NeurIPS 2025 Papers

The rapid development of multilingual large language models (LLMs) highlights the need for high-quality, diverse, and well-curated multilingual datasets. In this paper, we introduce DCAD-2000 (Data Cleaning as Anomaly Detection), a large-scale multilingual corpus constructed from newly extracted Common Crawl data and existing multilingual sources. DCAD-2000 covers 2,282 languages, 46.72TB of text, and 8.63 billion documents, spanning 155 high- and medium-resource languages and 159 writing scripts. To overcome the limitations of existing data cleaning approaches, which rely on manually designed heuristic thresholds, we reframe data cleaning as an anomaly detection problem. This dynamic filtering paradigm substantially improves data quality by automatically identifying and removing noisy or anomalous content. By fine-tuning LLMs on DCAD-2000, we demonstrate notable improvements in data quality, robustness of the cleaning pipeline, and downstream performance, particularly for low-resource languages across multiple multilingual benchmarks.

We study the implicit bias of flatness / low (loss) curvature and its effects on generalization in two-layer overparameterized ReLU networks with multivariate inputs---a problem well motivated by the minima stability and edge-of-stability phenomena in gradient-descent training. Existing work either requires interpolation or focuses only on univariate inputs. This paper presents new and somewhat surprising theoretical results for multivariate inputs. On two natural settings (1) generalization gap for flat solutions, and (2) mean-squared error (MSE) in nonparametric function estimation by stable minima, we prove upper and lower bounds, which establish that while flatness does imply generalization, the resulting rates of convergence necessarily deteriorate exponentially as the input dimension grows. This gives an exponential separation between the flat solutions compared to low-norm solutions (i.e., weight decay), which are known not to suffer from the curse of dimensionality. In particular, our minimax lower bound construction, based on a novel packing argument with boundary-localized ReLU neurons, reveals how flat solutions can exploit a kind of "neural shattering" where neurons rarely activate, but with high weight magnitudes. This leads to poor performance in high dimensions. We corroborate these theoretical findings with extensive numerical simulations. To the best of our knowledge, our analysis provides the first systematic explanation for why flat minima may fail to generalize in high dimensions.

We show that across architecture (Transformer vs. Mamba vs. RWKV), training dataset (OpenWebText vs. The Pile), and scale (14 million parameters to 12 billion parameters), autoregressive language models exhibit highly consistent patterns of change in their behavior over the course of pretraining. Based on our analysis of over 1,400 language model checkpoints on over 110,000 tokens of English, we find that up to 98\% of the variance in language model behavior at the word level can be explained by three simple heuristics: the unigram probability (frequency) of a given word, the $n$-gram probability of the word, and the semantic similarity between the word and its context. Furthermore, we see consistent behavioral phases in all language models, with their predicted probabilities for words overfitting to those words' $n$-gram probabilities for increasing $n$ over the course of training. Taken together, these results suggest that learning in neural language models may follow a similar trajectory irrespective of model details.

This work introduces Structured Linear Controlled Differential Equations (SLiCEs), a unifying framework for sequence models with structured, input-dependent state-transition matrices that retain the maximal expressivity of dense matrices whilst being cheaper to compute. The framework encompasses existing architectures, such as input-dependent block-diagonal linear recurrent neural networks and DeltaNet's diagonal-plus-low-rank structure, as well as two novel variants based on sparsity and the Walsh-Hadamard transform. We prove that, unlike the diagonal state-transition matrices of S4D and Mamba, SLiCEs employing block-diagonal, sparse, or Walsh-Hadamard matrices match the maximal expressivity of dense matrices. Empirically, SLiCEs solve the $A_5$ state-tracking benchmark with a single layer, achieve best-in-class length generalisation on regular language tasks among parallel-in-time models, and match the performance of log neural controlled differential equations on six multivariate time-series classification datasets while cutting the average time per training step by a factor of twenty.

With video games leading in entertainment revenues, optimizing game development workflows is critical to the industry’s long-term success. Recent advances in vision-language models (VLMs) hold significant potential to automate and enhance various aspects of game development—particularly video game quality assurance (QA), which remains one of the most labor-intensive processes with limited automation. To effectively measure VLM performance in video game QA tasks and evaluate their ability to handle real-world scenarios, there is a clear need for standardized benchmarks, as current ones fall short in addressing this domain. To bridge this gap, we introduce VideoGameQA-Bench - a comprehensive benchmark designed to encompass a wide range of game QA activities, including visual unit testing, visual regression testing, needle-in-a-haystack, glitch detection, and bug report generation for both images and videos.

Vision-language models (VLMs) trained on internet-scale data achieve remarkable zero-shot detection performance on common objects like car, truck, and pedestrian. However, state-of-the-art models still struggle to generalize to out-of-distribution classes, tasks and imaging modalities not typically found in their pre-training. Rather than simply re-training VLMs on more visual data, we argue that one should align VLMs to new concepts with annotation instructions containing a few visual examples and rich textual descriptions. To this end, we introduce Roboflow100-VL, a large-scale collection of 100 multi-modal object detection datasets with diverse concepts not commonly found in VLM pre-training. We evaluate state-of-the-art models on our benchmark in zero-shot, few-shot, semi-supervised, and fully-supervised settings, allowing for comparison across data regimes. Notably, we find that VLMs like GroundingDINO and Qwen2.5-VL achieve less than 2% zero-shot accuracy on challenging medical imaging datasets within Roboflow100-VL, demonstrating the need for few-shot concept alignment. Lastly, we discuss our recent CVPR 2025 Foundational FSOD competition and share insights from the community. Notably, the winning team significantly outperforms our baseline by 17 mAP! Our code and dataset are available on GitHub and Roboflow.

Recent advances in LVLMs have improved vision-language understanding, but they still struggle with spatial perception, limiting their ability to reason about complex 3D scenes. Unlike previous approaches that incorporate 3D representations into models to improve spatial understanding, we aim to unlock the potential of VLMs by leveraging spatially relevant image data. To this end, we introduce a novel 2D spatial data generation and annotation pipeline built upon scene data with 3D ground-truth. This pipeline enables the creation of a diverse set of spatial tasks, ranging from basic perception tasks to more complex reasoning tasks. Leveraging this pipeline, we construct SPAR-7M, a large-scale dataset generated from thousands of scenes across multiple public datasets. In addition, we introduce SPAR-Bench, a benchmark designed to offer a more comprehensive evaluation of spatial capabilities compared to existing spatial benchmarks, supporting both single-view and multi-view inputs. Training on both SPAR-7M and large-scale 2D datasets enables our models to achieve state-of-the-art performance on 2D spatial benchmarks. Further fine-tuning on 3D task-specific datasets yields competitive results, underscoring the effectiveness of our dataset in enhancing spatial reasoning.

Although data that can be naturally represented as graphs is widespread in real-world applications across diverse industries, popular graph ML benchmarks for node property prediction only cover a surprisingly narrow set of data domains, and graph neural networks (GNNs) are often evaluated on just a few academic citation networks. This issue is particularly pressing in light of the recent growing interest in designing graph foundation models. These models are supposed to be able to transfer to diverse graph datasets from different domains, and yet the proposed graph foundation models are often evaluated on a very limited set of datasets from narrow applications. To alleviate this issue, we introduce GraphLand: a benchmark of 14 diverse graph datasets for node property prediction from a range of different industrial applications. GraphLand allows evaluating graph ML models on a wide range of graphs with diverse sizes, structural characteristics, and feature sets, all in a unified setting. Further, GraphLand allows investigating such previously underexplored research questions as how realistic temporal distributional shifts under transductive and inductive settings influence graph ML model performance. To mimic realistic industrial settings, we use GraphLand to compare GNNs with gradient-boosted decision trees (GBDT) models that are popular in industrial applications and show that GBDTs provided with additional graph-based input features can sometimes be very strong baselines. Further, we evaluate currently available general-purpose graph foundation models and find that they fail to produce competitive results on our proposed datasets.

Robust unlearning is crucial for safely deploying large language models (LLMs) in environments where data privacy, model safety, and regulatory compliance must be ensured. Yet the task is inherently challenging, partly due to difficulties in reliably measuring whether unlearning has truly occurred. Moreover, fragmentation in current methodologies and inconsistent evaluation metrics hinder comparative analysis and reproducibility. To unify and accelerate research efforts, we introduce OpenUnlearning, a standardized and extensible framework designed explicitly for benchmarking both LLM unlearning methods and metrics. OpenUnlearning integrates 13 state-of-the-art unlearning algorithms and 16 diverse evaluations across 3 leading benchmarks (TOFU, MUSE, and WMDP) and also enables analyses of forgetting behaviors across 450+ publicly released checkpoints. Leveraging OpenUnlearning, we propose a novel meta-evaluation benchmark focused specifically on assessing the faithfulness and robustness of evaluation metrics themselves. We also benchmark diverse unlearning methods and provide a comparative analysis against an extensive evaluation suite. Overall, we establish a clear, community-driven pathway toward rigorous development in LLM unlearning research.

We introduce Reasoning Gym, a library of reasoning environments for reinforcement learning with verifiable rewards (RLVR). It provides over 100 tasks spanning multiple domains including algebra, arithmetic, computation, cognition, geometry, graph theory, logic, and various common games. Its key innovation is the ability to generate virtually infinite training data with adjustable complexity, unlike most previous reasoning datasets, which are typically fixed. This procedural generation approach allows for continuous evaluation across varying difficulty levels and task configurations. Our experimental results demonstrate the efficacy of Reasoning Gym in both evaluating and reinforcement learning of reasoning models.

Power flow (PF) calculations are the backbone of real-time grid operations, across workflows such as contingency analysis (where repeated PF evaluations assess grid security under outages) and topology optimization (which involves PF-based searches over combinatorially large action spaces). Running these calculations at operational timescales or across large evaluation spaces remains a major computational bottleneck. Additionally, growing uncertainty in power system operations from the integration of renewables and climate-induced extreme weather also calls for tools that can accurately and efficiently simulate a wide range of scenarios and operating conditions. Machine learning methods offer a potential speedup over traditional solvers, but their performance has not been systematically assessed on benchmarks that capture real-world variability. This paper introduces PF∆, a benchmark dataset for power flow that captures diverse variations in load, generation, and topology. PF∆ contains 859,800 solved power flow instances spanning six different bus system sizes, capturing three types of contingency scenarios (N , N –1, and N –2), and including close-to-infeasible cases near steady-state voltage stability limits. We evaluate traditional solvers and GNN-based methods, highlighting key areas where existing approaches struggle, and identifying open problems for future research. Our dataset is available at https://huggingface.co/datasets/pfdelta/pfdelta/tree/main and our code with data generation scripts and model implementations is at https: //github.com/MOSSLab-MIT/pfdelta

Dark matter makes up approximately 85\% of total matter in our universe, yet it has never been directly observed in any laboratory on Earth. The origin of dark matter is one of the most important questions in contemporary physics, and a convincing detection of dark matter would be a Nobel-Prize-level breakthrough in fundamental science. The ABRACADABRA experiment was specifically designed to search for dark matter. Although it has not yet made a discovery, ABRACADABRA has produced several dark matter search results widely endorsed by the physics community. The experiment generates ultra-long time-series data at a rate of 10 million samples per second, where the dark matter signal would manifest itself as a sinusoidal oscillation mode within the ultra-long time series. In this paper, we present the TIDMAD --- a comprehensive data release from the ABRACADABRA experiment including three key components: an ultra-long time series dataset divided into training, validation, and science subsets; a carefully-designed denoising score for direct model benchmarking; and a complete analysis framework which produces a physics community-standard dark matter search result suitable for publication as a physics paper. This data release enables core AI algorithms to extract the dark matter signal and produce real physics results thereby advancing fundamental science.

Gaussian processes (GPs) play an essential role in biostatistics, scientific machine learning, and Bayesian optimization for their ability to provide probabilistic predictions and model uncertainty. However, GP inference struggles to scale to large datasets (which are common in modern applications), since it requires the solution of a linear system whose size scales quadratically with the number of samples in the dataset. We propose an approximate, distributed, accelerated sketch-and-project algorithm ($\texttt{ADASAP}$) for solving these linear systems, which improves scalability. We use the theory of determinantal point processes to show that the posterior mean induced by sketch-and-project rapidly converges to the true posterior mean. In particular, this yields the first efficient, condition number-free algorithm for estimating the posterior mean along the top spectral basis functions, showing that our approach is principled for GP inference. $\texttt{ADASAP}$ outperforms state-of-the-art solvers based on conjugate gradient and coordinate descent across several benchmark datasets and a large-scale Bayesian optimization task. Moreover, $\texttt{ADASAP}$ scales to a dataset with $> 3 \cdot 10^8$ samples, a feat which has not been accomplished in the literature.

The rapid advancement of Large Vision-Language Models (VLMs), both general-domain models and those specifically tailored for remote sensing, has demonstrated exceptional perception and reasoning capabilities in Earth observation tasks. However, a benchmark for systematically evaluating their capabilities in this domain is still lacking. To bridge this gap, we propose CHOICE, an extensive benchmark designed to objectively evaluate the hierarchical remote sensing capabilities of VLMs. Focusing on 2 primary capability dimensions essential to remote sensing: perception and reasoning, we further categorize 6 secondary dimensions and 23 leaf tasks to ensure a well-rounded assessment coverage. CHOICE guarantees the quality of all 10,507 problems through a rigorous process of data collection from 50 globally distributed cities, question construction, and quality control. The newly curated data and the format of multiple-choice questions with definitive answers allow for an objective and straightforward performance assessment. Our evaluation of 3 proprietary and 21 open-source VLMs highlights their critical limitations within this specialized context. We hope that CHOICE will serve as a valuable resource and offer deeper insights into the challenges and potential of VLMs in the field of remote sensing. Code and dataset are available at [this https URL](https://github.com/ShawnAn-WHU/CHOICE).

We introduce MIRAGE, a new benchmark for multimodal expert-level reasoning and decision-making in consultative interaction settings. Designed for the domain of agriculture, MIRAGE captures the full complexity of expert consultations by combining natural user queries, expert-authored responses, and image-based context, offering a high-fidelity benchmark for evaluating models on grounded reasoning, clarification strategies, and long-form generation in a real-world, knowledge-intensive domain. Grounded in over 35,000 real user-expert interactions, and curated through a carefully designed multi-step pipeline, MIRAGE spans diverse crop health, pest diagnosis, and crop management scenarios. The benchmark includes more than 7,000 unique biological entities, covering plant species, pests, and diseases, making it one of the most taxonomically diverse benchmarks available for vision-language models in real-world expert-guided domains. Unlike existing benchmarks that rely on well-specified user inputs, MIRAGE features underspecified, context-rich scenarios, requiring models to infer latent knowledge gaps and either proactively guide the interaction or respond. Our benchmark comprises two core components. The Single-turn Challenge to reason over a single user turn and image set, identify relevant entities, infer causal explanations, and generate actionable recommendations; and a Multi-Turn challenge for dialogue state tracking, goal-driven generation, and expert-level conversational decision-making. We evaluate more than 20 closed and open-source frontier vision-language models (VLMs), using three reasoning language models as evaluators, highlighting the significant challenges posed by MIRAGE in both single-turn and multi-turn interaction settings. Even the advanced GPT4.1 and GPT4o models achieve 44.6% and 40.9% accuracy, respectively, indicating significant room for improvement.

Integrating large language models (LLMs) into recommender systems has created new opportunities for improving recommendation quality. However, a comprehensive benchmark is needed to thoroughly evaluate and compare the recommendation capabilities of LLMs with traditional recommender systems. In this paper, we introduce \recbench{}, which systematically investigates various item representation forms (including unique identifier, text, semantic embedding, and semantic identifier) and evaluates two primary recommendation tasks, i.e., click-through rate prediction (CTR) and sequential recommendation (SeqRec). Our extensive experiments cover up to 17 large models and are conducted across five diverse datasets from fashion, news, video, books, and music domains. Our findings indicate that LLM-based recommenders outperform conventional recommenders, achieving up to a 5% AUC improvement in CTR and up to a 170% NDCG@10 improvement in SeqRec. However, these substantial performance gains come at the expense of significantly reduced inference efficiency, rendering LLMs impractical as real-time recommenders. We have released our code and data to enable other researchers to reproduce and build upon our experimental results.

Vehicle aerodynamics optimization has become critical for automotive electrification, where drag reduction directly determines electric vehicle range and energy efficiency. Traditional approaches face an intractable trade-off: computationally expensive Computational Fluid Dynamics (CFD) simulations requiring weeks per design iteration, or simplified models that sacrifice production-grade accuracy. While machine learning offers transformative potential, existing datasets exhibit fundamental limitations -- inadequate mesh resolution, missing vehicle components, and validation errors exceeding 5% -- preventing deployment in industrial workflows. We present DrivAerStar, comprising 12,000 industrial-grade automotive CFD simulations generated using STAR-CCM+${}^{\textregistered}$ software. The dataset systematically explores three vehicle configurations through 20 Computer Aided Design (CAD) parameters via Free Form Deformation (FFD) algorithms, including complete engine compartments and cooling systems with realistic internal airflow. DrivAerStar achieves wind tunnel validation accuracy below 1.04% -- a five-fold improvement over existing datasets -- through refined mesh strategies with strict wall $y^+$ control. Benchmarks demonstrate that models trained on this data achieve production-ready accuracy while reducing computational costs from weeks to minutes. This represents the first dataset bridging academic machine learning research and industrial CFD practice, establishing a new standard for data-driven aerodynamic optimization in automotive development. Beyond automotive applications, DrivAerStar demonstrates a paradigm for integrating high-fidelity physics simulations with Artificial Intelligence (AI) across engineering disciplines where computational constraints currently limit innovation.

In recent years, large language models with Transformer architecture as the core have made breakthrough progress in many fields. At the same time, there are also some weaknesses in the large language model that have prompted people to reflect, among which the most fundamental one is the reflection on the Transformer architecture. The Transformer architecture has high parallelism and can fully utilize the computing power of GPUs, thus replacing models such as LSTM in the past few years. However, high parallelism is not a free lunch, as it fundamentally limits the performance of models. Especially, the problems that logarithmic precision Transformer architecture can solve are strictly limited to the $TC^0$. And there are many important issues that are usually considered out of $TC^0$, such as Python code evaluation, entity tracking, chess, and other state tracking tasks. Meanwhile, some recent state space methods based on Delta Rule have been able to break through the $TC^0$ architecture, but they are limited by fixed size state spaces and perform poorly on many tasks. To this end, we have re-examined the Transformer from the perspective of a state space with kernel functions and propose an improved Transformer called DeltaFormer. We have theoretically and practically demonstrated that the proposed new architecture can break through the limitation of the inherent $TC^0$ expressivity of Transformers and verified that it is not weaker than standard Transformer in language modeling tasks. We hope our work can provide inspiration for designing more expressive models.

Securing software supply chains is a growing challenge due to the inadequacy of existing datasets in capturing the complexity of next-gen attacks, such as multiphase malware execution, remote access activation, and dynamic payload generation. Existing datasets, which rely on metadata inspection and static code analysis, are inadequate for detecting such attacks. This creates a critical gap because these datasets do not capture what happens during and after a package is installed. To address this gap, we present QUT-DV25, a dynamic analysis dataset specifically designed to support and advance research on detecting and mitigating supply chain attacks within the Python Package Index (PyPI) ecosystem. This dataset captures install and post-install-time traces from 14,271 Python packages, of which 7,127 are malicious. The packages are executed in an isolated sandbox environment using an extended Berkeley Packet Filter (eBPF) kernel and user-level probes. It captures 36 real-time features, that includes system calls, network traffic, resource usages, directory access patterns, dependency logs, and installation behaviors, enabling the study of next-gen attack vectors. ML analysis using the QUT-DV25 dataset identified four malicious PyPI packages previously labeled as benign, each with thousands of downloads. These packages deployed covert remote access and multi-phase payloads, were reported to PyPI maintainers, and subsequently removed. This highlights the practical value of QUT-DV25, as it outperforms reactive, metadata, and static datasets, offering a robust foundation for developing and benchmarking advanced threat detection within the evolving software supply chain ecosystem.