NeurIPS 2025 Papers

Computer-Aided Design (CAD) is a time-consuming and complex process, requiring precise, long-horizon user interactions with intricate 3D interfaces. While recent advances in AI-driven user interface (UI) agents show promise, most existing datasets and methods focus on short, low-complexity tasks in mobile or web applications, failing to capture the demands of professional engineering tools. In this work, we introduce VideoCAD, the first attempt to model UI interactions for precision engineering tasks. Specifically, VideoCAD is a large-scale synthetic dataset consisting of over 41K annotated video recordings of CAD operations, generated using an automated framework for collecting high-fidelity UI action data from human-made CAD designs. Compared to existing datasets, VideoCAD offers an order-of-magnitude increase in complexity for real-world engineering UI tasks, with time horizons up to $20\times$ longer than those in other datasets. We show two important downstream applications of VideoCAD: (1) learning UI interactions from professional 3D CAD tools for precision tasks and (2) a visual question-answering (VQA) benchmark designed to evaluate multimodal large language models (LLMs) on spatial reasoning and video understanding. To learn the UI interactions, we propose VideoCADFormer, a state-of-the-art model for learning CAD interactions directly from video, which outperforms existing behavior cloning baselines. Both VideoCADFormer and the VQA benchmark derived from VideoCAD reveal key challenges in the current state of video-based UI understanding, including the need for precise action grounding, multi-modal and spatial reasoning, and long-horizon dependencies. Dataset and code available at: https://github.com/ghadinehme/VideoCAD.

Zero-shot object navigation (ZSON) allows robots to find target objects in unfamiliar environments using natural language instructions, without relying on pre-built maps or task-specific training. Recent general-purpose models, such as large language models (LLMs) and vision-language models (VLMs), equip agents with semantic reasoning abilities to estimate target object locations in a zero-shot manner. However, these models often greedily select the next goal without maintaining a global understanding of the environment and are fundamentally limited in the spatial reasoning necessary for effective navigation. To overcome these limitations, we propose a novel 3D voxel-based belief map that estimates the target’s prior presence distribution within a voxelized 3D space. This approach enables agents to integrate semantic priors from LLMs and visual embeddings with hierarchical spatial structure, alongside real-time observations, to build a comprehensive 3D global posterior belief of the target’s location. Building on this 3D voxel map, we introduce BeliefMapNav, an efficient navigation system with two key advantages: i) grounding LLM semantic reasoning within the 3D hierarchical semantics voxel space for precise target position estimation, and ii) integrating sequential path planning to enable efficient global navigation decisions. Experiments on HM3D and HSSD benchmarks show that BeliefMapNav achieves state-of-the-art (SOTA) Success Rate (SR) and Success weighted by Path Length (SPL), with a notable 9.7 SPL improvement over the previous best SR method, validating its effectiveness and efficiency.

Recent advances in Large Language Models (LLMs) have spurred transformative applications in various domains, ranging from open-source to proprietary LLMs. However, jailbreak attacks, which aim to break safety alignment and user compliance by tricking the target LLMs into answering harmful and risky responses, are becoming an urgent concern. The practice of red-teaming for LLMs is to proactively explore potential risks and error-prone instances before the release of frontier AI technology. This paper proposes an agentic workflow to automate and scale the red-teaming process of LLMs through the Composition-of-Principles (CoP) framework, where human users provide a set of red-teaming principles as instructions to an AI agent to automatically orchestrate effective red-teaming strategies and generate jailbreak prompts. Distinct from existing red-teaming methods, our CoP framework provides a unified and extensible framework to encompass and orchestrate human-provided red-teaming principles to enable the automated discovery of new red-teaming strategies. When tested against leading LLMs, CoP reveals unprecedented safety risks by finding novel jailbreak prompts and improving the best-known single-turn attack success rate by up to 13.8 times.

Multi-Agent Reinforcement Learning (MARL) has achieved remarkable success in various real-world scenarios, but its high cost of online training makes it impractical to learn each task from scratch. To enable effective policy reuse, we consider the problem of zero-shot generalization from offline data across multiple tasks. While prior work focuses on transferring individual skills of agents, we argue that the effective policy transfer across tasks should also capture the team-level coordination knowledge. In this paper, we propose Bi-Level Knowledge Transfer (BiKT) for Multi-Task MARL, which performs knowledge transfer at both the individual and team levels. At the individual level, we extract transferable individual skill embeddings from offline MARL trajectories. At the team level, we define tactics as coordinated patterns of skill combinations and capture them by leveraging the learned skill embeddings. We map skill combinations into compact tactic embeddings and then construct a tactic codebook. To incorporate both skills and tactics into decision-making, we design a bi-level decision transformer that infers them in sequence. Our BiKT leverages both the generalizability of individual skills and the diversity of tactics, enabling the learned policy to perform effectively across multiple tasks. Extensive experiments on SMAC and MPE benchmarks demonstrate that BiKT achieves strong generalization to previously unseen tasks.

The emergence of large language models (LLMs) enables the development of intelligent agents capable of engaging in complex and multi-turn dialogues. However, multi-agent collaboration faces critical safety challenges, such as hallucination amplification and error injection and propagation. This paper presents GUARDIAN, a unified method for detecting and mitigating multiple safety concerns in GUARDing Intelligent Agent collaboratioNs. By modeling the multi-agent collaboration process as a discrete-time temporal attributed graph, GUARDIAN explicitly captures the propagation dynamics of hallucinations and errors. The unsupervised encoder-decoder architecture incorporating an incremental training paradigm learns to reconstruct node attributes and graph structures from latent embeddings, enabling the identification of anomalous nodes and edges with unparalleled precision. Moreover, we introduce a graph abstraction mechanism based on the Information Bottleneck Theory, which compresses temporal interaction graphs while preserving essential patterns. Extensive experiments demonstrate GUARDIAN's effectiveness in safeguarding LLM multi-agent collaborations against diverse safety vulnerabilities, achieving state-of-the-art accuracy with efficient resource utilization. The code is available at https://github.com/JialongZhou666/GUARDIAN.

Vision-language models have demonstrated impressive capabilities as computer-use agents (CUAs) capable of automating diverse computer tasks. As their commercial potential grows, critical details of the most capable CUA systems remain closed. As these agents will increasingly mediate digital interactions and execute consequential decisions on our behalf, the research community needs access to open CUA frameworks to study their capabilities, limitations, and risks. To bridge this gap, we propose OpenCUA, a comprehensive open-source framework for scaling CUA data and foundation models. Our framework consists of: (1) an annotation infrastructure that seamlessly captures human computer-use demonstrations; (2) AgentNet, the first large-scale computer-use task dataset spanning 3 operating systems and 200+ applications and websites; (3) a scalable pipeline that transforms demonstrations into state–action pairs with reflective long Chain-of-Thought reasoning that sustain robust performance gains as data scales. Our end-to-end agent models demonstrate strong performance across CUA benchmarks. In particular, OpenCUA-72B achieves an average success rate of 45.0% on OSWorld‑Verified, establishing a new state-of-the-art (SOTA) among open-source models. Further analysis confirms that our approach generalizes well across domains and benefits significantly from increased test-time computation. We release our annotation tool, datasets, code, and models to build open foundations for further CUA research.

The recent advent of reasoning models like OpenAI's o1 was met with excited speculation by the AI community about the mechanisms underlying these capabilities in closed models, followed by a rush of replication efforts, particularly from the open source community. These speculations were largely settled by the demonstration from DeepSeek-R1 that chain-of-thought and reinforcement learning (RL) can effectively replicate reasoning on top of base LLMs. However, it remains valuable to explore alternative methods for theoretically eliciting reasoning that could help elucidate the underlying mechanisms, as well as providing additional methods that may offer complementary benefits. Here, we build on the long-standing literature in cognitive psychology and cognitive architectures, which postulates that reasoning arises from the orchestrated, sequential execution of a set of modular, predetermined cognitive operations. Crucially, we implement this key idea within a modern agentic tool-calling framework. In particular, we endow an LLM with a small set of "cognitive tools" encapsulating specific reasoning operations, each executed by the LLM itself. Surprisingly, this simple strategy results in considerable gains in performance on standard mathematical reasoning benchmarks compared to base LLMs, for both closed and open-weight models. For instance, providing our "cognitive tools" to GPT-4.1 increases its pass@1 performance on AIME2024 from 32\% to 53\%, even surpassing the performance of o1-preview. In addition to its practical implications, this demonstration contributes to the debate regarding the role of post-training methods in eliciting reasoning in LLMs versus the role of inherent capabilities acquired during pre-training, and whether post-training merely uncovers these latent abilities.

Large language model (LLM) agents for web interfaces have advanced rapidly, yet open-source systems still lag behind proprietary agents. Bridging this gap is key to enabling customizable, efficient, and privacy-preserving agents. Two challenges hinder progress: the reproducibility issues in RL and LLM agent training, where results often depend on sensitive factors like seeds and decoding parameters, and the focus of prior work on single-step tasks, overlooking the complexities of web-based, multi-step decision-making. We address these gaps by providing a statistically driven study of training LLM agents for web tasks. Our two-stage pipeline combines imitation learning from a Llama 3.3 70B teacher with on-policy fine-tuning via Group Relative Policy Optimization (GRPO) on a Llama 3.1 8B student. Through 240 configuration sweeps and rigorous bootstrapping, we chart the first compute allocation curve for open-source LLM web agents. Our findings show that dedicating one-third of compute to teacher traces and the rest to RL improves MiniWoB++ success by 6 points and closes 60\% of the gap to GPT-4o on WorkArena, while cutting GPU costs by 45\%. We introduce a principled hyperparameter sensitivity analysis, offering actionable guidelines for robust and cost-effective agent training.

Large Language Models (LLMs) are rapidly saturating existing benchmarks, necessitating new open-ended evaluations. We introduce the Factorio Learning Environment (FLE), based on the game of Factorio, that tests agents in long-term planning, spatial reasoning, program synthesis, and resource optimization. FLE provides exponentially scaling challenges -- from basic automation to complex factories processing millions of resource units per second. We provide two settings: (1) open-play with the open-ended task of building the largest factory on an procedurally generated map and (2) lab-play consisting of 33 bounded tasks accross three settings with fixed resources. We demonstrate across both settings that models still lack strong spatial reasoning. In lab-play, we find that LLMs exhibit promising short-horizon skills, yet are unable to operate effectively in constrained environments, reflecting limitations in error analysis. In open-play, while LLMs discover automation strategies that improve growth (e.g electric-powered drilling), they fail to achieve complex automation (e.g electronic-circuit manufacturing)

Despite recent progress in Language Models (LMs) for software engineering, collecting training data remains a significant pain point. Existing datasets are small, with at most 1,000s of training instances from 11 or fewer GitHub repositories. The procedures to curate such datasets are often complex, necessitating hundreds of hours of human labor; companion execution environments also take up several terabytes of storage, severely limiting their scalability and usability. To address this pain point, we introduce SWE-smith, a novel pipeline for generating software engineering training data at scale. Given any Python codebase, SWE-smith constructs a corresponding execution environment, then automatically synthesizes 100s to 1,000s of task instances that break existing test(s) in the codebase. Using SWE-smith, we create a dataset of 50k instances sourced from 128 GitHub repositories, an order of magnitude larger than all previous works. We train SWE-agent-LM-32B, achieving 40.2% Pass@1 resolve rate on the SWE-bench Verified benchmark, state of the art among open source models. We open source SWE-smith (collection procedure, task instances, trajectories, models) to lower the barrier of entry for research in LM systems for automated software engineering. All assets available at \url{https://swesmith.com}.

One of the principal challenges in building VLM-powered GUI agents is visual grounding—localizing the appropriate screen region for action execution based on both the visual content and the textual plans. Most existing work formulates this as a text-based coordinate generation task. However, these approaches suffer from several limitations: weak spatial-semantic alignment due to lack of explicit spatial supervision; inability to handle ambiguous supervision targets, as single-point predictions penalize valid variations; and a mismatch between the dense nature of screen coordinates and the coarse, patch-level granularity of visual features extracted by models like Vision Transformers. In this paper, we propose **GUI-Actor**, a VLM-based method for coordinate-free GUI grounding. At its core, **GUI-Actor** introduces an attention-based action head that learns to align a dedicated `<ACTOR>` token with all relevant visual patch tokens, enabling the model to propose one or more action regions in a single forward pass. In line with this, we further design a grounding verifier to evaluate and select the most plausible action region from the candidates proposed for action execution. Extensive experiments show that **GUI-Actor** outperforms prior state-of-the-art methods on multiple GUI action grounding benchmarks, with improved generalization to unseen screen resolutions and layouts. Notably, **GUI-Actor-7B** achieves scores of **40.7** with Qwen2-VL and **44.6** with Qwen2.5-VL as backbones, outperforming **UI-TARS-72B (38.1)** on ScreenSpot-Pro, with significantly fewer parameters and training data. Furthermore, by incorporating the verifier, we find that fine-tuning only the newly introduced action head (~100M parameters for 7B model) while keeping the VLM backbone frozen is sufficient to achieve performance comparable to previous state-of-the-art models, highlighting that **GUI-Actor** can endow the underlying VLM with effective grounding capabilities without compromising its general-purpose strengths. Project page: [https://aka.ms/GUI-Actor](https://aka.ms/GUI-Actor)

We introduce MLE-Dojo, a Gym-style framework for systematically reinforcement learning, evaluating, and improving autonomous large language model (LLM) agents in iterative machine learning engineering (MLE) workflows. Unlike existing benchmarks that primarily rely on static datasets or single-attempt evaluations, MLE-Dojo provides an interactive environment enabling agents to iteratively experiment, debug, and refine solutions through structured feedback loops. Built upon 200+ real-world Kaggle challenges, MLE-Dojo covers diverse, open-ended MLE tasks carefully curated to reflect realistic engineering scenarios such as data processing, architecture search, hyperparameter tuning, and code debugging. Its fully executable environment supports comprehensive agent training via both supervised fine-tuning and reinforcement learning, facilitating iterative experimentation, realistic data sampling, and real-time outcome verification. Extensive evaluations of eight frontier LLMs reveal that while current models achieve meaningful iterative improvements, they still exhibit significant limitations in autonomously generating long-horizon solutions and efficiently resolving complex errors. Furthermore, MLE-Dojo’s flexible and extensible architecture seamlessly integrates diverse data sources, tools, and evaluation protocols, uniquely enabling model-based agent tuning and promoting interoperability, scalability, and reproducibility. We open-source our framework and benchmarks to foster community-driven innovation towards next-generation MLE agents.

In this paper, we consider fully decentralized cooperative multi-agent reinforcement learning, where each agent has access only to the states, its local actions, and the shared rewards. The absence of information about other agents' actions typically leads to the non-stationarity problem during per-agent value function updates, and the relative overgeneralization issue during value function estimation. However, existing works fail to address both issues simultaneously, as they lack the capability to model the agents' joint policy in a fully decentralized setting. To overcome this limitation, we propose a simple yet effective method named Return-Aware Context (RAC). RAC formalizes the dynamically changing task, as locally perceived by each agent, as a contextual Markov Decision Process (MDP), and addresses both non-stationarity and relative overgeneralization through return-aware context modeling. Specifically, the contextual MDP attributes the non-stationary local dynamics of each agent to switches between contexts, each corresponding to a distinct joint policy. Then, based on the assumption that the joint policy changes only between episodes, RAC distinguishes different joint policies by the training episodic return and constructs contexts using discretized episodic return values. Accordingly, RAC learns a context-based value function for each agent to address the non-stationarity issue during value function updates. For value function estimation, an individual optimistic marginal value is constructed to encourage the selection of optimal joint actions, thereby mitigating the relative overgeneralization problem. Experimentally, we evaluate RAC on various cooperative tasks (including matrix game, predator and prey, and SMAC), and its significant performance validates its effectiveness.

Large language model (LLM) based agentic workflows have become a popular paradigm for coordinating multiple specialized agents to solve complex tasks. To improve serving efficiency, existing LLM systems employ prefix caching to reuse key-value (KV) tensors corresponding to agents' fixed prompts, thereby avoiding redundant computation across repeated invocations. However, current systems typically evict KV caches using a Least Recently Used (LRU) policy, which fails to anticipate future agent usage and often discards KV caches shortly before their reuse. This leads to frequent cache misses and substantial recomputation or swap- ping overhead. We present KVFlow, a workflow-aware KV cache management framework tailored for agentic workloads. KVFlow abstracts the agent execution schedule as an Agent Step Graph and assigns each agent a steps-to-execution value that estimates its temporal proximity to future activation. These values guide a fine-grained eviction policy at the KV node level, allowing KVFlow to preserve entries likely to be reused and efficiently manage shared prefixes in tree-structured caches. Moreover, KVFlow introduces a fully overlapped KV prefetching mecha- nism, which proactively loads required tensors from CPU to GPU in background threads for agents scheduled in the next step, thereby avoiding cache miss stalls during generation. Compared to SGLang with hierarchical radix cache, KVFlow achieves up to 1.83× speedup for single workflows with large prompts, and up to 2.19× speedup for scenarios with many concurrent workflows.

Instruction fine-tuning is crucial in NLP tasks, enhancing pretrained models' instruction-following capabilities and task-specific performance. However, obtaining high-quality fine-tuning data for large models is challenging due to data collection difficulties and high production costs. To address this, we propose MASTER, a novel data augmentation method that enriches original data through interactions among multiple agents with varying cognitive levels. We simulate three pedagogically grounded teaching scenarios, leveraging multi-agent conversations to generate high-quality teacher-student interaction data. Utilizing MASTER, we construct BOOST-QA, a fine-tuning dataset augmented from existing datasets like Orca-Math-200k, ProcQA, and OpenHermes2.5. Experiments show that models fine-tuned with BOOST-QA perform excellently across multiple benchmarks, demonstrating strong multitask generalization. Notably, MASTER significantly improves models' reasoning abilities in complex tasks, providing valuable insights for future research.

Adaptive cooperation in multi-agent reinforcement learning (MARL) requires policies to express homogeneous, specialised, or mixed behaviours, yet achieving this adaptivity remains a critical challenge. While parameter sharing (PS) is standard for efficient learning, it notoriously suppresses the behavioural diversity required for specialisation. This failure is largely due to cross-agent gradient interference, a problem we find is surprisingly exacerbated by the common practice of *coupling agent IDs with observations*. Existing remedies typically add complexity through altered objectives, manual preset diversity levels, or sequential updates -- raising a fundamental question: *can shared policies adapt without these intricacies?* We propose a solution built on a key insight: an agent-conditioned hypernetwork can generate agent-specific parameters and *decouple* observation- and agent-conditioned gradients, directly countering the interference from coupling agent IDs with observations. Our resulting method, **HyperMARL**, avoids the complexities of prior work and empirically reduces policy gradient variance. Across diverse MARL benchmarks (22 scenarios, up to 30 agents), HyperMARL achieves performance competitive with six key baselines while preserving behavioural diversity comparable to non-parameter sharing methods, establishing it as a versatile and principled approach for adaptive MARL. The code is publicly available at https://github.com/KaleabTessera/HyperMARL.

Conventional wisdom dictates that small batch sizes make language model pretraining and fine-tuning unstable, motivating gradient accumulation, which trades off the number of optimizer steps for a proportional increase in batch size. While it is common to decrease the learning rate for smaller batch sizes, other hyperparameters are often held fixed. In this work, we revisit small batch sizes all the way down to batch size one, and we propose a rule for scaling Adam hyperparameters to small batch sizes. In particular, rather than holding the decay rate of the second moment fixed across batch sizes, we propose to hold its half-life fixed in terms of tokens. We find that small batch sizes (1) train stably, (2) are consistently more robust to hyperparameter choices, (3) achieve equal or better per-FLOP performance than larger batch sizes, and (4) notably enable stable language model training with vanilla SGD, even without momentum, despite storing no optimizer state. Building on these results, we provide practical recommendations for selecting a batch size and setting optimizer hyperparameters. We further recommend against gradient accumulation unless training on multiple devices with multiple model replicas. Finally, we show that a small batch size combined with an optimizer with a small state size can provide the performance benefits of full fine-tuning while maintaining a similar memory footprint to LoRA.

As multimodal large models (MLLMs) continue to advance across challenging tasks, a key question emerges: \textbf{\textit{What essential capabilities are still missing? }} A critical aspect of human learning is continuous interaction with the environment -- not limited to language, but also involving multimodal understanding and generation. To move closer to human-level intelligence, models must similarly support \textbf{multi-turn}, \textbf{multimodal interaction}. In particular, they should comprehend interleaved multimodal contexts and respond coherently in ongoing exchanges. In this work, we present \textbf{an initial exploration} through the \textsc{InterMT} -- \textbf{the first preference dataset for \textit{multi-turn} multimodal interaction}, grounded in real human feedback. In this exploration, we particularly emphasize the importance of human oversight, introducing expert annotations to guide the process, motivated by the fact that current MLLMs lack such complex interactive capabilities. \textsc{InterMT} captures human preferences at both global and local levels into nine sub-dimensions, consists of 15.6k prompts, 52.6k multi-turn dialogue instances, and 32.4k human-labeled preference pairs. To compensate for the lack of capability for multi-modal understanding and generation, we introduce an agentic workflow that leverages tool-augmented MLLMs to construct multi-turn QA instances. To further this goal, we introduce \textsc{InterMT-Bench} to assess the ability of MLLMs in assisting judges with multi-turn, multimodal tasks. We demonstrate the utility of \textsc{InterMT} through applications such as judge moderation and further reveal the \textit{multi-turn scaling law} of judge model. We hope the open-source of our data can help facilitate further research on aligning current MLLMs to the next step.

A challenge in human-AI decision-making is to balance three factors: the *correctness* of predictions, the *cost* of knowledge and reasoning complexity, and the confidence about whether to *abstain* from automated answers or escalate to human experts. In this work, we present a cascaded LLM decision framework that adaptively delegates tasks across multiple tiers of expertise -- a base model for initial candidate answers, a more capable and knowledgeable (but costlier) large model, and a human expert for when the model cascade abstains. Our method proceeds in two stages. First, a deferral policy determines whether to accept the base model’s answer or regenerate it with the large model based on the confidence score. Second, an abstention policy decides whether the cascade model response is sufficiently certain or requires human intervention. Moreover, to overcome static policies and accommodate changing task difficulty, we incorporate an online learning mechanism which uses human feedback. We demonstrate this approach to general question-answering (ARC-Easy, ARC-Challenge, and MMLU) and medical question-answering (MedQA and MedMCQA). Our results demonstrate that our cascaded strategy outperforms single-model baselines in most cases, achieving higher accuracy while reducing costs and providing a principled approach to handling abstentions.