ICML 2026 论文列表

Reinforcement Learning (RL) has become pivotal for improving model capabilities yet suffers from rollout efficiency bottlenecks due to the long-tail response length distribution. While existing works mitigate the impact of long tails via prompt-level tail scheduling, we focus on the root source of inefficiency: the distribution itself. Specifically, we characterize the long-tail distribution at a finer granularity, identifying intra-prompt long tails, and revealing that they frequently consist of ineffective verbosity. To address this, we propose a novel paradigm of active distribution shaping to shape the rollout distribution towards conciseness and certainty, thereby fundamentally resolving tail-induced overheads. We achieve this through a distribution-aware trajectory sampling mechanism, which selects trajectories from a redundant exploration space for each prompt, and an adaptive redundancy allocation scheme to maximize both shaping effectiveness and system efficiency. Experiments demonstrate significant acceleration over state-of-the-art systems by up to 1.77$\times$ without compromising model performance.

We address the challenging task of 3D object segmentation in complex scene point clouds without relying on any scene-level human annotations during training. Existing methods are typically constrained to identifying simple objects, primarily due to insufficient object priors in the learning process. In this paper, we present FoundObj, a novel framework featuring a superpoint-based object discovery agent that incrementally merges suitable neighboring superpoints, guided by our innovative semantic and geometric reward modules. These modules synergistically leverage semantic and geometric priors from self-supervised 2D/3D foundation models, providing complementary feedback to the object discovery agent and enabling robust identification of multi-class objects through reinforcement learning. Extensive experiments on diverse benchmarks demonstrate that our approach consistently outperforms existing baselines. Notably, our method exhibits strong generalization in zero-shot and long-tail scenarios, underscoring its potential for scalable, label-free 3D object segmentation. Code is available at https://github.com/vLAR-group/FoundObj

Reinforcement learning with verifiable rewards (RLVR) efficiently scales the reasoning ability of large language models (LLMs) but is bottlenecked by scarce labeled data. Reinforcement learning with intrinsic rewards (RLIR) offers a scalable alternative via self-rewarding, yet often suffers from instability and inferior performance. We trace this gap to a systemic bias in confidence-coupled self-rewarding: the model tends to over-reward high-confidence mistakes, forming a \textbf{self-confirming loop}. We quantify this feedback-loop bias with three metrics: reward noise magnitude ($\rho_{\text{noise}}$), policy–reward coupling ($\rho_{\text{selfbias}}$), and over-/under-reward skew ($\rho_{\text{symbias}}$). Our analyses show a compounding effect where strong coupling amplifies confidence-conditioned errors and drives a drift toward over-reward, leading to instability and a lower performance ceiling. To mitigate this, we propose reinforcement learning with ensembled rewards (\textbf{RLER}), which aggregates diverse models with adaptive reward interpolation and disagreement-aware rollout selection to reduce coupling and suppress over-reward drift. Extensive experiments show that RLER improves by 6.2\% over the best RLIR baseline and is within 3.6\% of RLVR, while exhibiting stable scaling on unlabeled samples.

Planning-driven model-based (modelic) reinforcement learning has achieved impressive success in continuous control tasks but predominantly relies on zero-order optimizers like Model Predictive Path Integral (MPPI). While robust for global exploration, MPPI updates actions solely through sampling and neglects the smooth return gradients inherent in structured dynamics that guide fine-grained search. To complement MPPI’s robustness with gradient-guided precision, we first propose \textbf{La}ngevin \textbf{R}ollout \textbf{O}ptimization (LaRO), which leverages return gradients to refine actions via Langevin dynamics, achieving reliable local convergence without sacrificing multimodal exploration. This is supported by a score-augmented world model that jointly learns dynamics and a score function within a unified latent space, facilitating efficient and accurate gradient estimation for real-time planning. Second, we combine MPPI and LaRO through a simple yet effective choice mechanism, termed \textbf{M}aximum \textbf{L}ook-\textbf{A}head \textbf{P}lanning (MLAP). Finally, we instantiate MLAP within the latest BOOM algorithm, replacing its MPPI-only planner and yielding BOOM-L. Empirical results on the DeepMind Control Suite and Humanoid Bench demonstrate that BOOM-L consistently outperforms strong baselines in both sample efficiency and final performance.

Recent advancements in aligning large language models via reinforcement learning have achieved remarkable gains in solving complex reasoning problems, but at the cost of expensive on-policy rollouts and limited exploration of diverse reasoning paths. In this work, we introduce TreePO, involving a self-guided rollout algorithm that views sequence generation as a tree-structured searching process. Composed of dynamic tree sampling policy and fixed-length segment decoding, TreePO leverages local uncertainty to warrant additional branches. By amortizing computation across common prefixes and pruning low-value paths early, TreePO essentially reduces the per-update compute burden while preserving or enhancing exploration diversity. Key contributions include: (1) a segment-wise sampling algorithm that alleviates the KV cache burden through contiguous segments and spawns new branches along with an early-stop mechanism; (2) a tree-based segment-level advantage estimation that considers both global and local proximal policy optimization. and (3) analysis on the effectiveness of probability and quality-driven dynamic divergence and fallback strategy. We empirically validate the performance gain of TreePO on a set reasoning benchmarks and the efficiency saving of GPU hours from 22% up to 43% of the sampling design for the trained models, meanwhile showing up to 40% reduction at trajectory-level and 35% at token-level sampling compute for the existing models. While offering a free lunch of inference efficiency, TreePO reveals a practical path toward scaling RL-based post-training with fewer samples and less compute.

Efficient exploration remains a key challenge in deep reinforcement learning, especially for embodied agents operating in realistic environments with high-dimensional observations and complex dynamics. Recent latent exploration methods define bonuses in a learned latent space, but often struggle in these settings where (i) representations can be noisy or policy-dependent, and (ii) common strategies such as randomized latent objectives or fixed directional spanning are brittle and fail to improve global coverage. We propose Covariance Volume Maximization (CVM), a coverage-driven latent exploration framework with two key components. First, we learn a behavioral state encoder using a *policy-mixture* objective to reduce representation drift under rapidly changing exploration policies, yielding stable and behaviorally meaningful latent displacements. Second, CVM rewards each transition by its exact increase in the log-determinant of the covariance of recent latent displacements, explicitly expanding the explored region and prioritizing under-covered directions. This objective coincides with the classical *D-optimal* design criterion, providing an information-efficiency justification. Extensive experiments on embodied navigation and manipulation tasks demonstrate that CVM substantially improves exploration efficiency and robustness, and scales effectively to different environments.

While recent video diffusion models (VDMs) produce visually impressive results, they fundamentally struggle to maintain 3D structural consistency, often resulting in object deformation or spatial drift. We hypothesize that these failures arise because standard denoising objectives lack explicit incentives for geometric coherence. To address this, we introduce VideoGPA (Video Geometric Preference Alignment), a data-efficient self-supervised framework that leverages a geometry foundation model to automatically derive dense preference signals that guide VDMs via Direct Preference Optimization (DPO). This approach effectively steers the generative distribution toward inherent 3D consistency without requiring human annotations. VideoGPA significantly enhances temporal stability, geometric plausibility, and motion coherence using minimal preference pairs, consistently outperforming state-of-the-art baselines in extensive experiments.

Reinforcement Learning with Verifiable Reward (RLVR) on preference data has become the mainstream approach for training Generative Reward Models (GRMs). Typically, GRMs generate reasoning chains ending with critiques and preference labels, with RLVR using label correctness as the training reward. However, we demonstrate that such binary classification tasks make GRMs susceptible to guessing correct outcomes without sound critiques, introducing noise into the reward signal and impairing learning effectiveness. To address this, we propose Reward Modeling from Natural Language Human Feedback (RM-NLHF), which leverages natural language feedback to obtain process reward signals. Specifically, we compute the similarity between GRM-generated and human critiques as the process reward, providing more accurate signals than outcome-only supervision. Considering that human critiques are difficult to scale, we introduce MetaRM which learns to predict process reward from datasets with human critiques and generalizes to data without them. Experiments on multiple benchmarks demonstrate that RM-NLHF consistently outperforms state-of-the-art models trained with outcome reward, confirming the superiority of natural language over binary feedback. Our code is publicly available at https://github.com/Tongyi-ConvAI/Qwen-Character/tree/main/Character-GenRM-NLHF.

We address the problem of fine-tuning pre-trained generative policies with reinforcement learning (RL) while preserving the multimodality of their action distributions. Existing methods for RL fine-tuning of generative policies (e.g., diffusion policies) improve task performance but often collapse diverse behaviors into a single reward-maximizing mode. To mitigate this issue, we propose an unsupervised mode discovery framework that uncovers latent behavioral modes within generative policies. The discovered modes enable the use of mutual information as an intrinsic reward, regularizing RL fine-tuning to enhance task success while maintaining behavioral diversity. Experiments on robotic manipulation tasks demonstrate that our method consistently outperforms conventional fine-tuning approaches, achieving higher success rates and preserving richer multimodal action distributions.

Safe reinforcement learning (Safe RL) seeks to optimize long-term performance while ensuring adherence to safety constraints. However, most existing approaches address safety in a simplified manner, typically by linearly combining rewards and costs, which provides limited guidance when safety and performance interact in complex, nonlinear ways. We present USB-RL (Unsupervised Score-Balanced Reinforcement Learning), a model-based framework that learns implicit safety–performance preferences directly from experience. Our approach infers a monotone partial-order score through unsupervised pairwise comparisons of long-horizon outcomes, capturing nuanced trade-offs without relying on manually tuned cost weights. The learned score guides model-based policy optimization by dynamically balancing safety and performance, enabling flexible and adaptive multi-step planning in imagination-based control. Across diverse safety benchmarks, USB-RL achieves strong returns while substantially reducing safety violations, demonstrating stable and interpretable safety–performance trade-offs.

World models enable learning policies via latent imagination, offering benefits such as history compression and sample efficiency. The primary challenge in applying world models to multi-agent tasks is that modeling multi-agent dynamics in latent space requires integrating information from different agents, often creating spurious correlations between their latent states. Existing methods either reconstruct the observation for each agent or employ communication to maintain correlation during execution, failing to learn disentangled latent states that are crucial for effective decentralized control. To address this, we present the Disentangled Multi-Agent World Model (DMAWM). It facilitates learning decentralized policies in the latent space through a novel architecture comprising independent agent modules and a shared environment module. During real-environment execution, agent modules independently process local information to form a factorized latent representation. The environment module is then trained to mirror the factorized structure generated by the agent modules, effectively disentangling individual latent states from the interaction dynamics. Consequently, imaginary rollouts generated by the environment module more faithfully simulate decentralized execution dynamics, facilitating the transfer of policies from imagination to decentralized execution. Empirically, DMAWM outperforms existing model-based and model-free approaches in convergence speed and final performance, with additional visualization demonstrating its efficacy in capturing agent interactions.

Reinforcement learning with verifiable rewards has driven recent advances in LLM post-training, in particular for reasoning. Policy optimization algorithms generate a number of responses for a given prompt and then effectively weight the corresponding gradients depending on the rewards. The most popular algorithms including GRPO, DAPO, and RLOO focus on ambiguous prompts, i.e., prompts with intermediate success probability, while downgrading gradients with very easy and very hard prompts. In this paper, we consider asymmetric prompt weightings that assign higher weights to prompts with low, or even zero, empirical success probability. We find that asymmetric weighting particularly benefits from-scratch RL (as in R1-Zero), where training traverses a wide accuracy range, and less so in post-SFT RL where the model already starts at high accuracy. We also provide theory that characterizes prompt weights which minimize the time needed to raise success probability from an initial level to a target accuracy under a fixed update budget. In low-success regimes, where informative responses are rare and response cost dominates, these optimal weights become asymmetric, upweighting low success probabilities and thereby accelerating effective-time convergence.

Integrating graph knowledge into Large Language Models (LLMs) via passive representation faces critical bottlenecks: limited context windows, unreliable numerical computation, and structural hallucinations. To solve this, we propose **GRASP** (**G**raph **R**easoning via **A**gentic **S**olving and **P**robing), shifting the paradigm from passive ingestion to proactive agentic exploration. By interleaving Neighbor Retrieval for on-demand probing with Code Interpreter as a deterministic solver, GRASP enables LLMs to autonomously navigate and compute over complex topologies. We employ a staged reinforcement learning strategy (GRPO) that transitions from visible tuning to a structure-blind environment, forcing the agent to develop genuine topological awareness. Evaluated on multi-domain graph reasoning benchmarks, our 4B model achieves a 53.06% average performance boost, surpassing SOTA baselines like DeepSeek-V3.2 and successfully generalizing to unseen tasks, with high potential for tackling sampling on million-node graphs and solving Hard-level LeetCode graph problems. Our implementation is open-sourced at https://github.com/PKU-ML/GRASP, with models hosted on Huggingface collection https://huggingface.co/collections/PKU-ML/grasp.

Traditional robot co-design approaches typically converge to one configuration, which do not explore the flexibility from reconfiguration on heterogeneous environments. On the other hand, existing designs for reconfigurable robots require human-designed configurations. We present Learning to Reconfigure, a holistic pipeline for configuration-control co-optimization of reconfigurable robots in heterogeneous locomotion tasks consisting of several sub-tasks. Our pipeline proposes low-level specialized primitives with a high-level scheduler. To jointly optimize configuration design and control, our primitives employ a multi-tail architecture that disentangles these distinct objectives. Building on this, the scheduler learns to dynamically switch configurations based on global task progress. We evaluate our pipeline on locomotion tasks across walking, flying, and swimming, and compare with the state-of-the-art baselines, including single-robot control and multi-morphology co-optimization algorithms. Quantitative results based on traversal progress show that our pipeline outperforms single-robot baselines by 5.95x average progress. Compared with the reconfiguration-free design given by the co-design algorithms, our robots also exhibit 9.81x progress on average. These results highlight the critical role of configuration adaptation in achieving versatile robotic autonomy in complex worlds.

Communication is a key component in multi-agent reinforcement learning (MARL) for mitigating partial observability, yet prior approaches often rely on inefficient information exchange or fail to transmit sufficient state information. To address this, we propose LLM-driven Multi-Agent Communication (LMAC), which leverages an LLM's reasoning capability to design a communication protocol that enables all agents to reconstruct the underlying state as accurately and uniformly as possible. LMAC iteratively refines the protocol using an explicit state-awareness criterion, improving state recovery while narrowing differences in agents' knowledge. Experiments on diverse MARL benchmarks show that LMAC improves state reconstruction across agents and yields substantial performance gains over prior communication baselines.

Computing the differential entropy of distributions known only up to a normalization constant is a fundamental challenge with broad theoretical and practical significance. While variational inference is highly scalable for density approximation from samples, its application to unnormalized densities remains under-explored due to the difficulty of constructing variational distributions that exploit the structure of the unnormalized density and are simultaneously expressive, tractable, efficiently samplable. Recently, Messaoud et al. (ICLR 24) introduced _P-SVGD_, a Stein variational inference method for this setting. We show, however, that \textit{P-SVGD} fails to scale to high dimensions due to incorrect invertibility assumptions, omission of a critical trace-of-Hessian term, and unstable divergence-control heuristics. We propose _MET-SVGD_, a principled extension of _P-SVGD_ that provides a general framework for stable _SVGD_ hyperparameter selection with global invertibility and convergence guarantees. Empirically, _MET-SVGD_ achieves up to $12\times$ and $16\times$ lower entropy estimation error than _P-SVGD_ and existing _SVGD_ baselines, respectively. On CIFAR-10 energy-based image generation, it improves FID by $80.4\%$ and yields $64\times$ higher training stability. In maximum-entropy reinforcement learning, it achieves up to $16\%$ higher returns than _P-SVGD_. Code is available at https://shorturl.at/fTG0G.

Reinforcement learning (RL) has become a dominant paradigm for enhancing LLMs' reasoning capabilities. However, RL algorithms with PPO-Clip are inherently limited by exploration collapse. Subsequent works remain primarily heuristic and fail to identify the essential cause of PPO-Clip’s failure. This work reveals the fundamental flaw of PPO-Clip: it implicitly measures policy discrepancy using Euclidean metric, which is theoretically inconsistent with the intrinsic geometry on the policy Riemannian manifold. This geometric mismatch results in overly conservative updates in low-probability regions while aggressive in high-probability regions, ultimately collapsing exploration. To correct this geometric flaw, we propose Riemannian Isometric Policy Optimization (RIPO), which guarantees isometric policy updates on the Riemannian manifold, effectively balancing exploration and exploitation. We further show that RIPO achieves a favorable bias-variance trade-off, which stabilizes optimization. Extensive experiments demonstrate that RIPO significantly surpasses existing LLM RL algorithms across seven competition-level benchmarks (up to 60% improvement over GRPO on AIME24).

Process reward models (PRMs) allow for fine-grained credit assignment in reinforcement learning (RL), and seemingly contrast with outcome reward models (ORMs), which assign a single reward to an entire trajectory. However, we provide theoretical proof in this work that the Group Relative Policy Optimization (GRPO) RL algorithm equipped with an ORM is in fact equivalent to a PRM-aware RL objective equipped with a non-trivial, Monte-Carlo-based PRM (given mild assumptions). Leveraging the framework of GRPO-as-a-PRM, we identify a flaw in the GRPO objective that interacts with imbalanced process steps and rewards to hinder both exploration and exploitation (under different conditions). We propose a simple modification to the algorithm to mitigate this defect ($\lambda$-GRPO), and show that LLMs tuned with $\lambda$-GRPO outperform LLMs tuned with standard GRPO on downstream reasoning tasks$\textemdash$and reach peak performance more rapidly. These results show that we can leverage the hidden, built-in PRM structure within the vanilla GRPO algorithm to boost model performance without employing an explicit PRM, and with a negligible impact on training time and cost.

Programmatic policies are often reported to generalize better than neural policies in reinforcement learning (RL) benchmarks. We revisit some of these claims and show that much of the observed gap arises from uncontrolled experimental factors rather than intrinsic representational reasons. Re-evaluating three core benchmarks used in influential papers---TORCS, Karel, and Parking---we find that neural policies, when trained with a few modifications, such as sparse observations and cautious intrinsic reward functions, can match or exceed the out-of-distribution (OOD) generalization of programmatic policies. We argue that a representation enables OOD generalization if (i) the policy space it induces includes a generalizing policy and (ii) the search algorithm can find it. The neural and programmatic policies in prior work are comparable in OOD generalization because the domain-specific languages used induce policy spaces similar to those of neural networks, and our modifications help the gradient search find generalizing solutions. By disentangling representational factors from experimental confounds, we advance our understanding of what makes a representation succeed or fail at OOD generalization.