ICML 2025 accepted papers on Reinforcement learning

Visual and representation RL · 73 papers

• Empirical Design in Reinforcement Learning (7 citations)
• DINO-WM: World Models on Pre-trained Visual Features enable Zero-shot Planning (5 citations)
• Benchmarking Quantum Reinforcement Learning (3 citations)
• EARL-BO: Reinforcement Learning for Multi-Step Lookahead, High-Dimensional Bayesian Optimization (1 citation)
• Hyper: Hyperparameter Robust Efficient Exploration in Reinforcement Learning (1 citation)
• Improving Transformer World Models for Data-Efficient RL (1 citation)
• MENTOR: Mixture-of-Experts Network with Task-Oriented Perturbation for Visual Reinforcement Learning (1 citation)
• Reinforcement Learning for Quantum Control under Physical Constraints (1 citation)
• Stealing That Free Lunch: Exposing the Limits of Dyna-Style Reinforcement Learning (1 citation)
• Trajectory World Models for Heterogeneous Environments (1 citation)
• A Multi-Region Brain Model to Elucidate the Role of Hippocampus in Spatially Embedded Decision-Making
• Action-Dependent Optimality-Preserving Reward Shaping

and 61 more in the interactive map.

Policy gradient and actor-critic · 51 papers

• A Model of Place Field Reorganization During Reward Maximization (3 citations)
• Gradient Boosting Reinforcement Learning (1 citation)
• The Perils of Optimizing Learned Reward Functions: Low Training Error Does Not Guarantee Low Regret (1 citation)
• A Theoretical Justification for Asymmetric Actor-Critic Algorithms
• Actor-Critics Can Achieve Optimal Sample Efficiency
• Analytical Lyapunov Function Discovery: An RL-based Generative Approach
• Beyond CVaR: Leveraging Static Spectral Risk Measures for Enhanced Decision-Making in Distributional Reinforcement Learning
• Calibrated Value-Aware Model Learning with Probabilistic Environment Models
• Categorical Distributional Reinforcement Learning with Kullback-Leibler Divergence: Convergence and Asymptotics
• Controlling Underestimation Bias in Constrained Reinforcement Learning for Safe Exploration
• Convergence Analysis of Policy Gradient Methods with Dynamic Stochasticity
• Decoding Rewards in Competitive Games: Inverse Game Theory with Entropy Regularization

and 39 more in the interactive map.

RL from feedback · 33 papers

• SFT Memorizes, RL Generalizes: A Comparative Study of Foundation Model Post-training (6 citations)
• Theoretical guarantees on the best-of-n alignment policy (1 citation)
• BOPO: Neural Combinatorial Optimization via Best-anchored and Objective-guided Preference Optimization
• Can RLHF be More Efficient with Imperfect Reward Models? A Policy Coverage Perspective
• CLARIFY: Contrastive Preference Reinforcement Learning for Untangling Ambiguous Queries
• Comparing Comparisons: Informative and Easy Human Feedback with Distinguishability Queries
• Deep Reinforcement Learning from Hierarchical Preference Design
• Design Considerations in Offline Preference-based RL
• Do We Need to Verify Step by Step? Rethinking Process Supervision from a Theoretical Perspective
• DPO Meets PPO: Reinforced Token Optimization for RLHF
• Enhancing Rating-Based Reinforcement Learning to Effectively Leverage Feedback from Large Vision-Language Models
• InfAlign: Inference-aware language model alignment

and 21 more in the interactive map.

Offline RL · 26 papers

• Adaptive Learn-then-Test: Statistically Valid and Efficient Hyperparameter Selection
• Behavior-Regularized Diffusion Policy Optimization for Offline Reinforcement Learning
• C2IQL: Constraint-Conditioned Implicit Q-learning for Safe Offline Reinforcement Learning
• Diversity By Design: Leveraging Distribution Matching for Offline Model-Based Optimization
• Flow Q-Learning
• Habitizing Diffusion Planning for Efficient and Effective Decision Making
• In-Context Reinforcement Learning From Suboptimal Historical Data
• Learning to Trust Bellman Updates: Selective State-Adaptive Regularization for Offline RL
• Meta-Black-Box-Optimization through Offline Q-function Learning
• MODULI: Unlocking Preference Generalization via Diffusion Models for Offline Multi-Objective Reinforcement Learning
• Offline-to-Online Reinforcement Learning with Classifier-Free Diffusion Generation
• Provable Zero-Shot Generalization in Offline Reinforcement Learning

and 14 more in the interactive map.

MDPs and planning · 22 papers

• Overcoming the Curse of Dimensionality in Reinforcement Learning Through Approximate Factorization (2 citations)
• Accelerating Quantum Reinforcement Learning with a Quantum Natural Policy Gradient Based Approach (1 citation)
• A Computationally Efficient Algorithm for Infinite-Horizon Average-Reward Linear MDPs
• A Reductions Approach to Risk-Sensitive Reinforcement Learning with Optimized Certainty Equivalents
• A Sharper Global Convergence Analysis for Average Reward Reinforcement Learning via an Actor-Critic Approach
• Adaptive Exploration for Multi-Reward Multi-Policy Evaluation
• Intersectional Fairness in Reinforcement Learning with Large State and Constraint Spaces
• Learning Policy Committees for Effective Personalization in MDPs with Diverse Tasks
• Learning Utilities from Demonstrations in Markov Decision Processes
• Model-Based Exploration in Monitored Markov Decision Processes
• Multiple-policy Evaluation via Density Estimation
• Partially Observable Reinforcement Learning with Memory Traces

and 10 more in the interactive map.

Imitation learning · 11 papers

• Action-Constrained Imitation Learning
• Active Fine-Tuning of Multi-Task Policies
• ELEMENTAL: Interactive Learning from Demonstrations and Vision-Language Models for Reward Design in Robotics
• IL-SOAR : Imitation Learning with Soft Optimistic Actor cRitic
• Learning from Suboptimal Data in Continuous Control via Auto-Regressive Soft Q-Network
• Provably Efficient Exploration in Inverse Constrained Reinforcement Learning
• Reward-free World Models for Online Imitation Learning
• Score-Based Diffusion Policy Compatible with Reinforcement Learning via Optimal Transport
• Zero-Shot Offline Imitation Learning via Optimal Transport
• Diversifying Robot Locomotion Behaviors with Extrinsic Behavioral Curiosity
• Policy Regularization on Globally Accessible States in Cross-Dynamics Reinforcement Learning