AAAI 2026 accepted papers on Reinforcement learning

Visual and representation RL · 62 papers

• WorldRFT: Latent World Model Planning with Reinforcement Fine-Tuning for Autonomous Driving (5 citations)
• Actor-Critic for Continuous Action Chunks: A Reinforcement Learning Framework for Long-Horizon Robotic Manipulation with Sparse Reward (2 citations)
• Delphi: A Neuro-Symbolic Framework for Individualized, Safe and Interpretable Treatment Recommendation (2 citations)
• Dexterous Manipulation Transfer via Progressive Kinematic-Dynamic Alignment (2 citations)
• DRMD: Deep Reinforcement Learning for Malware Detection Under Concept Drift (2 citations)
• Learning Diverse Bimanual Dexterous Manipulation Skills from Human Demonstrations (2 citations)
• AcoustoReinforce: Multi-Particle Acoustophoretic Path Planning with Deep Reinforcement Learning (1 citation)
• Brownian Bridge Augmented Surrogate Simulation and Injection Planning for Geological CO2 Storage (1 citation)
• FoAM: Foresight-Augmented Multi-Task Imitation Policy for Robotic Manipulation (1 citation)
• Improving Deepfake Detection with Reinforcement Learning-Based Adaptive Data Augmentation (1 citation)
• Out-of-Distribution Generalization with a SPARC: Racing 100 Unseen Vehicles with a Single Policy (1 citation)
• VILTA: A VLM-in-the-Loop Adversary for Enhancing Driving Policy Robustness (1 citation)

and 50 more in the interactive map.

RL from feedback · 25 papers

• UI-R1: Enhancing Efficient Action Prediction of GUI Agents by Reinforcement Learning (6 citations)
• GUI-G²: Gaussian Reward Modeling for GUI Grounding (3 citations)
• InfiGUI-G1: Advancing GUI Grounding with Adaptive Exploration Policy Optimization (1 citation)
• MedGR2: Breaking the Data Barrier for Medical Reasoning via Generative Reward Learning (1 citation)
• Preference Is More than Comparisons: Rethinking Dueling Bandits with Augmented Human Feedback (1 citation)
• Reward Model Evaluation via Automatically-Ranked Policy Alignment (1 citation)
• A Rolling Stone Gathers No Moss: Adaptive Policy Optimization for Stable Self-Evaluation in Large Multimodal Models
• Belief-Driven Value Alignment for Human-Robot Collaboration
• Facilitating Early Maladaptive Schema–Guided Polite and Empathetic Psychotherapeutic Support: An LLM-Driven MoE-RL-Based Dialogue System
• FDC-Ground: Improving GRPO for GUI Grounding via Exponential Rewards and Fact-Aligned Pruning
• MetaAct-RL: Training Language Models for Reasoning Through Meta-Action-Based Reinforcement Learning
• MRACL: Multi-Reward Space Guided Adaptive Curriculum Reinforcement Learning for LLMs

and 13 more in the interactive map.

Policy gradient and actor-critic · 23 papers

• DIMM: Decoupled Multi-hierarchy Kalman Filter via Reinforcement Learning (3 citations)
• CHDP: Cooperative Hybrid Diffusion Policies for Reinforcement Learning in Parameterized Action Space (2 citations)
• Enhancing Control Policy Smoothness by Aligning Actions with Predictions from Preceding States (1 citation)
• One-Step Generative Policies with Q-Learning: A Reformulation of MeanFlow (1 citation)
• Proactive Constrained Policy Optimization with Preemptive Penalty (1 citation)
• Stability-Aware Reinforcement Learning for Robust Class Integration Test Order Generation (1 citation)
• A Differential Perspective on Distributional Reinforcement Learning
• Behaviour Policy Optimization: Provably Lower Variance Return Estimates for Off-Policy Reinforcement Learning
• DiffOP: Reinforcement Learning of Optimization-Based Control Policies via Implicit Policy Gradients
• Improving Stochastic Action-Constrained Reinforcement Learning via Truncated Distributions
• Learning Branching Policies for MILPs with Proximal Policy Optimization
• LPPG-RL: Lexicographically Projected Policy Gradient Reinforcement Learning with Subproblem Exploration

and 11 more in the interactive map.

MDPs and planning · 14 papers

• ORVIT: Near-Optimal Online Distributionally Robust Reinforcement Learning (1 citation)
• Asymptotic and Finite Sample Analysis of Nonexpansive Stochastic Approximations with Markovian Noise
• Distributionally Robust Online Markov Game with Linear Function Approximation
• Do It for HER: First-Order Temporal Logic Reward Specification in Reinforcement Learning
• Efficient Multiagent Planning via Shared Action Suggestions
• Policy Newton Methods for Distortion Riskmetrics
• Potential Outcome Rankings for Counterfactual Decision Making
• Best-Effort Policies for Robust Markov Decision Processes
• Constrained and Robust Policy Synthesis with Satisfiability-Modulo-Probabilistic-Model-Checking
• Efficient Solution and Learning of Robust Factored MDPs
• Good-for-MDP State Reduction for Stochastic LTL Planning
• Online Robust Planning Under Model Uncertainty: A Sample-Based Approach

and 2 more in the interactive map.

Offline RL · 10 papers

• Treatment Stitching with Schrödinger Bridge for Enhancing Offline Reinforcement Learning in Adaptive Treatment Strategies (2 citations)
• Balancing Signal and Variance: Adaptive Offline RL Post-Training for VLA Flow Models (1 citation)
• Offline Meta-Reinforcement Learning with Flow-Based Task Inference and Adaptive Correction of Feature Overgeneralization (1 citation)
• Behavior Regularization with Flow Latent Policy for Offline Reinforcement Learning
• Enhancing Diffusion Policies with Distribution-Matching Generator in Offline Reinforcement Learning
• MetaTrader: Learning to Generalize RL Trading Policies Beyond Offline Data
• Reliability-Guaranteed and Reward-Seeking Sequence Modeling for Model-Based Offline Reinforcement Learning
• State Proficiency-Based Adaptive Fine-Tuning for Offline-to-Online Reinforcement Learning
• Variational OOD State Correction for Offline Reinforcement Learning
• Advancing Safe Mechanical Ventilation Using Offline RL with Hybrid Actions and Clinically Aligned Rewards

Imitation learning · 5 papers

• Hexaïssa: Standing on Giants’ Shoulders—Routing the Best Chess Engines with Mixture-of-Experts and Latent Reward Learning
• Inferring Heterogeneous Private Valuations from Offline Market Data via Entropic Risk-Sensitive Utility Maximization
• Misalignment from Treating Means as Ends
• Reward Redistribution via Gaussian Process Likelihood Estimation
• SafeMIL: Learning Offline Safe Imitation Policy from Non-Preferred Trajectories