NeurIPS 2026 Papers

Many emerging agentic paradigms require agents to collaborate with one another (or people) to achieve shared goals. However, existing approaches to learning policies for such collaborative problems produce brittle solutions that fail when paired with new partners. We attribute these failures to a combination of free-riding during training and a lack of strategic robustness. To address these problems, we study the concept of strategic risk aversion and interpret it as a principled inductive bias for generalizable cooperation with unseen partners. While strategically risk-averse players are robust to deviations in their partner's behavior by design, we show that, in collaborative games, they also (1) can have better equilibrium outcomes than those at classical game-theoretic concepts like Nash, and (2) exhibit less or no free-riding. Inspired by these insights, we develop a multi-agent reinforcement learning (MARL) algorithm that integrates strategic risk aversion into standard policy optimization methods. Our empirical results across collaborative benchmarks (including an LLM collaboration task) validate our theory and demonstrate that our approach consistently achieves reliable collaboration with heterogeneous and previously unseen partners across collaborative tasks.

Inference-time scaling provides a promising path for improving agentic reasoning performance, and one widely adopted form is sample-then-select: first sample a pool of candidate reasoning trajectories, then select from the resulting candidates. While this keeps scaling lightweight without updating the policy, it exposes a training-deployment mismatch. Standard trajectory-wise Reinforcement Learning (RL) objectives optimize normalized signals for individual trajectories, whereas choosing from a large candidate pool with the policy's own selection signal is inherently set-conditional: it requires calibrated ranking and confidence aggregation across correlated trajectories. To address this challenge, we propose AutoPortfolio, a set-conditional policy alignment framework for efficient and effective sample-then-select scaling. AutoPortfolio builds a tractable environment-informed target on the sampled candidates, and uses its induced Plackett-Luce (PL) ranking to calibrate the policy which trajectories to prioritize within that set. It trains the policy with a balanced PL-motivated objective that reallocates probability mass among competing trajectories, preserving reward-aligned alternatives while sharpening away from low-quality distractors. Aligned with our policy training, our lightweight inference-time Mass Aggregation scaling adaptively perceives confidence from candidate trajectories and selects the most promising answer(s). Theoretically, we show that reducing our set-conditional loss narrows the discrepancy between policy-induced and environment-informed PL rankings on the sampled set. Empirically, across 12 challenging reasoning benchmarks, AutoPortfolio improves over strong agentic RL baselines, with the clearest gains in larger candidate pools where the set-level selection is crucial.

Large language model (LLM) agents are increasingly deployed in scientific research, where reliability is critical, and the underlying knowledge is densely interconnected. In such settings, hallucinations are particularly damaging: a single erroneous claim on a foundational concept can propagate through multi-step reasoning and corrupt entire trajectories. Existing hallucination benchmarks largely operate at the surface level, treating facts in isolation and relying on uniform accuracy metrics that ignore this topological structure. We address this gap with \textsc{Schema}, the first evidence-grounded, topology-aware evaluation framework for scientific agents. Built on an automated concept-graph construction pipeline, \textsc{Schema} provides two complementary diagnostic instruments. A trajectory hallucination pipeline audits intermediate reasoning at scale via a topology-weighted severity score ($\mathrm{HS}^w$), while a multi-agent counterfactual attribution module pinpoints the causal mechanism behind selected failures. The framework supports diverse task formats spanning multi-hop reasoning, claim verification, and programmatic experimental execution. Across two biomedical subdomains and eleven contemporary LLM agents, \textsc{Schema} reveals that hallucinations concentrate at a small set of highly connected knowledge hubs, and that final-answer accuracy decouples from trajectory honesty—models often reach correct conclusions through structurally flawed reasoning. These results indicate that for high-stakes scientific applications, terminal accuracy alone is an insufficient signal of agent reliability, motivating mechanism-level evaluation grounded in knowledge topology\footnote{Code is available at \url{https://anonymous.4open.science/r/SCHEMA_NIPS2026}}.

While LLMs excel at isolated coding tasks, their ability to perform autonomous, long-horizon machine learning (ML) development remains bottlenecked by poor strategic planning. Because current benchmarks rely on outcome-based metrics, they obscure process-level failures, making it impossible to diagnose where agent strategies diverge from human expertise. To solve this, we introduce TraceML, the first large-scale dataset of paired human and agent trajectories for long-horizon ML tasks. TraceML contains 4{,}672 Kaggle trajectories across 134 MLE-bench competitions, with 430 expert humans paired head-to-head with 207 LLM-agent runs on 7 of those competitions. Using a unified extraction schema to analyze step-by-step behavior, we reveal systematic gaps: agents execute local edits well but exhibit unstructured exploration, poor experiment prioritization, and weak long-term memory attribution compared to experts. Translating these diagnostic insights into action, we design a novel, planning-oriented agent harness based on human priors that significantly improves both process efficiency and final competitive performance.

Equipping large language models (LLMs) with external memory banks is a critical stepping stone towards autonomous agents. Current approaches have concentrated substantial efforts on the design of memory construction and retrieval mechanisms. However, they typically relegate memory utilization to the simplistic concatenation of explicit texts for LLM inputs, thereby remaining constrained by the fundamental bottlenecks of context overload and semantic dilution. In response to this dilemma, we present UniMem, a unified memory adaptation framework that empowers LLM agents to holistically harness memory information, extending beyond textual signals into latent and parametric dimensions. Operating along these pathways, UniMem not only elevates contextual density through latent compression, but also facilitates semantic grounding via parametric modulation. Extensive experiments show that UniMem (I) achieves superior performance across diverse benchmarks, including long-term dialogue, memory personalization, and open-domain question answering; and (II) can seamlessly complement prevailing memory construction and retrieval mechanisms for synergistic enhancements.

Adaptive prompt and program search makes LLM evaluation selection-sensitive. Once benchmark items are reused inside tuning, the observed winner's score need not estimate the fresh-data performance of the full tune-then-deploy procedure. We study inference for this procedure-level target under explicit tuning budgets. We propose SIREN, a selection-aware repeated-split reporting protocol that freezes the post-search shortlist, separates splitwise selection from held-out evaluation, and uses an item-level Gaussian multiplier bootstrap for uncertainty quantification. In a fixed-shortlist regime with smooth stabilized selection, the estimator admits a first-order item-level representation, and the bootstrap yields valid simultaneous inference on a finite budget grid. This supports confidence intervals for procedure-performance curves and pre-specified equal-budget and cross-budget comparisons. Controlled simulations and MMLU-Pro tuning experiments show that winner-based reporting can be optimistic and can change deployment conclusions, while SIREN remains close to the finite-sample reporting target.

Instruction following is a core capability of modern foundation models, underpinning assistant-style interaction, agentic workflows, visual understanding, and multimodal reasoning. Existing instruction-following data and methods, however, remain substantially misaligned with realistic user behavior: users often converse with models over multiple turns, provide interleaved images and videos, switch modalities, or implicitly expect earlier constraints to persist. To bridge this gap, we first identify and formalize six core challenges for realistic conversational multimodal instruction following: Constraint-Guided Visual Reasoning, Cross-Modal Context Transfer, Persistent Instruction Tracking, User-Aware Context Modeling, Iterative Grounded Response Refinement, and Visual Consistency Alignment. Driven by these challenges, we develop a multi-agent data construction pipeline that generates diverse, constraint-rich, image- and video-grounded conversations, followed by human verification. We further improve model behavior through a rubric-guided multi-stage training pipeline that decomposes complex instructions into atomic supervision signals and trains models from basic visual constraint following to advanced multi-turn conversational constraint satisfaction. Finally, we establish Multimodal MultiChallenge (MMMC) as a systematic benchmark for realistic conversational multimodal instruction following. The results show that realistic conversational multimodal instruction following remains challenging for frontier models, while targeted rubric-guided post-training improves constraint satisfaction without sacrificing general multimodal ability.

Semantic fidelity is essential for multi-agent communication under non-ideal channels, comprising transmission fidelity and conversion fidelity. Yet existing numerical and neural message formats cannot simultaneously achieve both. To address this, we propose a novel message format-\textbf{H}yperdimensional \textbf{S}ymbolic \textbf{M}essages (HSM), and its corresponding communication framework \textbf{Herm}. For transmission fidelity, Herm suppresses semantic entanglement via an entity-attribute-value structured encoding pipeline. For conversion fidelity, Herm first leverages a receiver-centric ego-binding mechanism to align semantics. Then, a similarity-aware dual-branch fusion module is proposed to prevent critical semantic dilution. To realize explicit message-semantic conversion, we devise an iterative decoding scheme via algebraic inverse operation. Experiments exhibit that Herm outperforms baselines across various noisy scenarios while maintaining superior structural readability.

Modern autonomous driving systems widely adopt multi-view Bird’s-Eye-View (BEV) based 3D perception models due to their superior performance. Despite their success, the robustness of these models against adversarial camouflage attacks remains largely unexplored. Existing camouflage attacks primarily target single-view 2D detectors, limiting their effectiveness in multi-view 3D perception settings. To address this gap, we propose BEVCA, a novel adversarial camouflage generation framework tailored for multi-view BEV-based 3D perception. Our framework integrates a BEV-feature-based adversarial loss with a multi-view neural rendering module, enabling effective and transferable camouflage attacks across different models and tasks. Extensive experiments in both digital and physical settings demonstrate that BEVCA outperforms state-of-the-art baselines and exhibits strong robustness under real-world conditions.

Large language models increasingly serve as execution engines for agentic systems, yet they still consume context through a sequential text interface. This creates a mismatch with modern structured agent workflows, in which independent branches explore subtasks, retrieve evidence, or generate candidate solutions before a final synthesis step. Existing systems typically merge these branches by concatenating their textual outputs, which discards the parallel structure and incurs redundant prefill computation. In this work, we introduce Parallel-Synthesis, a plug-and-play framework that enables a synthesizer to directly consume the KV caches produced by parallel worker agents. Parallel-Synthesis combines a cache mapper that calibrates independently generated branch caches with a fine-tuned synthesizer adapter that enables generation from this non-sequential cache interface. We train Parallel-Synthesis using data that exposes the synthesizer to parallel cache contexts, teaches aggregation across cached branches, and distills reasoning behavior from standard text-concatenation-based synthesis. Across nine downstream datasets spanning math, science QA, code generation, GAIA, and multi-agent database diagnosis, Parallel-Synthesis matches or outperforms text-based synthesis on seven datasets and remains close on the other two. It also reduces time-to-first-token by 2.5×--11×, suggesting that direct cache-based synthesis is a promising interface for more native and efficient synthesis over parallel agent branches.

Online robust zero-sum Markov games pose a coupled operator-learning and game-solving problem: from nominal interaction data, the learner must recover a worst-case minimax Bellman operator while maintaining strategic coverage and controlling stage-game approximation error. Existing robust reinforcement-learning theory does not yet provide such an operator-learning perspective beyond specialized multi-agent settings. We develop an optimistic dual robust Bellman framework that combines dual robust Bellman fitting, optimistic minimax planning, backward-validated confidence sets, and exploiter-mix data collection. Our analysis is organized around the robust Bellman residuals induced by the candidate value class and yields a regret reduction that separates exploration complexity, robust operator-estimation error, stage-game solution error, and misspecification. Under explicit interpretable assumptions, we obtain conditional theorem interfaces for general and projected function approximation, together with closed results in several structured regimes, including tabular, epochwise/cross-fit linear, self-normalized pure-online linear / finite-rank kernel, and projected RKHS truncation. We also provide numerical experiments that validate the theoretical predictions.

Empowered by advances in Language Model agents, systems have made substantial strides in code generation and understanding. However, these approaches often rely on access to the relevant code, an assumption which does not hold when dealing with external APIs. In this work, we introduce the PAU (Python API Understanding) benchmark, where we provide models with black-box, API-level access to code snippets. Models must query the API with exploratory inputs and draw insights from the resulting outputs, with the goal of describing the snippet's true functionality. By treating the code snippets as external tools that must be understood via interaction alone, PAU studies the more general problem of unsupervised tool understanding, specifically for tools implemented as Python methods. Despite recent progress in coding agents, even frontier models struggle to achieve high performance on PAU, with the best model (Claude-4-Opus) failing to understand over 45% of the PAU test set. An investigation into the common error modes reveals that models are overconfident; they often overrate the quality of their current hypothesis, leading to insufficient exploration and premature termination. Finally, we take inspiration from the Asymmetric Actor Critic (AAC) paradigm, frequently used in robot learning, to post-train models for interactive code understanding. Models trained with AAC conduct more active exploration of the APIs, with an AAC-tuned Qwen3-8B model matching the performance of GPT-5-mini.

Printed circuit boards (PCBs) are the hardware foundation of modern electronic systems. However, automating PCB schematic design remains difficult. A valid schematic must respect real-IC pin-level constraints and cross-component topological relations, neither of which is captured by generic simulation, local rule checks, or direct LLM generation. To address this challenge, we present TopoScope, a training-free LLM agent that uses the schematic graph not only as the final output, but also to guide generation and verification at runtime. TopoScope couples two mechanisms: Topology-Guided Context Scoping, which computes a scoped view for each component-level wiring decision, and Pattern-Based Rule Verification, which expresses datasheet-derived electrical constraints as graph patterns and uses them as in-loop feedback. We also introduce a 20-design benchmark spanning three complexity tiers and six application domains, with functional assertions evaluated on the generated schematic graph and calibrated against expert judgments. Across seven recent LLMs, TopoScope improves functional correctness, reduces invalid graph references, and enables more reliable generation of medium and hard multi-IC schematics compared with knowledge-provided baselines.

Large Language Models (LLMs) have demonstrated remarkable capabilities in orchestrating tools for reasoning tasks. However, existing methods rely on a step-wise paradigm that lacks a global perspective, which causes error accumulation over long horizons and restricts generalization to unseen tools. To overcome these limitations, we propose Tools as Continuous Flow for Evolving Agentic Reasoning (FlowAgent), which reconceptualizes tool chaining as continuous trajectory generation within a semantic space. To systematically evaluate this paradigm, we introduce the first plan-level closed-loop benchmark dedicated to plan-level agentic reasoning in dynamic real-world environments. Specifically, the proposed FlowAgent leverages conditional flow matching to generate continuous latent trajectories, providing a global planning perspective to ensure coherent and robust tool execution. Theoretically, we establish formal bounds on utility convergence and prove that our continuous formulation fundamentally guarantees robust generalization and error attenuation. Empirical evaluations show that FlowAgent achieves superior robustness and adaptability in long-horizon reasoning tasks.

Large language models (LLMs) excel at natural language processing but struggle to interact with external environments. Tool learning provides a promising way to extend LLMs into actionable agents, where tool selection is a critical prerequisite for successful tool use. Existing work often assumes a small or predefined set of tools, leaving large-scale tool selection underexplored. Real-world repositories contain a vast and diverse array of tools, making it difficult for LLMs to effectively search, distinguish, and compose tools under context-length constraints. We identify large-scale tool selection as a new challenge for agentic reinforcement learning, highlighting that existing RL methods for knowledge-based question answering are inadequate for selecting tools while considering compatibility. To address this challenge, we propose ToolSearcher, a novel RL framework for effective multi-turn search and fine-grained optimization in large-scale tool selection. Specifically, we introduce category-constrained tool discrimination to improve the model's ability to distinguish functionally similar tools, event-level search modeling to explicitly optimize the discovery of target tools during multi-turn search, and trajectory-aligned credit allocation to provide fine-grained reward signals for different stages of the search-selection process. Extensive experiments on large-scale tool selection benchmarks demonstrate that ToolSearcher consistently outperforms a set of strong baselines in challenging settings involving iterative search and complex tool composition.

Computer Use Agents (CUAs) can act through both atomic GUI actions ($\textit{e.g., click, type}$) and high-level tool calls ($\textit{e.g., API-based file operations}$), but they are often confused by this hybrid action space: they do not know when to continue with GUI actions and when to switch to tools, and finally fail to select the optimal execution path. We view this orchestration problem as GUI-Tool path selection: deciding when the agent should continue with GUI actions and when it should switch to tool calls to form an effective execution trajectory. This difficulty stems from two issues. First, high-quality interleaved GUI-Tool trajectories are scarce, and collecting real tool trajectories is expensive and brittle. Second, existing supervision provides limited guidance for GUI-Tool path selection, as most methods focus on step-level action imitation or final task completion and offer little trajectory-level feedback on whether GUI-Tool switching leads to a more effective execution path. In this paper, we propose $\textbf{ToolCUA}$, an end-to-end agent designed to learn optimal GUI-Tool path selection through a staged training paradigm. We first introduce an $\textbf{Interleaved GUI-Tool Trajectory Scaling Pipeline}$ that repurposes abundant static GUI trajectories and synthesizes a grounded library of tools, making it possible to scale diverse GUI-Tool trajectories without manual engineering or real tool-trajectory collection.Based on this data, we perform Tool-Bootstrapped GUI RFT, which combines warmup SFT with single-turn RL to improve decisions at critical GUI-Tool switching points. Finally, we further optimize ToolCUA with $\textbf{Online Agentic RL}$ in a high-fidelity GUI-Tool environment, using a Tool-Efficient Path Reward that encourages both appropriate tool use and shorter execution paths. Experiments on OSWorld-MCP show that ToolCUA achieves 46.85% accuracy and outperforms the baseline by over 60% relatively, establishing a new state of the art among models of comparable scale. It also improves by 3.9% over GUI-only settings, demonstrating effective GUI-Tool orchestration. The results further suggest that training in a hybrid action space is a promising paradigm for real-world digital agents.

Large language model (LLM) routing is widely used to advance the cost-quality Pareto frontier of modern serving systems. While coarse-grained routing at the session or query level has been widely adopted in production systems, recent algorithm work shows substantial efficiency and quality advantages of finer token-level routing. However, efficiently serving token-level routed inference poses significant challenges to existing system design. Built on single-LLM assumptions, current systems struggle with step desynchronization and frequent batch admission delays under token-level routing, and they also impose high implementation complexity on developers. To address this, we propose TokenRouter, an efficient and user-friendly serving system for token-level routed LLM inference. TokenRouter shifts from the conventional request-centric paradigm to a model-centric one. Instead of treating each request as a synchronized decoding stream, TokenRouter launches a subserver for each candidate LLM and dispatches requests asynchronously according to routing decisions. Each subserver employs a delayed-batching scheduler, whose hyperparameters are derived mathematically from a throughput model of the system. Across diverse routing algorithms, workloads, and model pairs, TokenRouter achieves 1.83-64.29× higher decoding throughput than existing systems, substantially advancing the serving efficiency of token-level LLM routing.

Large Language Model-based Multi-Agent Systems (LLM-MAS) have demonstrated strong capabilities in solving complex tasks but remain vulnerable when agents receive unreliable messages. This vulnerability stems from a fundamental gap: LLM agents treat all incoming messages equally without evaluating their trustworthiness. While some existing studies approach trustworthiness, they focus on a single type of harmfulness rather than analyze it in a holistic approach from multiple trustworthiness perspectives. We address this gap by proposing a comprehensive definition of trustworthiness inspired by human communication theory \citep{grice1975logic}. Our definition identifies six orthogonal trust dimensions that provide interpretable measures of trustworthiness. Building on this definition, we introduce the Attention Trust Score (A‑Trust), a lightweight, attention‑based method for evaluating the trustworthiness of messages. We then develop a principled trust management system (TMS) for LLM‑MAS that supports both message‑level and agent‑level trust assessments. Experiments across diverse multi‑agent settings and tasks demonstrate that our TMS significantly improves robustness against malicious inputs.

As web agents close the gap with humans on benchmarks, it raises the question: *Do today's agents perform just as well on tomorrow's web?* We introduce TimeWarp, a benchmark that emulates the evolving web. TimeWarp consists of three web environments, each with six UI versions spanning UI design and frontend code from different eras of the internet. We pair TimeWarp with a set of complex, realistic tasks covering different forms of web navigation. Our experiments reveal vulnerabilities of web agents, especially visual ones, to changes and the limitations of behavior cloning (BC) on complex trajectories from a single version. To address this, we propose TimeTraj, a simple yet effective algorithm that uses plan distillation to collect trajectories across multiple versions. By training agents on teacher rollouts using our BC-variant, we achieve substantial performance gains: $20.4$ %$\rightarrow37.7$% for Qwen-3 4B and $0$ % $\rightarrow27.0$% for Llama-3.1 8B models. Our work helps study generalization across web designs and enables a new paradigm for collecting plans rather than trajectories to improve the robustness of web agents.