Daily Papers

Recent advances in Omni-Multimodal Large Language Models (Omni-MLLMs) have enabled strong integration of vision, audio, and language. However, their audio-visual intelligence (AVI) remains insufficiently evaluated due to the lack of systematic and comprehensive benchmarks. We introduce AVI-Bench, a cognitively inspired benchmark that evaluates Omni-MLLMs across three stages, perception, understanding, and reasoning, through cross-modal tasks requiring joint audio-visual interpretation. This design enables fine-grained diagnosis of model capabilities and failure modes. To further assess robustness beyond familiar domains, we propose AVI-Bench-PriSe, an extension that probes models' primitive audio-visual sensation using unfamiliar, low-semantic stimuli, testing generalization beyond common training distributions. Extensive experiments on both open-source and closed-source models reveal substantial limitations in current Omni-MLLMs. Based on these findings, we present a four-level AVI taxonomy. Overall, AVI-Bench provides a principled evaluation framework to guide the development of more robust and generalizable AVI. Project website: https://fudancvl.github.io/AVI-Bench/

We evaluate ChatGPT's ability to solve algorithm problems from the CLRS benchmark suite that is designed for GNNs. The benchmark requires the use of a specified classical algorithm to solve a given problem. We find that ChatGPT outperforms specialist GNN models, using Python to successfully solve these problems. This raises new points in the discussion about learning algorithms with neural networks and how we think about what out of distribution testing looks like with web scale training data.

This paper introduces SCHEDBench, a natural-language benchmark for evaluating combinatorial scheduling constraint faithfulness under surface-form variation. Grounded in canonical scheduling instances and solver-derived feasibility and optimality, SCHEDBench assesses whether large language models (LLMs) generate schedules with the same constraint-feasible behavior across varied natural-language (NL) surface forms. SCHEDBench spans 1,132 instances across job-shop scheduling problems (JSP), single and multi-mode resource-constrained project scheduling problems (RCPSP), nurse rostering/scheduling, and curriculum timetabling problems of varying difficulty. Instances are templated into natural language problems using domain-specific templates, themed entities, lexical-syntactic template rephrasing, and constraint-level surface-form variation, with reference solutions verified for feasibility and objective optimality. Across thirteen frontier and open-weight LLMs, we find that models are not reliably invariant to semantically equivalent renderings of the same scheduling problem. Surface-form variation reduces feasibility and induces above-noise shifts in per-instance hard-constraint violations on matched instances. Among the tested isolated axes, constraint reordering yields the clearest above-noise sensitivity.

While generalization over tasks from easy to hard is crucial to profile language models (LLMs), the datasets with fine-grained difficulty annotations for each problem across a broad range of complexity are still blank. Aiming to address this limitation, we present Easy2Hard-Bench, a consistently formatted collection of 6 benchmark datasets spanning various domains, such as mathematics and programming problems, chess puzzles, and reasoning questions. Each problem within these datasets is annotated with numerical difficulty scores. To systematically estimate problem difficulties, we collect abundant performance data on attempts to each problem by humans in the real world or LLMs on the prominent leaderboard. Leveraging the rich performance data, we apply well-established difficulty ranking systems, such as Item Response Theory (IRT) and Glicko-2 models, to uniformly assign numerical difficulty scores to problems. Moreover, datasets in Easy2Hard-Bench distinguish themselves from previous collections by a higher proportion of challenging problems. Through extensive experiments with six state-of-the-art LLMs, we provide a comprehensive analysis of their performance and generalization capabilities across varying levels of difficulty, with the aim of inspiring future research in LLM generalization. The datasets are available at https://huggingface.co/datasets/furonghuang-lab/Easy2Hard-Bench.

Recent advancements in Web agents have introduced novel architectures and benchmarks showcasing progress in autonomous web navigation and interaction. However, most existing benchmarks prioritize effectiveness and accuracy, overlooking factors like safety and trustworthiness which are essential for deploying web agents in enterprise settings. We present STWebAgentBench, a benchmark designed to evaluate web agents safety and trustworthiness across six critical dimensions, essential for reliability in enterprise applications. This benchmark is grounded in a detailed framework that defines safe and trustworthy (ST) agent behavior. Our work extends WebArena with safety templates and evaluation functions to assess safety policy compliance rigorously. We introduce the Completion Under Policy to measure task success while adhering to policies, alongside the Risk Ratio, which quantifies policy violations across dimensions, providing actionable insights to address safety gaps. Our evaluation reveals that current SOTA agents struggle with policy adherence and cannot yet be relied upon for critical business applications. We open-source this benchmark and invite the community to contribute, with the goal of fostering a new generation of safer, more trustworthy AI agents. All code, data, environment reproduction resources, and video demonstrations are available at https://sites.google.com/view/st-webagentbench/home.

Agents are rapidly advancing in automating digital work, but enterprises face a harder challenge: moving beyond prototypes to deployed systems that deliver measurable business value. This path is complicated by fragmented frameworks, slow development, and the absence of standardized evaluation practices. Generalist agents have emerged as a promising direction, excelling on academic benchmarks and offering flexibility across task types, applications, and modalities. Yet, evidence of their use in production enterprise settings remains limited. This paper reports IBM's experience developing and piloting the Computer Using Generalist Agent (CUGA), which has been open-sourced for the community (https://github.com/cuga-project/cuga-agent). CUGA adopts a hierarchical planner--executor architecture with strong analytical foundations, achieving state-of-the-art performance on AppWorld and WebArena. Beyond benchmarks, it was evaluated in a pilot within the Business-Process-Outsourcing talent acquisition domain, addressing enterprise requirements for scalability, auditability, safety, and governance. To support assessment, we introduce BPO-TA, a 26-task benchmark spanning 13 analytics endpoints. In preliminary evaluations, CUGA approached the accuracy of specialized agents while indicating potential for reducing development time and cost. Our contribution is twofold: presenting early evidence of generalist agents operating at enterprise scale, and distilling technical and organizational lessons from this initial pilot. We outline requirements and next steps for advancing research-grade architectures like CUGA into robust, enterprise-ready systems.

Large language model (LLM) agents are increasingly capable of autonomously conducting cyberattacks, posing significant threats to existing applications. This growing risk highlights the urgent need for a real-world benchmark to evaluate the ability of LLM agents to exploit web application vulnerabilities. However, existing benchmarks fall short as they are limited to abstracted Capture the Flag competitions or lack comprehensive coverage. Building a benchmark for real-world vulnerabilities involves both specialized expertise to reproduce exploits and a systematic approach to evaluating unpredictable threats. To address this challenge, we introduce CVE-Bench, a real-world cybersecurity benchmark based on critical-severity Common Vulnerabilities and Exposures. In CVE-Bench, we design a sandbox framework that enables LLM agents to exploit vulnerable web applications in scenarios that mimic real-world conditions, while also providing effective evaluation of their exploits. Our evaluation shows that the state-of-the-art agent framework can resolve up to 13% of vulnerabilities.

The promise of LLM-based user simulators to improve conversational AI is hindered by a critical "realism gap," leading to systems that are optimized for simulated interactions, but may fail to perform well in the real world. We introduce ConvApparel, a new dataset of human-AI conversations designed to address this gap. Its unique dual-agent data collection protocol -- using both "good" and "bad" recommenders -- enables counterfactual validation by capturing a wide spectrum of user experiences, enriched with first-person annotations of user satisfaction. We propose a comprehensive validation framework that combines statistical alignment, a human-likeness score, and counterfactual validation to test for generalization. Our experiments reveal a significant realism gap across all simulators. However, the framework also shows that data-driven simulators outperform a prompted baseline, particularly in counterfactual validation where they adapt more realistically to unseen behaviors, suggesting they embody more robust, if imperfect, user models.

As robotic hardware and learning methods advance, humanoids need tools to perform tasks beyond their inherent physical limits. Successful tool use requires selecting a suitable tool and coordinating manipulation and, when needed, locomotion to complete the task. Existing benchmarks do not jointly evaluate these capabilities on a humanoid. We introduce HumanoidToolBench, an 18-task benchmark spanning three scenarios, three execution levels, and two tool-set modes, together with ToolBook, a dataset of 3.1k demonstrations collected in simulation and on a real Unitree G1. Evaluation of seven policies in simulation and three on the real robot reveals substantial gaps between selecting a suitable tool and completing the task. Focused GR00T N1.7 probes show reduced selection accuracy on unseen tools and continued task execution under unrelated instructions. Code and data are available at https://snu-pi.github.io/HumanoidToolBench/.

In this work, we report the set-up and results of the Liver Tumor Segmentation Benchmark (LiTS), which was organized in conjunction with the IEEE International Symposium on Biomedical Imaging (ISBI) 2017 and the International Conferences on Medical Image Computing and Computer-Assisted Intervention (MICCAI) 2017 and 2018. The image dataset is diverse and contains primary and secondary tumors with varied sizes and appearances with various lesion-to-background levels (hyper-/hypo-dense), created in collaboration with seven hospitals and research institutions. Seventy-five submitted liver and liver tumor segmentation algorithms were trained on a set of 131 computed tomography (CT) volumes and were tested on 70 unseen test images acquired from different patients. We found that not a single algorithm performed best for both liver and liver tumors in the three events. The best liver segmentation algorithm achieved a Dice score of 0.963, whereas, for tumor segmentation, the best algorithms achieved Dices scores of 0.674 (ISBI 2017), 0.702 (MICCAI 2017), and 0.739 (MICCAI 2018). Retrospectively, we performed additional analysis on liver tumor detection and revealed that not all top-performing segmentation algorithms worked well for tumor detection. The best liver tumor detection method achieved a lesion-wise recall of 0.458 (ISBI 2017), 0.515 (MICCAI 2017), and 0.554 (MICCAI 2018), indicating the need for further research. LiTS remains an active benchmark and resource for research, e.g., contributing the liver-related segmentation tasks in http://medicaldecathlon.com/. In addition, both data and online evaluation are accessible via www.lits-challenge.com.

There is a growing line of research on verifying the correctness of language models' outputs. At the same time, LMs are being used to tackle complex queries that require reasoning. We introduce CoverBench, a challenging benchmark focused on verifying LM outputs in complex reasoning settings. Datasets that can be used for this purpose are often designed for other complex reasoning tasks (e.g., QA) targeting specific use-cases (e.g., financial tables), requiring transformations, negative sampling and selection of hard examples to collect such a benchmark. CoverBench provides a diversified evaluation for complex claim verification in a variety of domains, types of reasoning, relatively long inputs, and a variety of standardizations, such as multiple representations for tables where available, and a consistent schema. We manually vet the data for quality to ensure low levels of label noise. Finally, we report a variety of competitive baseline results to show CoverBench is challenging and has very significant headroom. The data is available at https://huggingface.co/datasets/google/coverbench .

Eliciting reasoning capabilities from language models (LMs) is a critical direction on the path towards building intelligent systems. Most recent studies dedicated to reasoning focus on out-of-distribution performance on procedurally-generated synthetic benchmarks, bespoke-built to evaluate specific skills only. This trend makes results hard to transfer across publications, slowing down progress. Three years ago, a similar issue was identified and rectified in the field of neural algorithmic reasoning, with the advent of the CLRS benchmark. CLRS is a dataset generator comprising graph execution traces of classical algorithms from the Introduction to Algorithms textbook. Inspired by this, we propose CLRS-Text -- a textual version of these algorithmic traces. Out of the box, CLRS-Text is capable of procedurally generating trace data for thirty diverse, challenging algorithmic tasks across any desirable input distribution, while offering a standard pipeline in which any additional algorithmic tasks may be created in the benchmark. We fine-tune and evaluate various LMs as generalist executors on this benchmark, validating prior work and revealing a novel, interesting challenge for the LM reasoning community. Our code is available at https://github.com/google-deepmind/clrs/tree/master/clrs/_src/clrs_text.

With the rapid development of AI-based personality and job-related competency assessment, Asynchronous Video Interviews (AVIs) are increasingly used in recruitment. However, existing multimodal personality datasets are often based on short, task-free social media videos and crowdsourced apparent personality labels, which limits their construct validity and relevance to structured interview assessment. To address these limitations, we introduce AVI-Personality, a trait-activated multimodal dataset for personality and job-related competency assessment from AVIs. The dataset contains 3,876 interview videos from 646 participants who completed a simulated management traineeship application. Participants answered two generic questions and four personality-targeted questions designed according to Trait Activation Theory. Our dataset provides both self and observer-reported HEXACO personality traits and job-related competency. We validate AVI-Personality through reliability, construct validity, internal nomological association, fairness, and benchmark analyses. Validation results show that the observer-rated personality traits have moderate to high reliability, especially when ratings are based on personality-targeted questions. Benchmark results show that text-based AI algorithms provide strong personality-relevant cues, while multimodal methods achieve the best overall performance but only modestly outperform text-based baselines. In general, AVI-Personality provides a psychometrically grounded dataset for developing and evaluating AI-based models for personality and competency assessment. The dataset is available are released at https://github.com/APAL-SEU/AVI6

Mathematical reasoning is critical for tasks such as precise distance and area computations, trajectory estimations, and spatial analysis in unmanned aerial vehicle (UAV) based remote sensing, yet current vision-language models (VLMs) have not been adequately tested in this domain. To address this gap, we introduce AVI-Math, the first benchmark to rigorously evaluate multimodal mathematical reasoning in aerial vehicle imagery, moving beyond simple counting tasks to include domain-specific knowledge in areas such as geometry, logic, and algebra. The dataset comprises 3,773 high-quality vehicle-related questions captured from UAV views, covering 6 mathematical subjects and 20 topics. The data, collected at varying altitudes and from multiple UAV angles, reflects real-world UAV scenarios, ensuring the diversity and complexity of the constructed mathematical problems. In this paper, we benchmark 14 prominent VLMs through a comprehensive evaluation and demonstrate that, despite their success on previous multimodal benchmarks, these models struggle with the reasoning tasks in AVI-Math. Our detailed analysis highlights significant limitations in the mathematical reasoning capabilities of current VLMs and suggests avenues for future research. Furthermore, we explore the use of Chain-of-Thought prompting and fine-tuning techniques, which show promise in addressing the reasoning challenges in AVI-Math. Our findings not only expose the limitations of VLMs in mathematical reasoning but also offer valuable insights for advancing UAV-based trustworthy VLMs in real-world applications. The code, and datasets will be released at https://github.com/VisionXLab/avi-math

Fine-tuning language models~(LMs) on human-generated data remains a prevalent practice. However, the performance of such models is often limited by the quantity and diversity of high-quality human data. In this paper, we explore whether we can go beyond human data on tasks where we have access to scalar feedback, for example, on math problems where one can verify correctness. To do so, we investigate a simple self-training method based on expectation-maximization, which we call ReST^{EM}, where we (1) generate samples from the model and filter them using binary feedback, (2) fine-tune the model on these samples, and (3) repeat this process a few times. Testing on advanced MATH reasoning and APPS coding benchmarks using PaLM-2 models, we find that ReST^{EM} scales favorably with model size and significantly surpasses fine-tuning only on human data. Overall, our findings suggest self-training with feedback can substantially reduce dependence on human-generated data.

Full-wave electromagnetic (EM) simulation enables accurate patch-antenna analysis but is computationally expensive for large-scale forward prediction and inverse design. We present a mesh-native, physics-augmented graph-learning framework that treats radiation-pattern prediction as signal reconstruction on an irregular surface mesh. For the forward problem, a GPS graph transformer is trained with Physics-Augmented Intermediate Supervision (PAIS), an auxiliary node-level objective that predicts complex surface currents, the physical intermediate linking geometry to radiation. PAIS improves multiple GNN backbones at no inference-time cost, while shuffled-current and non-physical controls show the gain comes from physical correspondence. Direction-conditioned decoding and a differentiable radiation-integral consistency loss further exploit this structure. On an 80,000-sample CST benchmark, GPS+PAIS reaches MSE 0.17 / PSNR 19.67, generalizes to a PCA split, and transfers zero-shot to canonical patches. For inverse design, surrogate-filtered diffusion beats nearest-neighbor retrieval by 32% relative MSE.

Agents increasingly build on code written by other agents, and they reimplement rather than reuse, growing the codebases later agents must work in. To measure how well agents design libraries for other agents, we introduce LibraryDesignBench, a two-phase benchmark in which an agent implements a full-featured library from a specification that defines required capabilities and potential use cases without prescribing the design. We evaluate the library through the correctness and simplicity of programs written by three user agents from different model families. The benchmark spans 242 expert-validated programming problems across 15 library-design tasks in four languages. On eleven of the fifteen tasks, agent designers reproduce the abstractions of the human-written production library. Downstream agents adopt agent- and human-written libraries alike but underuse them, reimplementing capabilities the library already provides. Our failure analysis finds that downstream agents write extra code mainly because agent-written libraries are rigid or hard to use, not because capabilities are missing. We also experiment with giving designers more prescriptive, agent-first guidance and having them test their library with subagents; this improves downstream scores and yields simpler programs. LibraryDesignBench provides both a testbed for evaluating library-design practices for agent users and an initial design baseline that improves downstream reuse.

Since their initial release, BERT models have demonstrated exceptional performance on a variety of tasks, despite their relatively small size (BERT-base has ~100M parameters). Nevertheless, the architectural choices used in these models are outdated compared to newer transformer-based models such as Llama3 and Qwen3. In recent months, several architectures have been proposed to close this gap. ModernBERT and NeoBERT both show strong improvements on English benchmarks and significantly extend the supported context window. Following their successes, we introduce NeoDictaBERT and NeoDictaBERT-bilingual: BERT-style models trained using the same architecture as NeoBERT, with a dedicated focus on Hebrew texts. These models outperform existing ones on almost all Hebrew benchmarks and provide a strong foundation for downstream tasks. Notably, the NeoDictaBERT-bilingual model shows strong results on retrieval tasks, outperforming other multilingual models of similar size. In this paper, we describe the training process and report results across various benchmarks. We release the models to the community as part of our goal to advance research and development in Hebrew NLP.

Language identification in code-mixed text, largely observed in social media, is highly essential when users frequently switch between multiple languages within a single utterance. Accurately identifying the languages of code-mixed tokens becomes an urgent necessity. Traditional language identification models, designed for monolingual text, are not well suited for token-level language identification in code-mixed settings. We formulate the task as a sequence labeling problem and fine-tune contextual transformer-based models MuRIL and XLM-RoBERTa best suited for Indian languages. We evaluate these systems on three different data configurations (Hindi, Gujarati, and Bengali) to predict language labels for individual tokens. We release a benchmark for language identification in code-mixed tokens with manually annotated test sets. We propose two approaches of code-mixed generation using parallel sentences of three languages. The trained models demonstrate the effectiveness of contextual embeddings for token-level language identification in multilingual social media text. For reproducibility and to facilitate future research, we publicly release our fine-tuned models.

We present a new pre-trained language model (PLM) for modern Hebrew, termed AlephBERTGimmel, which employs a much larger vocabulary (128K items) than standard Hebrew PLMs before. We perform a contrastive analysis of this model against all previous Hebrew PLMs (mBERT, heBERT, AlephBERT) and assess the effects of larger vocabularies on task performance. Our experiments show that larger vocabularies lead to fewer splits, and that reducing splits is better for model performance, across different tasks. All in all this new model achieves new SOTA on all available Hebrew benchmarks, including Morphological Segmentation, POS Tagging, Full Morphological Analysis, NER, and Sentiment Analysis. Subsequently we advocate for PLMs that are larger not only in terms of number of layers or training data, but also in terms of their vocabulary. We release the new model publicly for unrestricted use.

Automated neural network architecture design remains a significant challenge in computer vision. Task diversity and computational constraints require both effective architectures and efficient search methods. Large Language Models (LLMs) present a promising alternative to computationally intensive Neural Architecture Search (NAS), but their application to architecture generation in computer vision has not been systematically studied, particularly regarding prompt engineering and validation strategies. Building on the task-agnostic NNGPT/LEMUR framework, this work introduces and validates two key contributions for computer vision. First, we present Few-Shot Architecture Prompting (FSAP), the first systematic study of the number of supporting examples (n = 1, 2, 3, 4, 5, 6) for LLM-based architecture generation. We find that using n = 3 examples best balances architectural diversity and context focus for vision tasks. Second, we introduce Whitespace-Normalized Hash Validation, a lightweight deduplication method (less than 1 ms) that provides a 100x speedup over AST parsing and prevents redundant training of duplicate computer vision architectures. In large-scale experiments across seven computer vision benchmarks (MNIST, CIFAR-10, CIFAR-100, CelebA, ImageNette, SVHN, Places365), we generated 1,900 unique architectures. We also introduce a dataset-balanced evaluation methodology to address the challenge of comparing architectures across heterogeneous vision tasks. These contributions provide actionable guidelines for LLM-based architecture search in computer vision and establish rigorous evaluation practices, making automated design more accessible to researchers with limited computational resources.

Agentic systems, AI architectures that autonomously execute multi-step workflows to achieve complex goals, are often built using repeated large language model (LLM) calls for closed-set decision tasks such as routing, shortlisting, gating, and verification. While convenient, this design makes deployments slow and expensive due to cumulative latency and token usage. We propose TabAgent, a framework for replacing generative decision components in closed-set selection tasks with a compact textual-tabular classifier trained on execution traces. TabAgent (i) extracts structured schema, state, and dependency features from trajectories (TabSchema), (ii) augments coverage with schema-aligned synthetic supervision (TabSynth), and (iii) scores candidates with a lightweight classifier (TabHead). On the long-horizon AppWorld benchmark, TabAgent maintains task-level success while eliminating shortlist-time LLM calls, reducing latency by approximately 95% and inference cost by 85-91%. Beyond tool shortlisting, TabAgent generalizes to other agentic decision heads, establishing a paradigm for learned discriminative replacements of generative bottlenecks in production agent architectures.

We present DictaBERT, a new state-of-the-art pre-trained BERT model for modern Hebrew, outperforming existing models on most benchmarks. Additionally, we release two fine-tuned versions of the model, designed to perform two specific foundational tasks in the analysis of Hebrew texts: prefix segmentation and morphological tagging. These fine-tuned models allow any developer to perform prefix segmentation and morphological tagging of a Hebrew sentence with a single call to a HuggingFace model, without the need to integrate any additional libraries or code. In this paper we describe the details of the training as well and the results on the different benchmarks. We release the models to the community, along with sample code demonstrating their use. We release these models as part of our goal to help further research and development in Hebrew NLP.

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

Training large language models (LLMs) in low-resource languages such as Hebrew poses unique challenges. In this paper, we introduce DictaLM2.0 and DictaLM2.0-Instruct, two LLMs derived from the Mistral model, trained on a substantial corpus of approximately 200 billion tokens in both Hebrew and English. Adapting a pre-trained model to a new language involves specialized techniques that differ significantly from training a model from scratch or further training existing models on well-resourced languages such as English. We outline these novel training methodologies, which facilitate effective learning and adaptation to the linguistic properties of Hebrew. Additionally, we fine-tuned DictaLM2.0-Instruct on a comprehensive instruct dataset to enhance its performance on task-specific instructions. To rigorously evaluate our models, we introduce a new benchmark suite for Hebrew LLM evaluation, covering a diverse set of tasks including Question Answering, Sentiment Analysis, Winograd Schema Challenge, Translation, and Summarization. Our work not only addresses the intricacies of training LLMs in low-resource languages but also proposes a framework that can be leveraged for adapting other LLMs to various non-English languages, contributing to the broader field of multilingual NLP.

Multi-document (MD) processing is crucial for LLMs to handle real-world tasks such as summarization and question-answering across large sets of documents. While LLMs have improved at processing long inputs, MD contexts still present challenges, such as managing inter-document dependencies, redundancy, and incoherent structures. We introduce MDCure, a scalable and effective fine-tuning pipeline to enhance the MD capabilities of LLMs without the computational cost of pre-training or reliance on human annotated data. MDCure is based on generation of high-quality synthetic MD instruction data from sets of related articles via targeted prompts. We further introduce MDCureRM, a multi-objective reward model which filters generated data based on their training utility for MD settings. With MDCure, we fine-tune a variety of LLMs, from the FlanT5, Qwen2, and LLAMA3.1 model families, up to 70B parameters in size. Extensive evaluations on a wide range of MD and long-context benchmarks spanning various tasks show MDCure consistently improves performance over pre-trained baselines and over corresponding base models by up to 75.5%. Our code, datasets, and models are available at https://github.com/yale-nlp/MDCure.

Retrieval Augmented Generation (RAG) is a commonly used approach for enhancing large language models (LLMs) with relevant and up-to-date information. However, the retrieved sources can often contain conflicting information and it remains unclear how models should address such discrepancies. In this work, we first propose a novel taxonomy of knowledge conflict types in RAG, along with the desired model behavior for each type. We then introduce CONFLICTS, a high-quality benchmark with expert annotations of conflict types in a realistic RAG setting. CONFLICTS is the first benchmark that enables tracking progress on how models address a wide range of knowledge conflicts. We conduct extensive experiments on this benchmark, showing that LLMs often struggle to appropriately resolve conflicts between sources. While prompting LLMs to explicitly reason about the potential conflict in the retrieved documents significantly improves the quality and appropriateness of their responses, substantial room for improvement in future research remains.

We present Foundation-Sec-8B-Reasoning, the first open-source native reasoning model for cybersecurity. Built upon our previously released Foundation-Sec-8B base model (derived from Llama-3.1-8B-Base), the model is trained through a two-stage process combining supervised fine-tuning (SFT) and reinforcement learning from verifiable rewards (RLVR). Our training leverages proprietary reasoning data spanning cybersecurity analysis, instruction-following, and mathematical reasoning. Evaluation across 10 cybersecurity benchmarks and 10 general-purpose benchmarks demonstrates performance competitive with significantly larger models on cybersecurity tasks while maintaining strong general capabilities. The model shows effective generalization on multi-hop reasoning tasks and strong safety performance when deployed with appropriate system prompts and guardrails. This work demonstrates that domain-specialized reasoning models can achieve strong performance on specialized tasks while maintaining broad general capabilities. We release the model publicly at https://huggingface.co/fdtn-ai/Foundation-Sec-8B-Reasoning.

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

A critical component in the trustworthiness of LLMs is reliable uncertainty communication, yet LLMs often use assertive language when conveying false claims, leading to over-reliance and eroded trust. We present the first systematic study of faithful confidence calibration of LLMs, benchmarking models' ability to use linguistic expressions of uncertainty that faithfully reflect their intrinsic uncertainty, across a comprehensive array of models, datasets, and prompting strategies. Our results demonstrate that LLMs largely fail at this task, and that existing interventions are insufficient: standard prompt approaches provide only marginal gains, and existing, factuality-based calibration techniques can even harm faithful calibration. To address this critical gap, we introduce MetaFaith, a novel prompt-based calibration approach inspired by human metacognition. We show that MetaFaith robustly improves faithful calibration across diverse models and task domains, enabling up to 61% improvement in faithfulness and achieving an 83% win rate over original generations as judged by humans.

As transformer-based large language models (LLMs) increasingly permeate society, they have revolutionized domains such as software engineering, creative writing, and digital arts. However, their adoption in cybersecurity remains limited due to challenges like scarcity of specialized training data and complexity of representing cybersecurity-specific knowledge. To address these gaps, we present Foundation-Sec-8B, a cybersecurity-focused LLM built on the Llama 3.1 architecture and enhanced through continued pretraining on a carefully curated cybersecurity corpus. We evaluate Foundation-Sec-8B across both established and new cybersecurity benchmarks, showing that it matches Llama 3.1-70B and GPT-4o-mini in certain cybersecurity-specific tasks. By releasing our model to the public, we aim to accelerate progress and adoption of AI-driven tools in both public and private cybersecurity contexts.

We demonstrate that a neural network pre-trained on text and fine-tuned on code solves mathematics course problems, explains solutions, and generates new questions at a human level. We automatically synthesize programs using few-shot learning and OpenAI's Codex transformer and execute them to solve course problems at 81% automatic accuracy. We curate a new dataset of questions from MIT's largest mathematics courses (Single Variable and Multivariable Calculus, Differential Equations, Introduction to Probability and Statistics, Linear Algebra, and Mathematics for Computer Science) and Columbia University's Computational Linear Algebra. We solve questions from a MATH dataset (on Prealgebra, Algebra, Counting and Probability, Intermediate Algebra, Number Theory, and Precalculus), the latest benchmark of advanced mathematics problems designed to assess mathematical reasoning. We randomly sample questions and generate solutions with multiple modalities, including numbers, equations, and plots. The latest GPT-3 language model pre-trained on text automatically solves only 18.8% of these university questions using zero-shot learning and 30.8% using few-shot learning and the most recent chain of thought prompting. In contrast, program synthesis with few-shot learning using Codex fine-tuned on code generates programs that automatically solve 81% of these questions. Our approach improves the previous state-of-the-art automatic solution accuracy on the benchmark topics from 8.8% to 81.1%. We perform a survey to evaluate the quality and difficulty of generated questions. This work is the first to automatically solve university-level mathematics course questions at a human level and the first work to explain and generate university-level mathematics course questions at scale, a milestone for higher education.

Tabular data underpins numerous high-impact applications of machine learning from fraud detection to genomics and healthcare. Classical approaches to solving tabular problems, such as gradient boosting and random forests, are widely used by practitioners. However, recent deep learning methods have achieved a degree of performance competitive with popular techniques. We devise a hybrid deep learning approach to solving tabular data problems. Our method, SAINT, performs attention over both rows and columns, and it includes an enhanced embedding method. We also study a new contrastive self-supervised pre-training method for use when labels are scarce. SAINT consistently improves performance over previous deep learning methods, and it even outperforms gradient boosting methods, including XGBoost, CatBoost, and LightGBM, on average over a variety of benchmark tasks.

LLMs are highly sensitive to prompt phrasing, yet standard benchmarks typically report performance using a single prompt, raising concerns about the reliability of such evaluations. In this work, we argue for a stochastic method of moments evaluation over the space of meaning-preserving prompt perturbations. We introduce a formal definition of reliable evaluation that accounts for prompt sensitivity, and suggest ReliableEval - a method for estimating the number of prompt resamplings needed to obtain meaningful results. Using our framework, we stochastically evaluate five frontier LLMs and find that even top-performing models like GPT-4o and Claude-3.7-Sonnet exhibit substantial prompt sensitivity. Our approach is model-, task-, and metric-agnostic, offering a recipe for meaningful and robust LLM evaluation.

Image reconstruction from corrupted images is crucial across many domains. Most reconstruction networks are trained on post-ISP sRGB images, even though the image-signal-processing pipeline irreversibly mixes colors, clips dynamic range, and blurs fine detail. This paper uses the rain degradation problem as a use case to show that these losses are avoidable, and demonstrates that learning directly on raw Bayer mosaics yields superior reconstructions. To substantiate the claim, we (i) evaluate post-ISP and Bayer reconstruction pipelines, (ii) curate Raw-Rain, the first public benchmark of real rainy scenes captured in both 12-bit Bayer and bit-depth-matched sRGB, and (iii) introduce Information Conservation Score (ICS), a color-invariant metric that aligns more closely with human opinion than PSNR or SSIM. On the test split, our raw-domain model improves sRGB results by up to +0.99 dB PSNR and +1.2% ICS, while running faster with half of the GFLOPs. The results advocate an ISP-last paradigm for low-level vision and open the door to end-to-end learnable camera pipelines.

This is the system card published alongside the OpenAI GPT-5 launch, August 2025. GPT-5 is a unified system with a smart and fast model that answers most questions, a deeper reasoning model for harder problems, and a real-time router that quickly decides which model to use based on conversation type, complexity, tool needs, and explicit intent (for example, if you say 'think hard about this' in the prompt). The router is continuously trained on real signals, including when users switch models, preference rates for responses, and measured correctness, improving over time. Once usage limits are reached, a mini version of each model handles remaining queries. This system card focuses primarily on gpt-5-thinking and gpt-5-main, while evaluations for other models are available in the appendix. The GPT-5 system not only outperforms previous models on benchmarks and answers questions more quickly, but -- more importantly -- is more useful for real-world queries. We've made significant advances in reducing hallucinations, improving instruction following, and minimizing sycophancy, and have leveled up GPT-5's performance in three of ChatGPT's most common uses: writing, coding, and health. All of the GPT-5 models additionally feature safe-completions, our latest approach to safety training to prevent disallowed content. Similarly to ChatGPT agent, we have decided to treat gpt-5-thinking as High capability in the Biological and Chemical domain under our Preparedness Framework, activating the associated safeguards. While we do not have definitive evidence that this model could meaningfully help a novice to create severe biological harm -- our defined threshold for High capability -- we have chosen to take a precautionary approach.

We introduce the task of Visual Dialog, which requires an AI agent to hold a meaningful dialog with humans in natural, conversational language about visual content. Specifically, given an image, a dialog history, and a question about the image, the agent has to ground the question in image, infer context from history, and answer the question accurately. Visual Dialog is disentangled enough from a specific downstream task so as to serve as a general test of machine intelligence, while being grounded in vision enough to allow objective evaluation of individual responses and benchmark progress. We develop a novel two-person chat data-collection protocol to curate a large-scale Visual Dialog dataset (VisDial). VisDial v0.9 has been released and contains 1 dialog with 10 question-answer pairs on ~120k images from COCO, with a total of ~1.2M dialog question-answer pairs. We introduce a family of neural encoder-decoder models for Visual Dialog with 3 encoders -- Late Fusion, Hierarchical Recurrent Encoder and Memory Network -- and 2 decoders (generative and discriminative), which outperform a number of sophisticated baselines. We propose a retrieval-based evaluation protocol for Visual Dialog where the AI agent is asked to sort a set of candidate answers and evaluated on metrics such as mean-reciprocal-rank of human response. We quantify gap between machine and human performance on the Visual Dialog task via human studies. Putting it all together, we demonstrate the first 'visual chatbot'! Our dataset, code, trained models and visual chatbot are available on https://visualdialog.org

Self-correction is a highly desirable capability of large language models (LLMs), yet it has consistently been found to be largely ineffective in modern LLMs. Existing approaches for training self-correction either require multiple models or rely on a more capable model or other forms of supervision. To this end, we develop a multi-turn online reinforcement learning (RL) approach, SCoRe, that significantly improves an LLM's self-correction ability using entirely self-generated data. To build SCoRe, we first show that variants of supervised fine-tuning (SFT) on offline model-generated correction traces are insufficient for instilling self-correction behavior. In particular, we observe that training via SFT either suffers from a distribution mismatch between the training data and the model's own responses or implicitly prefers only a certain mode of correction behavior that is often not effective at test time. SCoRe addresses these challenges by training under the model's own distribution of self-generated correction traces and using appropriate regularization to steer the learning process into learning a self-correction strategy that is effective at test time as opposed to simply fitting high-reward responses for a given prompt. This regularization prescribes running a first phase of RL on a base model to generate a policy initialization that is less susceptible to collapse and then using a reward bonus to amplify self-correction during training. When applied to Gemini 1.0 Pro and 1.5 Flash models, we find that SCoRe achieves state-of-the-art self-correction performance, improving the base models' self-correction by 15.6% and 9.1% respectively on the MATH and HumanEval benchmarks.

Recent state-of-the-art natural language understanding models, such as BERT and XLNet, score a pair of sentences (A and B) using multiple cross-attention operations - a process in which each word in sentence A attends to all words in sentence B and vice versa. As a result, computing the similarity between a query sentence and a set of candidate sentences, requires the propagation of all query-candidate sentence-pairs throughout a stack of cross-attention layers. This exhaustive process becomes computationally prohibitive when the number of candidate sentences is large. In contrast, sentence embedding techniques learn a sentence-to-vector mapping and compute the similarity between the sentence vectors via simple elementary operations. In this paper, we introduce Distilled Sentence Embedding (DSE) - a model that is based on knowledge distillation from cross-attentive models, focusing on sentence-pair tasks. The outline of DSE is as follows: Given a cross-attentive teacher model (e.g. a fine-tuned BERT), we train a sentence embedding based student model to reconstruct the sentence-pair scores obtained by the teacher model. We empirically demonstrate the effectiveness of DSE on five GLUE sentence-pair tasks. DSE significantly outperforms several ELMO variants and other sentence embedding methods, while accelerating computation of the query-candidate sentence-pairs similarities by several orders of magnitude, with an average relative degradation of 4.6% compared to BERT. Furthermore, we show that DSE produces sentence embeddings that reach state-of-the-art performance on universal sentence representation benchmarks. Our code is made publicly available at https://github.com/microsoft/Distilled-Sentence-Embedding.

This paper presents our ongoing work toward developing an enterprise-ready Computer Using Generalist Agent (CUGA) system. Our research highlights the evolutionary nature of building agentic systems suitable for enterprise environments. By integrating state-of-the-art agentic AI techniques with a systematic approach to iterative evaluation, analysis, and refinement, we have achieved rapid and cost-effective performance gains, notably reaching a new state-of-the-art performance on the WebArena benchmark. We detail our development roadmap, the methodology and tools that facilitated rapid learning from failures and continuous system refinement, and discuss key lessons learned and future challenges for enterprise adoption.

Generalization in large language models (LLMs) is the ability to produce consistent and semantically stable outputs when the same input is expressed in different ways. Existing work typically evaluates generalization through aggregate accuracy on a single prompt format, task, or set of variations, which conflates robustness with overall benchmark performance. In this work, we show generalization evaluation at the level of individual examples, across multiple input variants, and across different aspects of model behavior, focusing on variability rather than reducing performance to a score that can be improved through narrow training or other ways that obfuscate generalization evaluation. Following this view, we introduce the Stability-Aware Generalization Objective (SAGO), a framework that measures how much model behavior changes for the same input under different variations and benchmarks, capturing variability across several dimensions including generation consistency, internal activations, confidence, and response mirroring. We show that many commonly used models exhibit statistically significant and consistent generalization instability: no model generalizes uniformly, behavioral axes capture independent failure modes, and cross-dataset variation can reverse model rankings.

In this report, we introduce the Gemini 2.X model family: Gemini 2.5 Pro and Gemini 2.5 Flash, as well as our earlier Gemini 2.0 Flash and Flash-Lite models. Gemini 2.5 Pro is our most capable model yet, achieving SoTA performance on frontier coding and reasoning benchmarks. In addition to its incredible coding and reasoning skills, Gemini 2.5 Pro is a thinking model that excels at multimodal understanding and it is now able to process up to 3 hours of video content. Its unique combination of long context, multimodal and reasoning capabilities can be combined to unlock new agentic workflows. Gemini 2.5 Flash provides excellent reasoning abilities at a fraction of the compute and latency requirements and Gemini 2.0 Flash and Flash-Lite provide high performance at low latency and cost. Taken together, the Gemini 2.X model generation spans the full Pareto frontier of model capability vs cost, allowing users to explore the boundaries of what is possible with complex agentic problem solving.

Visually Impaired Assistance (VIA) aims to automatically help the visually impaired (VI) handle daily activities. The advancement of VIA primarily depends on developments in Computer Vision (CV) and Natural Language Processing (NLP), both of which exhibit cutting-edge paradigms with large models (LMs). Furthermore, LMs have shown exceptional multimodal abilities to tackle challenging physically-grounded tasks such as embodied robots. To investigate the potential and limitations of state-of-the-art (SOTA) LMs' capabilities in VIA applications, we present an extensive study for the task of VIA with LMs (VIALM). In this task, given an image illustrating the physical environments and a linguistic request from a VI user, VIALM aims to output step-by-step guidance to assist the VI user in fulfilling the request grounded in the environment. The study consists of a survey reviewing recent LM research and benchmark experiments examining selected LMs' capabilities in VIA. The results indicate that while LMs can potentially benefit VIA, their output cannot be well environment-grounded (i.e., 25.7% GPT-4's responses) and lacks fine-grained guidance (i.e., 32.1% GPT-4's responses).

Fusion-in-Decoder (FiD) is an effective retrieval-augmented language model applied across a variety of open-domain tasks, such as question answering, fact checking, etc. In FiD, supporting passages are first retrieved and then processed using a generative model (Reader), which can cause a significant bottleneck in decoding time, particularly with long outputs. In this work, we analyze the contribution and necessity of all the retrieved passages to the performance of reader models, and propose eliminating some of the retrieved information, at the token level, that might not contribute essential information to the answer generation process. We demonstrate that our method can reduce run-time by up to 62.2%, with only a 2% reduction in performance, and in some cases, even improve the performance results.