NeurIPS 2025 accepted papers on Reinforcement learning
323 of the 5,812 papers accepted at NeurIPS 2025 (5.6%) are on Reinforcement learning, compared with 5.9% at NeurIPS 2024. They fall into 6 topics.
See NeurIPS 2025 in the interactive map
Working on something in this area? Describe your idea in scime atlas to see the closest accepted papers by meaning and how crowded the spot has become.
Topics and papers
Visual and representation RL · 108 papers
- Gymnasium: A Standard Interface for Reinforcement Learning Environments (21 citations)
- Compliant Residual DAgger: Improving Real-World Contact-Rich Manipulation with Human Corrections (5 citations)
- Meta-World+: An Improved, Standardized, RL Benchmark (5 citations)
- RAD: Training an End-to-End Driving Policy via Large-Scale 3DGS-based Reinforcement Learning (4 citations)
- From Forecasting to Planning: Policy World Model for Collaborative State-Action Prediction (2 citations)
- MetaBox-v2: A Unified Benchmark Platform for Meta-Black-Box Optimization (2 citations)
- Mixture-of-Experts Meets In-Context Reinforcement Learning (2 citations)
- NAVIX: Scaling MiniGrid Environments with JAX (2 citations)
- NeurIPS should lead scientific consensus on AI policy (2 citations)
- Offline Goal-conditioned Reinforcement Learning with Quasimetric Representations (2 citations)
- VLMLight: Safety-Critical Traffic Signal Control via Vision-Language Meta-Control and Dual-Branch Reasoning Architecture (2 citations)
- WHAT MAKES MATH PROBLEMS HARD FOR REINFORCEMENT LEARNING: A CASE STUDY (2 citations)
and 96 more in the interactive map.
RL from feedback · 80 papers
- Flow-GRPO: Training Flow Matching Models via Online RL (36 citations)
- Learning to Reason under Off-Policy Guidance (6 citations)
- Open-Reasoner-Zero: An Open Source Approach to Scaling Up Reinforcement Learning on the Base Model (5 citations)
- Group-in-Group Policy Optimization for LLM Agent Training (4 citations)
- GUI-G1: Understanding R1-Zero-Like Training for Visual Grounding in GUI Agents (4 citations)
- SE-GUI: Enhancing Visual Grounding for GUI Agents via Self-Evolutionary Reinforcement Learning (4 citations)
- CPPO: Accelerating the Training of Group Relative Policy Optimization-Based Reasoning Models (3 citations)
- DriveDPO: Policy Learning via Safety DPO For End-to-End Autonomous Driving (3 citations)
- R1-ShareVL: Incentivizing Reasoning Capabilities of Multimodal Large Language Models via Share-GRPO (3 citations)
- Stop Summation: Min-Form Credit Assignment Is All Process Reward Model Needs for Reasoning (3 citations)
- Aligning Transformers with Continuous Feedback via Energy Rank Alignment (2 citations)
- Reasoning Gym: Reasoning Environments for Reinforcement Learning with Verifiable Rewards (2 citations)
and 68 more in the interactive map.
Policy gradient and actor-critic · 55 papers
- Proxy Target: Bridging the Gap Between Discrete Spiking Neural Networks and Continuous Control (2 citations)
- RCCDA: Adaptive Model Updates in the Presence of Concept Drift under a Constrained Resource Budget (2 citations)
- CHPO: Constrained Hybrid-action Policy Optimization for Reinforcement Learning (1 citation)
- Compute-Optimal Scaling for Value-Based Deep RL (1 citation)
- Constrained Diffusers for Safe Planning and Control (1 citation)
- Continuous Q-Score Matching: Diffusion Guided Reinforcement Learning for Continuous-Time Control (1 citation)
- FairDICE: Fairness-Driven Offline Multi-Objective Reinforcement Learning (1 citation)
- Flow-Based Policy for Online Reinforcement Learning (1 citation)
- GenPO: Generative Diffusion Models Meet On-Policy Reinforcement Learning (1 citation)
- Scaling Off-Policy Reinforcement Learning with Batch and Weight Normalization (1 citation)
- A Differential and Pointwise Control Approach to Reinforcement Learning
- A Principled Path to Fitted Distributional Evaluation
and 43 more in the interactive map.
MDPs and planning · 39 papers
- Faster Fixed-Point Methods for Multichain MDPs (1 citation)
- Optimal Single-Policy Sample Complexity and Transient Coverage for Average-Reward Offline RL (1 citation)
- Skill-Driven Neurosymbolic State Abstractions (1 citation)
- A Generalized Bisimulation Metric of State Similarity between Markov Decision Processes: From Theoretical Propositions to Applications
- Beyond Scalar Rewards: An Axiomatic Framework for Lexicographic MDPs
- Breaking the Order Barrier: Off-Policy Evaluation for Confounded POMDPs
- Computational Hardness of Reinforcement Learning with Partial $q^{\pi}$-Realizability
- Deployment Efficient Reward-Free Exploration with Linear Function Approximation
- Efficient Policy Optimization in Robust Constrained MDPs with Iteration Complexity Guarantees
- Exploration from a Primal-Dual Lens: Value-Incentivized Actor-Critic Methods for Sample-Efficient Online RL
- Finite-Sample Analysis of Policy Evaluation for Robust Average Reward Reinforcement Learning
- Finite-Time Bounds for Average-Reward Fitted Q-Iteration
and 27 more in the interactive map.
Offline RL · 25 papers
- Value-Guided Decision Transformer: A Unified Reinforcement Learning Framework for Online and Offline Settings (2 citations)
- Forecasting in Offline Reinforcement Learning for Non-stationary Environments (1 citation)
- Offline Guarded Safe Reinforcement Learning for Medical Treatment Optimization Strategies (1 citation)
- Online Optimization for Offline Safe Reinforcement Learning (1 citation)
- Reinforcement Learning with Action Chunking (1 citation)
- A Clean Slate for Offline Reinforcement Learning
- Adaptive Neighborhood-Constrained Q Learning for Offline Reinforcement Learning
- ADG: Ambient Diffusion-Guided Dataset Recovery for Corruption-Robust Offline Reinforcement Learning
- Analytic Energy-Guided Policy Optimization for Offline Reinforcement Learning
- Boundary-to-Region Supervision for Offline Safe Reinforcement Learning
- BraVE: Offline Reinforcement Learning for Discrete Combinatorial Action Spaces
- Don’t Trade Off Safety: Diffusion Regularization for Constrained Offline RL
and 13 more in the interactive map.
Imitation learning · 16 papers
- ReSim: Reliable World Simulation for Autonomous Driving (5 citations)
- Adversarial Locomotion and Motion Imitation for Humanoid Policy Learning (2 citations)
- Learning from Demonstrations via Capability-Aware Goal Sampling (1 citation)
- A Smooth Sea Never Made a Skilled SAILOR: Robust Imitation via Learning to Search
- Blindfolded Experts Generalize Better: Insights from Robotic Manipulation and Videogames
- DPAIL: Training Diffusion Policy for Adversarial Imitation Learning without Policy Optimization
- Imitation Beyond Expectation Using Pluralistic Stochastic Dominance
- Interactive and Hybrid Imitation Learning: Provably Beating Behavior Cloning
- IOSTOM: Offline Imitation Learning from Observations via State Transition Occupancy Matching
- Latent Policy Barrier: Learning Robust Visuomotor Policies by Staying In-Distribution
- Learning Parameterized Skills from Demonstrations
- No Experts, No Problem: Avoidance Learning from Bad Demonstrations
and 4 more in the interactive map.