broad long-tail style coverage.In this work, we propose FreeStyle, a scalable dual-reference generation framework based on community LoRA mining.We treat community LoRAs as compositional anchors for style and content, and design a rigorous generation and filtering pipeline to construct large-scale Style-Reference and Content-Reference triplets across multiple base models.To address content leakage, we adopt a two-stage curriculum with stage-specific disentanglement mechanisms: an attention-level enrichment constraint that suppresses style-reference leakage in the style-transfer stage, and a frequency-aware RoPE modulation strategy that targets positional-correspondence-based leakage in the harder dual-reference stage.We also introduce a benchmark covering both style-reference and dual-reference generation, with evaluations on style similarity, content preservation, aesthetics, instruction following, and leakage rejection. The benchmark incorporates a style-invariant Content Alignment Score (CAS) and introduces a calibrated VLM-based Rejection Score for evaluating generation reliability and leakage suppression.Extensive experiments show that our model achieves a strong balance among style alignment, content preservation, and leakage suppression.

Text-to-image diffusion models have impactful applications in art, design, and entertainment, yet these technologies also pose significant risks by enabling the creation and dissemination of misinformation. Although recent advancements have produced AI-generated image detectors that claim robustness against various augmentations, their true effectiveness remains uncertain. Do these detectors reliably identify images with different levels of augmentation? Are they biased toward specific scenes or data distributions? To investigate, we introduce SEMI-TRUTHS, featuring 27,600 real images, 223,400 masks, and 1,472,700 AI-augmented images that feature targeted and localized perturbations produced using diverse augmentation techniques, diffusion models, and data distributions. Each augmented image is accompanied by metadata for standardized and targeted evaluation of detector robustness. Our findings suggest that state-of-the-art detectors exhibit varying sensitivities to the types and degrees of perturbations, data distributions, and augmentation methods used, offering new insights into their performance and limitations. The code for the augmentation and evaluation pipeline is available at https://github.com/J-Kruk/SemiTruths.

The prohibitive cost of evaluating large language models (LLMs) on comprehensive benchmarks necessitates the creation of small yet representative data subsets (i.e., tiny benchmarks) that enable efficient assessment while retaining predictive fidelity. Current methods for this task operate under a model-centric paradigm, selecting benchmarking items based on the collective performance of existing models. Such approaches are limited by large upfront costs, an inability to immediately handle new benchmarks (`cold-start'), and the fragile assumption that future models will share the failure patterns of their predecessors. In this work, we challenge this paradigm and propose a item-centric approach to benchmark subset selection, arguing that selection should be based on the intrinsic properties of the task items themselves, rather than on model-specific failure patterns. We instantiate this item-centric efficient benchmarking approach via a novel method, Scales++, where data selection is based on the cognitive demands of the benchmark samples. Empirically, we show Scales++ reduces the upfront selection cost by over 18x while achieving competitive predictive fidelity. On the Open LLM Leaderboard, using just a 0.5\% data subset, we predict full benchmark scores with a 2.9% mean absolute error. We demonstrate that this item-centric approach enables more efficient model evaluation without significant fidelity degradation, while also providing better cold-start performance and more interpretable benchmarking.

Frontier model progress is often measured by academic benchmarks, which offer a limited view of performance in real-world professional contexts. Existing evaluations often fail to assess open-ended, economically consequential tasks in high-stakes domains like Legal and Finance, where practical returns are paramount. To address this, we introduce Professional Reasoning Bench (PRBench), a realistic, open-ended, and difficult benchmark of real-world problems in Finance and Law. We open-source its 1,100 expert-authored tasks and 19,356 expert-curated criteria, making it, to our knowledge, the largest public, rubric-based benchmark for both legal and finance domains. We recruit 182 qualified professionals, holding JDs, CFAs, or 6+ years of experience, who contributed tasks inspired by their actual workflows. This process yields significant diversity, with tasks spanning 114 countries and 47 US jurisdictions. Our expert-curated rubrics are validated through a rigorous quality pipeline, including independent expert validation. Subsequent evaluation of 20 leading models reveals substantial room for improvement, with top scores of only 0.39 (Finance) and 0.37 (Legal) on our Hard subsets. We further catalog associated economic impacts of the prompts and analyze performance using human-annotated rubric categories. Our analysis shows that models with similar overall scores can diverge significantly on specific capabilities. Common failure modes include inaccurate judgments, a lack of process transparency and incomplete reasoning, highlighting critical gaps in their reliability for professional adoption.

Large language models (LLMs) show remarkable potential to act as computer agents, enhancing human productivity and software accessibility in multi-modal tasks that require planning and reasoning. However, measuring agent performance in realistic environments remains a challenge since: (i) most benchmarks are limited to specific modalities or domains (e.g. text-only, web navigation, Q&A, coding) and (ii) full benchmark evaluations are slow (on order of magnitude of days) given the multi-step sequential nature of tasks. To address these challenges, we introduce the Windows Agent Arena: a reproducible, general environment focusing exclusively on the Windows operating system (OS) where agents can operate freely within a real Windows OS and use the same wide range of applications, tools, and web browsers available to human users when solving tasks. We adapt the OSWorld framework (Xie et al., 2024) to create 150+ diverse Windows tasks across representative domains that require agent abilities in planning, screen understanding, and tool usage. Our benchmark is scalable and can be seamlessly parallelized in Azure for a full benchmark evaluation in as little as 20 minutes. To demonstrate Windows Agent Arena's capabilities, we also introduce a new multi-modal agent, Navi. Our agent achieves a success rate of 19.5% in the Windows domain, compared to 74.5% performance of an unassisted human. Navi also demonstrates strong performance on another popular web-based benchmark, Mind2Web. We offer extensive quantitative and qualitative analysis of Navi's performance, and provide insights into the opportunities for future research in agent development and data generation using Windows Agent Arena. Webpage: https://microsoft.github.io/WindowsAgentArena Code: https://github.com/microsoft/WindowsAgentArena

Large Language Models (LLMs) have demonstrated potential in assisting with Register Transfer Level (RTL) design tasks. Nevertheless, there remains to be a significant gap in benchmarks that accurately reflect the complexity of real-world RTL projects. To address this, this paper presents RTL-Repo, a benchmark specifically designed to evaluate LLMs on large-scale RTL design projects. RTL-Repo includes a comprehensive dataset of more than 4000 Verilog code samples extracted from public GitHub repositories, with each sample providing the full context of the corresponding repository. We evaluate several state-of-the-art models on the RTL-Repo benchmark, including GPT-4, GPT-3.5, Starcoder2, alongside Verilog-specific models like VeriGen and RTLCoder, and compare their performance in generating Verilog code for complex projects. The RTL-Repo benchmark provides a valuable resource for the hardware design community to assess and compare LLMs' performance in real-world RTL design scenarios and train LLMs specifically for Verilog code generation in complex, multi-file RTL projects. RTL-Repo is open-source and publicly available on Github.

Large Language Model agents increasingly operate external systems through programmatic interfaces, yet practitioners lack empirical guidance on how to structure the context these agents consume. Using SQL generation as a proxy for programmatic agent operations, we present a systematic study of context engineering for structured data, comprising 9,649 experiments across 11 models, 4 formats (YAML, Markdown, JSON, Token-Oriented Object Notation [TOON]), and schemas ranging from 10 to 10,000 tables. Our findings challenge common assumptions. First, architecture choice is model-dependent: file-based context retrieval improves accuracy for frontier-tier models (Claude, GPT, Gemini; +2.7%, p=0.029) but shows mixed results for open source models (aggregate -7.7%, p<0.001), with deficits varying substantially by model. Second, format does not significantly affect aggregate accuracy (chi-squared=2.45, p=0.484), though individual models, particularly open source, exhibit format-specific sensitivities. Third, model capability is the dominant factor, with a 21 percentage point accuracy gap between frontier and open source tiers that dwarfs any format or architecture effect. Fourth, file-native agents scale to 10,000 tables through domain-partitioned schemas while maintaining high navigation accuracy. Fifth, file size does not predict runtime efficiency: compact or novel formats can incur a token overhead driven by grep output density and pattern unfamiliarity, with the magnitude depending on model capability. These findings provide practitioners with evidence-based guidance for deploying LLM agents on structured systems, demonstrating that architectural decisions should be tailored to model capability rather than assuming universal best practices.

The advent of next-generation radio telescopes is set to transform radio astronomy by producing massive data volumes that challenge traditional processing methods. Deep learning techniques have shown strong potential in automating radio analysis tasks, yet are often constrained by the limited availability of large annotated datasets. Recent progress in self-supervised learning has led to foundational radio vision models, but adapting them for new tasks typically requires coding expertise, limiting their accessibility to a broader astronomical community. Text-based AI interfaces offer a promising alternative by enabling task-specific queries and example-driven learning. In this context, Large Language Models (LLMs), with their remarkable zero-shot capabilities, are increasingly used in scientific domains. However, deploying large-scale models remains resource-intensive, and there is a growing demand for AI systems that can reason over both visual and textual data in astronomical analysis. This study explores small-scale Vision-Language Models (VLMs) as AI assistants for radio astronomy, combining LLM capabilities with vision transformers. We fine-tuned the LLaVA VLM on a dataset of 59k radio images from multiple surveys, enriched with 38k image-caption pairs from the literature. The fine-tuned models show clear improvements over base models in radio-specific tasks, achieving ~30% F1-score gains in extended source detection, but they underperform pure vision models and exhibit ~20% drop on general multimodal tasks. Inclusion of caption data and LoRA fine-tuning enhances instruction-following and helps recover ~10% accuracy on standard benchmarks. This work lays the foundation for future advancements in radio VLMs, highlighting their potential and limitations, such as the need for better multimodal alignment, higher-quality datasets, and mitigation of catastrophic forgetting.

Recent years have witnessed the rapid development of general multimodal large language models (MLLMs). However, adapting general MLLMs to specific domains, such as scientific fields and industrial applications, remains less explored. This paper systematically investigates domain adaptation of MLLMs through post-training, focusing on data synthesis, training pipelines, and task evaluation. (1) Data Synthesis: Using open-source models, we develop a visual instruction synthesizer that effectively generates diverse visual instruction tasks from domain-specific image-caption pairs. Our synthetic tasks surpass those generated by manual rules, GPT-4, and GPT-4V in enhancing the domain-specific performance of MLLMs. (2) Training Pipeline: While the two-stage training--initially on image-caption pairs followed by visual instruction tasks--is commonly adopted for developing general MLLMs, we apply a single-stage training pipeline to enhance task diversity for domain-specific post-training. (3) Task Evaluation: We conduct experiments in two domains, biomedicine and food, by post-training MLLMs of different sources and scales (e.g., Qwen2-VL-2B, LLaVA-v1.6-8B, Llama-3.2-11B), and then evaluating MLLM performance on various domain-specific tasks. To support further research in MLLM domain adaptation, we will open-source our implementations.

Large language models (LLMs) have shown strong performance on isolated psychiatric tasks, including dialogue, diagnosis, and treatment planning, yet existing benchmarks rarely simulate complete psychiatric clinical encounters. We introduce MentalHospital, a virtual evaluation environment for LLM-based psychiatric clinical encounters. MentalHospital instantiates the Subjective Interviewing, Objective Examination, Diagnostic Assessment, and Treatment Planning (S.O.A.P.) workflow, using skill-augmented standardized patients constructed from 1,193 de-identified psychiatric electronic health record (EHR) cases spanning all major ICD-11 categories and 76 disorders. Each encounter is assessed through a dual-track protocol that combines objective comparison against EHR-derived references with subjective assessment of clinical process quality. To scale specialist judgment, we develop MentalEval, five domain-specific evaluators covering communication empathy, interviewing professionalism, clinical-note quality, diagnostic rigor, and treatment appropriateness, trained with rubric-grounded SFT and expert-guided DPO. Survey responses from 22 clinicians support MentalHospital's clinical fidelity (3.88/5), while MentalEval achieves strong expert alignment with an average QWK of 0.944. Benchmarking shows that even the strongest LLM trails clinicians by 37.28 percentage points in objective psychiatric competence, with mental status assessment as a key bottleneck.

The rapid advancement of large language models (LLMs) such as GPT-3, PaLM, and Llama has significantly transformed natural language processing, showcasing remarkable capabilities in understanding and generating language. However, these models often struggle with tasks requiring complex reasoning, particularly in mathematical problem-solving, due in part to the scarcity of large-scale, high-quality, domain-specific datasets necessary for training sophisticated reasoning abilities. To address this limitation, we introduce Template-based Data Generation (TDG), a novel approach that leverages LLMs (GPT-4) to automatically generate parameterized meta-templates, which are then used to synthesize a vast array of high-quality problems and solutions. Leveraging TDG, we create TemplateMath Part I: TemplateGSM, a dataset comprising over 7 million synthetically generated grade school math problems--each accompanied by code-based and natural language solutions--with the potential to generate an effectively unlimited number more. This dataset alleviates the scarcity of large-scale mathematical datasets and serves as a valuable resource for pre-training, fine-tuning, and evaluating LLMs in mathematical reasoning. Our method not only enables the generation of virtually infinite data but also elevates data augmentation to a new level by using GPT-4 for meta-template generation, ensuring diverse and high-quality problem structures. The TemplateMath Part I: TemplateGSM dataset is publicly available at https://huggingface.co/datasets/math-ai/TemplateGSM. The code is available at https://github.com/iiis-ai/TemplateMath.

Given that Large Language Models (LLMs) have made significant progress in writing code, can they now be used to autonomously reproduce results from research repositories? Such a capability would be a boon to the research community, helping researchers validate, understand, and extend prior work. To advance towards this goal, we introduce SUPER, the first benchmark designed to evaluate the capability of LLMs in setting up and executing tasks from research repositories. SUPERaims to capture the realistic challenges faced by researchers working with Machine Learning (ML) and Natural Language Processing (NLP) research repositories. Our benchmark comprises three distinct problem sets: 45 end-to-end problems with annotated expert solutions, 152 sub problems derived from the expert set that focus on specific challenges (e.g., configuring a trainer), and 602 automatically generated problems for larger-scale development. We introduce various evaluation measures to assess both task success and progress, utilizing gold solutions when available or approximations otherwise. We show that state-of-the-art approaches struggle to solve these problems with the best model (GPT-4o) solving only 16.3% of the end-to-end set, and 46.1% of the scenarios. This illustrates the challenge of this task, and suggests that SUPER can serve as a valuable resource for the community to make and measure progress.

Geospatial foundation models aim to learn representations that transfer across regions and sensors, yet evaluating them on specific tasks requires large, high-quality, multi-modal benchmarks that measure how well such models extract value from data. Concerning flood mapping, existing datasets rarely combine bi-temporal SAR and co-registered optical imagery at scale, leaving the value of foundation models for this downstream task largely untested. We introduce GEOID-Flood, a large-scale multi-modal flood segmentation benchmark, derived from Copernicus Emergency Management Service activations, spanning 219 events across 65 countries over ten years. The dataset provides more than 14,000 tiles with co-registered pre- and post-event Sentinel-1, in GRD and RTC format, pre-event Sentinel-2 composite, and DEM, including manually validated labels that separate background from permanent water and flooded water. Using this benchmark, we evaluate foundation models against conventional encoders across single-image, multi-temporal, and multi-modal protocols. We report three main findings: foundation models offer a consistent but modest advantage; optical-SAR fusion with finetuning best resolves transient flooding; and models trained on GEOID-Flood transfer to unseen events better than those trained on existing datasets. Dataset and code available at https://github.com/links-ads/geoid-flood.

Large language models (LLMs) are currently applied to scientific paper evaluation by assigning an absolute score to each paper independently. However, since score scales vary across conferences, time periods, and evaluation criteria, models trained on absolute scores are prone to fitting narrow, context-specific rules rather than developing robust scholarly judgment. To overcome this limitation, we propose shifting paper evaluation from isolated scoring to collaborative ranking. In particular, we design Comparison-Native framework for Paper Evaluation (CNPE), integrating comparison into both data construction and model learning. We first propose a graph-based similarity ranking algorithm to facilitate the sampling of more informative and discriminative paper pairs from a collection. We then enhance relative quality judgment through supervised fine-tuning and reinforcement learning with comparison-based rewards. At inference, the model performs pairwise comparisons over sampled paper pairs and aggregates these preference signals into a global relative quality ranking. Experimental results demonstrate that our framework achieves an average relative improvement of 21.8\% over the strong baseline DeepReview-14B, while exhibiting robust generalization to five previously unseen datasets. https://github.com/ECNU-Text-Computing/ComparisonReview{Code}.

Instruction-following capability has become a major ability to be evaluated for Large Language Models (LLMs). However, existing datasets, such as IFEval, are either predominantly monolingual and centered on English or simply machine translated to other languages, limiting their applicability in multilingual contexts. In this paper, we present an carefully-curated extension of IFEval to a localized multilingual version named Marco-Bench-MIF, covering 30 languages with varying levels of localization. Our benchmark addresses linguistic constraints (e.g., modifying capitalization requirements for Chinese) and cultural references (e.g., substituting region-specific company names in prompts) via a hybrid pipeline combining translation with verification. Through comprehensive evaluation of 20+ LLMs on our Marco-Bench-MIF, we found that: (1) 25-35% accuracy gap between high/low-resource languages, (2) model scales largely impact performance by 45-60% yet persists script-specific challenges, and (3) machine-translated data underestimates accuracy by7-22% versus localized data. Our analysis identifies challenges in multilingual instruction following, including keyword consistency preservation and compositional constraint adherence across languages. Our Marco-Bench-MIF is available at https://github.com/AIDC-AI/Marco-Bench-MIF.

As streaming video generation advances, enabling real-time interaction with generated video is becoming increasingly important. We introduce StreamEval, a unified benchmark for interactive streaming video generation spanning five tasks: avatar dialogue, avatar performance, world generation, instruction editing, and reference editing. StreamEval comprises 300 scripted test cases lasting 15-300s, each specifying initial assets, incrementally arriving multimodal requests, and explicit requirements for state changes and content preservation. Performance is assessed along four evaluation dimensions: task execution, state consistency, visual fidelity, and audiovisual coordination, instantiated through ten task-specific metrics on a 0-100 scale. Evaluating 37 distinct models across 300 test cases, we find that high visual quality and identity retention do not necessarily coincide with reliable request execution or sustained state consistency. In particular, request execution remains challenging: the highest task-specific scores are 36.90 for Following among seven avatar-performance models, 25.93 for Following in text-conditioned world generation, and 15.00 for Edit Success among three reported reference-editing models. These findings identify reliable request execution under timing and preservation constraints as a central challenge for advancing omni-modal real-time interactive systems. The leaderboard is available at https://streameval.ai

The rapid release of both language models and benchmarks makes it increasingly costly to evaluate every model on every dataset. In practice, models are often evaluated on different samples, making scores difficult to compare across studies. To address this, we propose a framework based on multidimensional Item Response Theory (IRT) that uses anchor items to calibrate new benchmarks to the evaluation suite while holding previously calibrated item parameters fixed. Our approach supports a realistic evaluation setting in which datasets are introduced over time and models are evaluated only on the datasets available at the time of evaluation, while a fixed anchor set for each dataset is used so that results from different evaluation periods can be compared directly. In large-scale experiments on more than 400 models, our framework predicts full-evaluation performance within 2-3 percentage points using only 100 anchor questions per dataset, with Spearman ρgeq 0.9 for ranking preservation, showing that it is possible to extend benchmark suites over time while preserving score comparability, at a constant evaluation cost per new dataset. Code available at https://github.com/eliyahabba/growing-pains

Despite the utility of Large Language Models (LLMs) across a wide range of tasks and scenarios, developing a method for reliably evaluating LLMs across varied contexts continues to be challenging. Modern evaluation approaches often use LLMs to assess responses generated by LLMs. However, the meta-evaluation conducted to assess the effectiveness of these LLMs as evaluators is typically constrained by the coverage of existing benchmarks or requires extensive human annotation. This underscores the urgency of methods for scalable meta-evaluation that can effectively, reliably, and efficiently evaluate the performance of LLMs as evaluators across diverse tasks and scenarios, particularly in potentially new, user-defined scenarios. To fill this gap, we propose ScaleEval, an agent-debate-assisted meta-evaluation framework that leverages the capabilities of multiple communicative LLM agents. This framework supports multi-round discussions to assist human annotators in discerning the most capable LLMs as evaluators, which significantly eases their workload in cases that used to require large-scale annotations during meta-evaluation. We release the code for our framework, which is publicly available at: https://github.com/GAIR-NLP/scaleeval.

Modern image generation and editing systems can produce photorealistic, prompt-aligned images, but still often render familiar objects at implausible relative sizes. To measure this failure mode, we introduce GenScale, a benchmark and evaluation protocol for real-world relative object scale in image generation and editing. GenScale contains 900 image-level entries and 1,643 pairwise anchor-target scale relations across common-object generation, human-product generation with metric dimensions, and scale correction from failed generations. We further design a human-calibrated ordinal judge for scalable pairwise scale evaluation. Last but not the least, we introduce Rescale, a model-agnostic post-processing agent for localized scale correction without modifying the source generator. Experiments reveal that state-of-the-art image generators and editors cannot reliably observe relative scale yet, while Rescale consistently improves scale plausibility across generated and edited images. Together, GenScale establishes relative object scale as a distinct, measurable, and actionable capability for image generation systems.

Scaling up language models has been empirically shown to improve performance on a wide range of downstream tasks. However, if we were to observe worse performance as a function of scale ("inverse scaling") on certain tasks, this would indicate that scaling can also encourage behaviors that are misaligned with human preferences. The Inverse Scaling Prize (McKenzie et al. 2022) identified eleven such inverse scaling tasks, evaluated on models of up to 280B parameters and up to 500 zettaFLOPs of training compute. This paper takes a closer look at these inverse scaling tasks. We evaluate models of up to 540B parameters, trained on five times more compute than those evaluated in the Inverse Scaling Prize. With this increased range of model sizes and training compute, only four out of the eleven tasks remain inverse scaling. Six out of the eleven tasks exhibit "U-shaped scaling", where performance decreases up to a certain size, and then increases again up to the largest model evaluated (the one remaining task displays positive scaling). In addition, we find that 1-shot examples and chain-of-thought can help mitigate undesirable scaling patterns even further. U-shaped scaling suggests that the inverse scaling trend observed in McKenzie et al. (2022) may not continue to hold for larger models, which we attribute to the presence of distractor tasks that only sufficiently large models can avoid.

Conversational interviews are commonly used to complement structured surveys by eliciting rich and contextualized responses, which are typically analyzed qualitatively. However, their potential contribution to quantitative measurement remains underexplored. In this paper, we introduce ConvScale, an AI-supported approach that transforms psychometric scales into natural conversational interviews while preserving the original measurement structure. Based on interview data, ConvScale predicts item-level scores and aggregates them to derive scale-based assessments. In a within-subjects study with 18 participants, our results show that ConvScale-derived scores align closely with participants' self-report scores at both the item and construct levels, while maintaining moderate internal reliability; however, the structural validity was inadequate. In light of this, we discussed the potential of supporting quantitative measurement through interviews and proposed implications for future designs.

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.)

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

Standard evaluation protocols reveal a counterintuitive phenomenon: on 7.7% of benchmark problems spanning five datasets, larger language models underperform smaller ones by 28.4 percentage points despite 10-100x more parameters. Through systematic evaluation of 31 models (0.5B-405B parameters) across 1,485 problems, we identify the mechanism as spontaneous scale-dependent verbosity that introduces errors through overelaboration. Causal intervention experiments demonstrate this reflects correctable prompt design rather than fundamental capability limitations. Constraining large models to produce brief responses improves accuracy by 26 percentage points and reduces performance gaps by up to two-thirds. Most critically, brevity constraints completely reverse performance hierarchies on mathematical reasoning and scientific knowledge benchmarks, with large models achieving 7.7-15.9 percentage point advantages over small models -- direct inversions of the original gaps. These reversals prove large models possess superior latent capabilities that universal prompting masks. We validate findings through three independent contamination tests and demonstrate inverse scaling operates continuously across the full parameter spectrum, with dataset-specific optimal scales ranging from 0.5B to 3.0B parameters. Our results establish that maximizing large model performance requires scale-aware prompt engineering rather than universal evaluation protocols, with immediate implications for deployment: prompt adaptation simultaneously improves accuracy and reduces computational costs.

Recent advances have made long-form report-generating systems widely available. This has prompted evaluation frameworks that use LLM-as-judge protocols and claim verification, along with meta-evaluation frameworks that seek to validate these methods. Many of the meta-evaluations estimate an evaluation quality's by comparing its assessments against human pairwise preferences. Prior work, however, suggests that human pairwise preference may be overly simplistic and can fail to capture nuances of expert expectations. We conduct a case study in meta-evaluation for long-form QA benchmarks using ScholarQA-CS2, a benchmark designed for assessing retrieval-augmented deep-research QA in the scientific domain. We comprehensively validate the benchmark through human pairwise preference judgments, then critically examine the strengths, weaknesses, and confounders of this approach. We show that pairwise preference rankings are best suited for system-level evaluation, while explicit metric-wise annotations and expert annotators are critical for reliable metric-level assessment, with subjectivity remaining a key challenge. Based on our findings, we offer practical guidelines for designing future meta-evaluations that better align evaluation methods, annotator expertise, and reporting practices. By surfacing these methodological challenges, we aim to advance evaluation standards for deep-research systems.

Predictable behavior from scaling advanced AI systems is an extremely desirable property. Although a well-established literature exists on how pretraining performance scales, the literature on how particular downstream capabilities scale is significantly muddier. In this work, we take a step back and ask: why has predicting specific downstream capabilities with scale remained elusive? While many factors are certainly responsible, we identify a new factor that makes modeling scaling behavior on widely used multiple-choice question-answering benchmarks challenging. Using five model families and twelve well-established multiple-choice benchmarks, we show that downstream performance is computed from negative log likelihoods via a sequence of transformations that progressively degrade the statistical relationship between performance and scale. We then reveal the mechanism causing this degradation: downstream metrics require comparing the correct choice against a small number of specific incorrect choices, meaning accurately predicting downstream capabilities requires predicting not just how probability mass concentrates on the correct choice with scale, but also how probability mass fluctuates on specific incorrect choices with scale. We empirically study how probability mass on the correct choice co-varies with probability mass on incorrect choices with increasing compute, suggesting that scaling laws for incorrect choices might be achievable. Our work also explains why pretraining scaling laws are commonly regarded as more predictable than downstream capabilities and contributes towards establishing scaling-predictable evaluations of frontier AI models.

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

Language model (LM) benchmarking faces several challenges: comprehensive evaluations are costly, benchmarks often fail to measure the intended capabilities, and evaluation quality can degrade due to labeling errors and benchmark saturation. Although various strategies have been proposed to mitigate these issues, they tend to address individual aspects in isolation, neglecting broader questions about overall evaluation quality. Here, we introduce Fluid Benchmarking, a new evaluation approach that advances LM benchmarking across multiple dimensions. Inspired by psychometrics, Fluid Benchmarking is based on the insight that the relative value of benchmark items depends on an LM's capability level, suggesting that evaluation should adapt to each LM. Methodologically, Fluid Benchmarking estimates an item response model based on existing LM evaluation results and uses the inferred quantities to select evaluation items dynamically, similar to computerized adaptive testing in education. In our experiments, we compare Fluid Benchmarking against the common practice of random item sampling as well as more sophisticated baselines, including alternative methods grounded in item response theory. We examine four dimensions -- efficiency, validity, variance, and saturation -- and find that Fluid Benchmarking achieves superior performance in all of them (e.g., higher validity and less variance on MMLU with fifty times fewer items). Our analysis shows that the two components of Fluid Benchmarking have distinct effects: item response theory, used to map performance into a latent ability space, increases validity, while dynamic item selection reduces variance. Overall, our results suggest that LM benchmarking can be substantially improved by moving beyond static evaluation.

Representation engineering methods have recently shown promise for enabling efficient steering of model behavior. However, evaluation pipelines for these methods have primarily relied on subjective demonstrations, instead of quantitative, objective metrics. We aim to take a step towards addressing this issue by advocating for four properties missing from current evaluations: (i) contexts sufficiently similar to downstream tasks should be used for assessing intervention quality; (ii) model likelihoods should be accounted for; (iii) evaluations should allow for standardized comparisons across different target behaviors; and (iv) baseline comparisons should be offered. We introduce an evaluation pipeline grounded in these criteria, offering both a quantitative and visual analysis of how effectively a given method works. We use this pipeline to evaluate two representation engineering methods on how effectively they can steer behaviors such as truthfulness and corrigibility, finding that some interventions are less effective than previously reported.

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

A benchmark is not merely a collection of test questions; it naturally induces an empirical measure over difficulty. In this work, we develop this measure-based view of benchmark evaluation. This perspective allows us to identify and formalize two central failure modes: measure mismatch, where the benchmark difficulty distribution differs from the intended difficulty distribution, and coverage gaps, where uncovered regions along the difficulty axis prevent benchmarks from detecting meaningful localized training progress. We propose two diagnostic metrics that capture these two failure modes and provide rigorous theoretical guarantees. Empirically, we analyze 25 popular benchmarks spanning 7 domains using carefully estimated item difficulties. Our results show that many benchmarks implicitly overweight certain difficulty regions, so their leaderboard conclusions should be interpreted with caution; others leave substantial portions of the difficulty axis uncovered, limiting their ability to monitor the model improvement process. Finally, we turn diagnosis into action by introducing BenchFill, a simple difficulty-targeted item rewriting algorithm that effectively reduces local coverage gaps.

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

Work on scaling laws has found that large language models (LMs) show predictable improvements to overall loss with increased scale (model size, training data, and compute). Here, we present evidence for the claim that LMs may show inverse scaling, or worse task performance with increased scale, e.g., due to flaws in the training objective and data. We present empirical evidence of inverse scaling on 11 datasets collected by running a public contest, the Inverse Scaling Prize, with a substantial prize pool. Through analysis of the datasets, along with other examples found in the literature, we identify four potential causes of inverse scaling: (i) preference to repeat memorized sequences over following in-context instructions, (ii) imitation of undesirable patterns in the training data, (iii) tasks containing an easy distractor task which LMs could focus on, rather than the harder real task, and (iv) correct but misleading few-shot demonstrations of the task. We release the winning datasets at https://inversescaling.com/data to allow for further investigation of inverse scaling. Our tasks have helped drive the discovery of U-shaped and inverted-U scaling trends, where an initial trend reverses, suggesting that scaling trends are less reliable at predicting the behavior of larger-scale models than previously understood. Overall, our results suggest that there are tasks for which increased model scale alone may not lead to progress, and that more careful thought needs to go into the data and objectives for training language models.

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

Language models demonstrate both quantitative improvement and new qualitative capabilities with increasing scale. Despite their potentially transformative impact, these new capabilities are as yet poorly characterized. In order to inform future research, prepare for disruptive new model capabilities, and ameliorate socially harmful effects, it is vital that we understand the present and near-future capabilities and limitations of language models. To address this challenge, we introduce the Beyond the Imitation Game benchmark (BIG-bench). BIG-bench currently consists of 204 tasks, contributed by 442 authors across 132 institutions. Task topics are diverse, drawing problems from linguistics, childhood development, math, common-sense reasoning, biology, physics, social bias, software development, and beyond. BIG-bench focuses on tasks that are believed to be beyond the capabilities of current language models. We evaluate the behavior of OpenAI's GPT models, Google-internal dense transformer architectures, and Switch-style sparse transformers on BIG-bench, across model sizes spanning millions to hundreds of billions of parameters. In addition, a team of human expert raters performed all tasks in order to provide a strong baseline. Findings include: model performance and calibration both improve with scale, but are poor in absolute terms (and when compared with rater performance); performance is remarkably similar across model classes, though with benefits from sparsity; tasks that improve gradually and predictably commonly involve a large knowledge or memorization component, whereas tasks that exhibit "breakthrough" behavior at a critical scale often involve multiple steps or components, or brittle metrics; social bias typically increases with scale in settings with ambiguous context, but this can be improved with prompting.

As language models (LMs) scale, they develop many novel behaviors, good and bad, exacerbating the need to evaluate how they behave. Prior work creates evaluations with crowdwork (which is time-consuming and expensive) or existing data sources (which are not always available). Here, we automatically generate evaluations with LMs. We explore approaches with varying amounts of human effort, from instructing LMs to write yes/no questions to making complex Winogender schemas with multiple stages of LM-based generation and filtering. Crowdworkers rate the examples as highly relevant and agree with 90-100% of labels, sometimes more so than corresponding human-written datasets. We generate 154 datasets and discover new cases of inverse scaling where LMs get worse with size. Larger LMs repeat back a dialog user's preferred answer ("sycophancy") and express greater desire to pursue concerning goals like resource acquisition and goal preservation. We also find some of the first examples of inverse scaling in RL from Human Feedback (RLHF), where more RLHF makes LMs worse. For example, RLHF makes LMs express stronger political views (on gun rights and immigration) and a greater desire to avoid shut down. Overall, LM-written evaluations are high-quality and let us quickly discover many novel LM behaviors.

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.

Rigorous and reproducible evaluation is critical for assessing the state of the art and for guiding scientific advances in Artificial Intelligence. Evaluation is challenging in practice due to several reasons, including benchmark saturation, lack of transparency in methods used for measurement, development challenges in extracting measurements for generative tasks, and, more generally, the extensive number of capabilities required for a well-rounded comparison across models. We make three contributions to alleviate the above challenges. First, we present Eureka, an open-source framework for standardizing evaluations of large foundation models beyond single-score reporting and rankings. Second, we introduce Eureka-Bench as an extensible collection of benchmarks testing capabilities that (i) are still challenging for state-of-the-art models and (ii) represent fundamental but overlooked language and multimodal capabilities. The inherent space for improvement in non-saturated benchmarks enables us to discover meaningful differences between models at a capability level. Third, using Eureka, we conduct an analysis of 12 state-of-the-art models, providing in-depth insights into failure understanding and model comparison, which can be leveraged to plan targeted improvements. In contrast to recent trends in reports and leaderboards showing absolute rankings and claims for one model or another to be the best, our analysis shows that there is no such best model. Different models have different strengths, but there are models that appear more often than others as best performers for some capabilities. Despite the recent improvements, current models still struggle with several fundamental capabilities including detailed image understanding, benefiting from multimodal input when available rather than fully relying on language, factuality and grounding for information retrieval, and over refusals.

Evaluation benchmarks are the cornerstone of measuring capabilities of large language models (LLMs), as well as driving progress in said capabilities. Originally designed to make claims about capabilities (or lack thereof) in fully pretrained models, evaluation benchmarks are now also extensively used to decide between various training choices. Despite this widespread usage, we rarely quantify the variance in our evaluation benchmarks, which dictates whether differences in performance are meaningful. Here, we define and measure a range of metrics geared towards measuring variance in evaluation benchmarks, including seed variance across initialisations, and monotonicity during training. By studying a large number of models -- both openly available and pretrained from scratch -- we provide empirical estimates for a variety of variance metrics, with considerations and recommendations for practitioners. We also evaluate the utility and tradeoffs of continuous versus discrete performance measures and explore options for better understanding and reducing this variance. We find that simple changes, such as framing choice tasks (like MMLU) as completion tasks, can often reduce variance for smaller scale (sim7B) models, while more involved methods inspired from human testing literature (such as item analysis and item response theory) struggle to meaningfully reduce variance. Overall, our work provides insights into variance in evaluation benchmarks, suggests LM-specific techniques to reduce variance, and more generally encourages practitioners to carefully factor in variance when comparing models.

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

Large language models (LLMs) have emerged as powerful tools for analyzing complex datasets. Recent studies demonstrate their potential to generate useful, personalized responses when provided with patient-specific health information that encompasses lifestyle, biomarkers, and context. As LLM-driven health applications are increasingly adopted, rigorous and efficient one-sided evaluation methodologies are crucial to ensure response quality across multiple dimensions, including accuracy, personalization and safety. Current evaluation practices for open-ended text responses heavily rely on human experts. This approach introduces human factors and is often cost-prohibitive, labor-intensive, and hinders scalability, especially in complex domains like healthcare where response assessment necessitates domain expertise and considers multifaceted patient data. In this work, we introduce Adaptive Precise Boolean rubrics: an evaluation framework that streamlines human and automated evaluation of open-ended questions by identifying gaps in model responses using a minimal set of targeted rubrics questions. Our approach is based on recent work in more general evaluation settings that contrasts a smaller set of complex evaluation targets with a larger set of more precise, granular targets answerable with simple boolean responses. We validate this approach in metabolic health, a domain encompassing diabetes, cardiovascular disease, and obesity. Our results demonstrate that Adaptive Precise Boolean rubrics yield higher inter-rater agreement among expert and non-expert human evaluators, and in automated assessments, compared to traditional Likert scales, while requiring approximately half the evaluation time of Likert-based methods. This enhanced efficiency, particularly in automated evaluation and non-expert contributions, paves the way for more extensive and cost-effective evaluation of LLMs in health.

Large language model (LLM) judges are often used as rewards to train policies on objectives that deterministic verifiers cannot capture. However, these rewards are often collapsed to pass/fail ({0, 1}), which reports the verdict but not how well a response met each criterion. We model each pass/fail verdict as a score on an unreported scale, compared with one cutoff. A stretch of that scale moves every score proportionally toward or away from the cutoff, but never across it, so no verdict changes. A policy is therefore free to apply any stretch without changing anything the panel reports. Under a joint-Gaussian model, a third grade adds a second threshold and removes this affine stretch ambiguity. On MATH and SciBench outputs from one seven-criterion judge, all 14 constructed criterionwise stretches were invisible after binarization but visible with three grades. At n=1{,}024, a test given both population laws had at least 96.5% power at a 1.5times stress. Retaining grades closes one blind spot created by binarization, but verdicts alone remain insufficient as some changes are still indistinguishable from genuine improvement. These include arbitrary within-grade changes and fixed-covariance, loading-aligned mean shifts -- the signature of a sycophancy-shaped lift the panel reads as competence. The shared-factor reference approximation fit MATH and SciBench but not HealthBench, delineating its empirical scope. We recommend keeping at least three grades (for example, asking the judge whether each criterion is fully, partially, or not met and rewarding {0, 0.5, 1}), and externally validating gains along the remaining direction, which no finer scale removes.

Clinical personality assessment screens response validity before interpreting substantive scales. LLM evaluation does not. We apply the validity scaling framework from the PAI and MMPI-3 to metacognitive probe data from 20 frontier models across 524 items. Six validity indices are operationalised: L (maintaining confidence on errors), K (betting on errors), F (withdrawing consensus-endorsed items), Fp (withdrawing correct answers), RBS (inverted monitoring), and TRIN (fixed responding). A tiered classification system identifies four models as construct-level invalid and two as elevated. Valid-profile models produce item-sensitive confidence (mean r = .18, 14 of 16 significant). Invalid-profile models do not (mean r = -.20, d = 2.17, p = .001). Chain-of-thought training produces two opposite response distortions. Two latent dimensions account for 94.6% of index variance. Companion papers extract a portable screening protocol (Cacioli, 2026e) and validate it against selective prediction (Cacioli, 2026f). All data and code: https://github.com/synthiumjp/validity-scaling-llm

Pathological images are inherently multi-scale, requiring pathologists to integrate evidence from global tissue architecture at low magnification to cellular morphology at higher magnification for accurate diagnosis. While existing pathological datasets for vision-language models (VLMs) include various scales, they often lack explicit cross-scale reasoning objectives. This limitation prevents VLMs from capturing essential cross-scale representations and learning evidence-based reasoning. To bridge this gap, we introduce the first cross-scale training and evaluation paradigm that formulates pathology interpretation as multi-magnification reasoning. However, creating such a task reveals a critical challenge: multi-image visual question answering (VQA) is prone to text-only shortcuts, which allow models to guess answers using magnification-dependent artifacts rather than visual evidence. To address this, we propose a leakage-aware curation pipeline that combines adversarial text-only screening with constraint-guided question design. Using this pipeline, we construct Scale-VQA, a high-quality benchmark with 4,685 multiple-choice questions grounded in 2,537 pathology images across multiple magnification levels. Finally, we present ScaleReasoner-R1, a model trained via reinforcement learning to optimize performance on cross-scale VQA tasks. ScaleReasoner-R1 achieves state-of-the-art performance on our cross-scale reasoning benchmark and generalizes to SOTA performance on established single-scale benchmarks. Findings suggest that even the limited cross-scale supervision can significantly improve pathological understanding. Code is available at https://github.com/iMVR-PL/ScaleReasoner-R1.

We present an online post-hoc calibration method, called Online Platt Scaling (OPS), which combines the Platt scaling technique with online logistic regression. We demonstrate that OPS smoothly adapts between i.i.d. and non-i.i.d. settings with distribution drift. Further, in scenarios where the best Platt scaling model is itself miscalibrated, we enhance OPS by incorporating a recently developed technique called calibeating to make it more robust. Theoretically, our resulting OPS+calibeating method is guaranteed to be calibrated for adversarial outcome sequences. Empirically, it is effective on a range of synthetic and real-world datasets, with and without distribution drifts, achieving superior performance without hyperparameter tuning. Finally, we extend all OPS ideas to the beta scaling method.

Large language models (LLMs) are commonly used as evaluators in tasks (e.g., reward modeling, LLM-as-a-judge), where they act as proxies for human preferences or judgments. This leads to the need for meta-evaluation: evaluating the credibility of LLMs as evaluators. However, existing benchmarks primarily focus on English, offering limited insight into LLMs' effectiveness as evaluators in non-English contexts. To address this, we introduce MM-Eval, a multilingual meta-evaluation benchmark that covers 18 languages across six categories. MM-Eval evaluates various dimensions, including language-specific challenges like linguistics and language hallucinations. Evaluation results show that both proprietary and open-source language models have considerable room for improvement. Further analysis reveals a tendency for these models to assign middle-ground scores to low-resource languages. We publicly release our benchmark and code.

The impressive performance of large language models (LLMs) has attracted considerable attention from the academic and industrial communities. Besides how to construct and train LLMs, how to effectively evaluate and compare the capacity of LLMs has also been well recognized as an important yet difficult problem. Existing paradigms rely on either human annotators or model-based evaluators to evaluate the performance of LLMs on different tasks. However, these paradigms often suffer from high cost, low generalizability, and inherited biases in practice, which make them incapable of supporting the sustainable development of LLMs in long term. In order to address these issues, inspired by the peer review systems widely used in academic publication process, we propose a novel framework that can automatically evaluate LLMs through a peer-review process. Specifically, for the evaluation of a specific task, we first construct a small qualification exam to select "reviewers" from a couple of powerful LLMs. Then, to actually evaluate the "submissions" written by different candidate LLMs, i.e., the evaluatees, we use the reviewer LLMs to rate or compare the submissions. The final ranking of evaluatee LLMs is generated based on the results provided by all reviewers. We conducted extensive experiments on text summarization tasks with eleven LLMs including GPT-4. The results demonstrate the existence of biasness when evaluating using a single LLM. Also, our PRE model outperforms all the baselines, illustrating the effectiveness of the peer review mechanism.

Every data selection method inherently has a target. In practice, these targets often emerge implicitly through benchmark-driven iteration: researchers develop selection strategies, train models, measure benchmark performance, then refine accordingly. This raises a natural question: what happens when we make this optimization explicit? To explore this, we propose benchmark-targeted ranking (BETR), a simple method that selects pretraining documents based on similarity to benchmark training examples. BETR embeds benchmark examples and a sample of pretraining documents in a shared space, scores this sample by similarity to benchmarks, then trains a lightweight classifier to predict these scores for the full corpus. We compare data selection methods by training over 500 models spanning 10^{19} to 10^{22} FLOPs and fitting scaling laws to them. From this, we find that simply aligning pretraining data to evaluation benchmarks using BETR achieves a 2.1x compute multiplier over DCLM-Baseline (4.7x over unfiltered data) and improves performance on 9 out of 10 tasks across all scales. BETR also generalizes well: when targeting a diverse set of benchmarks disjoint from our evaluation suite, it still matches or outperforms baselines. Our scaling analysis further reveals a clear trend: larger models require less aggressive filtering. Overall, our findings show that directly matching pretraining data to target tasks precisely shapes model capabilities and highlight that optimal selection strategies must adapt to model scale.

Humans regularly engage in analogical thinking, relating personal experiences to current situations (X is analogous to Y because of Z). Analogical thinking allows humans to solve problems in creative ways, grasp difficult concepts, and articulate ideas more effectively. Can language models (LMs) do the same? To answer this question, we propose ANALOBENCH, a benchmark to determine analogical reasoning ability in LMs. Our benchmarking approach focuses on aspects of this ability that are common among humans: (i) recalling related experiences from a large amount of information, and (ii) applying analogical reasoning to complex and lengthy scenarios. We test a broad collection of proprietary models (e.g., GPT family, Claude V2) and open source models such as LLaMA2. As in prior results, scaling up LMs results in some performance boosts. Surprisingly, scale offers minimal gains when, (i) analogies involve lengthy scenarios, or (ii) recalling relevant scenarios from a large pool of information, a process analogous to finding a needle in a haystack. We hope these observations encourage further research in this field.

Evaluating text-to-vision content hinges on two crucial aspects: visual quality and alignment. While significant progress has been made in developing objective models to assess these dimensions, the performance of such models heavily relies on the scale and quality of human annotations. According to Scaling Law, increasing the number of human-labeled instances follows a predictable pattern that enhances the performance of evaluation models. Therefore, we introduce a comprehensive dataset designed to Evaluate Visual quality and Alignment Level for text-to-vision content (Q-EVAL-100K), featuring the largest collection of human-labeled Mean Opinion Scores (MOS) for the mentioned two aspects. The Q-EVAL-100K dataset encompasses both text-to-image and text-to-video models, with 960K human annotations specifically focused on visual quality and alignment for 100K instances (60K images and 40K videos). Leveraging this dataset with context prompt, we propose Q-Eval-Score, a unified model capable of evaluating both visual quality and alignment with special improvements for handling long-text prompt alignment. Experimental results indicate that the proposed Q-Eval-Score achieves superior performance on both visual quality and alignment, with strong generalization capabilities across other benchmarks. These findings highlight the significant value of the Q-EVAL-100K dataset. Data and codes will be available at https://github.com/zzc-1998/Q-Eval.

The current focus of AI research is shifting from emphasizing model training towards enhancing evaluation quality, a transition that is crucial for driving further advancements in AI systems. Traditional evaluation methods typically rely on reward models assigning scalar preference scores to outputs. Although effective, such approaches lack interpretability, leaving users often uncertain about why a reward model rates a particular response as high or low. The advent of LLM-as-a-Judge provides a more scalable and interpretable method of supervision, offering insights into the decision-making process. Moreover, with the emergence of large reasoning models, which consume more tokens for deeper thinking and answer refinement, scaling test-time computation in the LLM-as-a-Judge paradigm presents an avenue for further boosting performance and providing more interpretability through reasoning traces. In this paper, we introduce J1-7B, which is first supervised fine-tuned on reflection-enhanced datasets collected via rejection-sampling and subsequently trained using Reinforcement Learning (RL) with verifiable rewards. At inference time, we apply Simple Test-Time Scaling (STTS) strategies for additional performance improvement. Experimental results demonstrate that J1-7B surpasses the previous state-of-the-art LLM-as-a-Judge by 4.8\% and exhibits a 5.1\% stronger scaling trend under STTS. Additionally, we present three key findings: (1) Existing LLM-as-a-Judge does not inherently exhibit such scaling trend. (2) Model simply fine-tuned on reflection-enhanced datasets continues to demonstrate similarly weak scaling behavior. (3) Significant scaling trend emerges primarily during the RL phase, suggesting that effective STTS capability is acquired predominantly through RL training.

Large Language Models (LLMs) are often described as being instances of foundation models - that is, models that transfer strongly across various tasks and conditions in few-show or zero-shot manner, while exhibiting scaling laws that predict function improvement when increasing the pre-training scale. These claims of excelling in different functions and tasks rely on measurements taken across various sets of standardized benchmarks showing high scores for such models. We demonstrate here a dramatic breakdown of function and reasoning capabilities of state-of-the-art models trained at the largest available scales which claim strong function, using a simple, short, conventional common sense problem formulated in concise natural language, easily solvable by humans. The breakdown is dramatic, as models also express strong overconfidence in their wrong solutions, while providing often non-sensical "reasoning"-like explanations akin to confabulations to justify and backup the validity of their clearly failed responses, making them sound plausible. Various standard interventions in an attempt to get the right solution, like various type of enhanced prompting, or urging the models to reconsider the wrong solutions again by multi step re-evaluation, fail. We take these initial observations to the scientific and technological community to stimulate urgent re-assessment of the claimed capabilities of current generation of LLMs, Such re-assessment also requires common action to create standardized benchmarks that would allow proper detection of such basic reasoning deficits that obviously manage to remain undiscovered by current state-of-the-art evaluation procedures and benchmarks. Code for reproducing experiments in the paper and raw experiments data can be found at https://github.com/LAION-AI/AIW

Large language models (LLMs) are increasingly evaluated and sometimes trained using automated graders such as LLM-as-judges that output scalar scores or preferences. While convenient, these approaches are often opaque: a single score rarely explains why an answer is good or bad, which requirements were missed, or how a system should be improved. This lack of interpretability limits their usefulness for model development, dataset curation, and high-stakes deployment. Query-specific rubric-based evaluation offers a more transparent alternative by decomposing quality into explicit, checkable criteria. However, manually designing high-quality, query-specific rubrics is labor-intensive and cognitively demanding and not feasible for deployment. While previous approaches have focused on generating intermediate rubrics for automated downstream evaluation, it is unclear if these rubrics are both interpretable and effective for human users. In this work, we investigate whether LLMs can generate useful, instance-specific rubrics as compared to human-authored rubrics, while also improving effectiveness for identifying good responses. Through our systematic study on two rubric benchmarks, and on multiple few-shot and post-training strategies, we find that off-the-shelf LLMs produce rubrics that are poorly aligned with human-authored ones. We introduce a simple strategy, RubricRAG, which retrieves domain knowledge via rubrics at inference time from related queries. We demonstrate that RubricRAG can generate more interpretable rubrics both for similarity to human-authored rubrics, and for improved downstream evaluation effectiveness. Our results highlight both the challenges and a promising approach of scalable, interpretable evaluation through automated rubric generation.

As Large Language Models (LLMs) are increasingly proposed for high-stakes medical applications, there has emerged a critical need for reliable and accurate evaluation methodologies. Traditional accuracy metrics fail inadequately as they neither capture question characteristics nor offer topic-specific insights. To address this gap, we introduce MedIRT, a rigorous evaluation framework grounded in Item Response Theory (IRT), the gold standard in high-stakes educational testing. Unlike previous research relying on archival data, we prospectively gathered fresh responses from 80 diverse LLMs on a balanced, 1,100-question USMLE-aligned benchmark. Using one unidimensional two-parameter logistic IRT model per topic, we estimate LLM's latent model ability jointly with question difficulty and discrimination, yielding more stable and nuanced performance rankings than accuracy alone. Notably, we identify distinctive ``spiky'' ability profiles, where overall rankings can be misleading due to highly specialized model abilities. While GPT-5 was the top performer in a majority of domains (8 of 11), it was outperformed in Social Science and Communication by Claude-3-opus, demonstrating that even an overall 23rd-ranked model can hold the top spot for specific competencies. Furthermore, we demonstrate IRT's utility in auditing benchmarks by identifying flawed questions. We synthesize these findings into a practical decision-support framework that integrates our multi-factor competency profiles with operational metrics. This work establishes a robust, psychometrically grounded methodology essential for the safe, effective, and trustworthy deployment of LLMs in healthcare.

The ``LLM-as-a-judge'' paradigm has become a standard method for evaluating open-ended generation. To address the quadratic scalability costs of pairwise comparisons, popular benchmarks like Arena-Hard and AlpacaEval compare all models against a single anchor. However, despite its widespread use, the impact of anchor selection on the reliability of the results remains largely unexplored. In this work, we systematically investigate the effect of anchor selection by evaluating 22 different anchors on the Arena-Hard-v2.0 dataset. We find that the choice of anchor is critical: a poor anchor can dramatically reduce correlation with human rankings. We identify that common anchor choices (best-performing and worst-performing models) make poor anchors. Because these extreme anchors are consistently better or worse than all other models, they are seldom indicative of the relative ranking of the models. We further quantify the effect size of anchor selection, showing it is comparable to the selection of a judge model. We conclude with actionable recommendations. First, we conduct a power analysis, and compute sufficient benchmark sizes for anchor-based evaluation, finding that standard benchmark sizes are insufficient for pairwise evaluation and fail to distinguish between competitive models reliably. Second, we provide guidelines for selecting informative anchors to ensure reliable and efficient evaluation practices.

Scaling test-time computation, or affording a generator large language model (LLM) extra compute during inference, typically employs the help of external non-generative evaluators (i.e., reward models). Concurrently, LLM-judges, models trained to generate evaluations and critiques (explanations) in natural language, are becoming increasingly popular in automatic evaluation. Despite judge empirical successes, their effectiveness as evaluators in test-time scaling settings is largely unknown. In this paper, we introduce the Judge Evaluation for Test-Time Scaling (JETTS) benchmark, which evaluates judge performance in three domains (math reasoning, code generation, and instruction following) under three task settings: response reranking, step-level beam search, and critique-based response refinement. We evaluate 10 different judge models (7B-70B parameters) for 8 different base generator models (6.7B-72B parameters). Our benchmark shows that while judges are competitive with outcome reward models in reranking, they are consistently worse than process reward models in beam search procedures. Furthermore, though unique to LLM-judges, their natural language critiques are currently ineffective in guiding the generator towards better responses.

Finetuning specialized generative evaluators has emerged as a popular paradigm to meet the increasing demand for scalable evaluation during both training and test-time. However, recent work has largely focused on applying new methodology, such as reinforcement learning (RL), to training evaluators, shying away from large-scale, data-driven development. In this work, we focus on data scaling, curating a set of 2.5M samples spanning five unique evaluation tasks (pairwise, step-level, reference-free and reference-based verification, and single rating) and multiple domains focused on reasoning evaluation. With our data, we train Foundational Automatic Reasoning Evaluators (FARE), a family of 8B and 20B (with 3.6B active) parameter evaluators, with a simple iterative rejection-sampling supervised finetuning (SFT) approach. FARE-8B challenges larger specialized RL-trained evaluators and FARE-20B sets the new standard for open-source evaluators, surpassing specialized 70B+ evaluators. Beyond static benchmarks, we evaluate FARE in real-world tasks: As inference-time rerankers, FARE-20B achieves near-oracle performance on MATH. As verifiers in RL training, FARE improves the downstream RL-trained model performance by up to 14.1% vs. string-matching verifiers. When initialized from FARE, a continually-finetuned FARE-Code outperforms gpt-oss-20B by 65% on evaluating test-case quality.

With the surge of large language models (LLMs) and their ability to produce customized output, style-personalized text generation--"write like me"--has become a rapidly growing area of interest. However, style personalization is highly specific, relative to every user, and depends strongly on the pragmatic context, which makes it uniquely challenging. Although prior research has introduced benchmarks and metrics for this area, they tend to be non-standardized and have known limitations (e.g., poor correlation with human subjects). LLMs have been found to not capture author-specific style well, it follows that the metrics themselves must be scrutinized carefully. In this work we critically examine the effectiveness of the most common metrics used in the field, such as BLEU, embeddings, and LLMs-as-judges. We evaluate these metrics using our proposed style discrimination benchmark, which spans eight diverse writing tasks across three evaluation settings: domain discrimination, authorship attribution, and LLM-generated personalized vs non-personalized discrimination. We find strong evidence that employing ensembles of diverse evaluation metrics consistently outperforms single-evaluator methods, and conclude by providing guidance on how to reliably assess style-personalized text generation.

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

Manual curation of evaluation datasets is struggling to keep up with the rapidly expanding capabilities and deployment scenarios of language models. Towards scalable model profiling, we introduce and validate a framework for evaluating LLMs, called Adaptive Evaluations. Adaptive evaluations use scaffolded language models (evaluator agents) to search through a target model's behavior on a domain dataset and create difficult questions (tasks) that can discover and probe the model's failure modes. We find that frontier models lack consistency when adaptively probed with our framework on a diverse suite of datasets and tasks, including but not limited to legal reasoning, forecasting, and online harassment. Generated questions pass human validity checks and often transfer to other models with different capability profiles, demonstrating that adaptive evaluations can also be used to create difficult domain-specific datasets.

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

Large language models can solve substantially harder reasoning problems with more inference-time compute. The term "test-time scaling," however, now covers diverse inference algorithms that extend deliberation along a single trajectory, sample completed candidates and aggregate them through voting or verification, or search over unfinished partial states. These algorithms differ in their statistical structure, compute accounting, and failure modes. Treating these procedures as interchangeable under a single scalar "budget," or reporting accuracy without the inference protocol that produced it, makes results difficult to compare across studies. We develop a systematic account of test-time scaling along three axes. First, we formalize test-time scaling as budgeted inference over the implicit prefix tree of an autoregressive model and distinguish three structural regimes: single-trajectory sequential scaling, leaf-level scaling with terminal reduction, and prefix-level scaling. Second, we treat the evaluated object as the entire inference system and develop evaluation principles that separate end-to-end system performance from candidate-bank diagnostics. We introduce an evaluation profile whose coordinates and simple functionals recover or bound common repeated-sampling metrics, and prescribe protocol-matched reporting of compute and uncertainty. Third, we specify reproducibility requirements for inference protocols, distinguishing exact replay from distributional reproducibility and identifying the artifacts needed to support each. We also organize the open-weight reasoning ecosystem by model-side and interface mechanisms, apply these principles to broad-knowledge, symbolic-reasoning, and competition-mathematics benchmarks, and assemble over 2 billion full reasoning traces for release with progressively richer verifier and token-level signals.

In Machine Learning, a benchmark refers to an ensemble of datasets associated with one or multiple metrics together with a way to aggregate different systems performances. They are instrumental in (i) assessing the progress of new methods along different axes and (ii) selecting the best systems for practical use. This is particularly the case for NLP with the development of large pre-trained models (e.g. GPT, BERT) that are expected to generalize well on a variety of tasks. While the community mainly focused on developing new datasets and metrics, there has been little interest in the aggregation procedure, which is often reduced to a simple average over various performance measures. However, this procedure can be problematic when the metrics are on a different scale, which may lead to spurious conclusions. This paper proposes a new procedure to rank systems based on their performance across different tasks. Motivated by the social choice theory, the final system ordering is obtained through aggregating the rankings induced by each task and is theoretically grounded. We conduct extensive numerical experiments (on over 270k scores) to assess the soundness of our approach both on synthetic and real scores (e.g. GLUE, EXTREM, SEVAL, TAC, FLICKR). In particular, we show that our method yields different conclusions on state-of-the-art systems than the mean-aggregation procedure while being both more reliable and robust.

We present a smoothly broken power law functional form (that we refer to as a Broken Neural Scaling Law (BNSL)) that accurately models & extrapolates the scaling behaviors of deep neural networks (i.e. how the evaluation metric of interest varies as amount of compute used for training (or inference), number of model parameters, training dataset size, model input size, number of training steps, or upstream performance varies) for various architectures & for each of various tasks within a large & diverse set of upstream & downstream tasks, in zero-shot, prompted, & finetuned settings. This set includes large-scale vision, language, audio, video, diffusion, generative modeling, multimodal learning, contrastive learning, AI alignment, AI capabilities, robotics, out-of-distribution (OOD) generalization, continual learning, transfer learning, uncertainty estimation / calibration, OOD detection, adversarial robustness, distillation, sparsity, retrieval, quantization, pruning, fairness, molecules, computer programming/coding, math word problems, "emergent phase transitions", arithmetic, supervised learning, unsupervised/self-supervised learning, & reinforcement learning (single agent & multi-agent). When compared to other functional forms for neural scaling, this functional form yields extrapolations of scaling behavior that are considerably more accurate on this set. Moreover, this functional form accurately models & extrapolates scaling behavior that other functional forms are incapable of expressing such as the nonmonotonic transitions present in the scaling behavior of phenomena such as double descent & the delayed, sharp inflection points present in the scaling behavior of tasks such as arithmetic. Lastly, we use this functional form to glean insights about the limit of the predictability of scaling behavior. Code is available at https://github.com/ethancaballero/broken_neural_scaling_laws

The task of personalized image aesthetic assessment seeks to tailor aesthetic score prediction models to match individual preferences with just a few user-provided inputs. However, the scalability and generalization capabilities of current approaches are considerably restricted by their reliance on an expensive curated database. To overcome this long-standing scalability challenge, we present a unique approach that leverages readily available databases for general image aesthetic assessment and image quality assessment. Specifically, we view each database as a distinct image score regression task that exhibits varying degrees of personalization potential. By determining optimal combinations of task vectors, known to represent specific traits of each database, we successfully create personalized models for individuals. This approach of integrating multiple models allows us to harness a substantial amount of data. Our extensive experiments demonstrate the effectiveness of our approach in generalizing to previously unseen domains-a challenge previous approaches have struggled to achieve-making it highly applicable to real-world scenarios. Our novel approach significantly advances the field by offering scalable solutions for personalized aesthetic assessment and establishing high standards for future research. https://yeolj00.github.io/personal-projects/personalized-aesthetics/