Distributionally Robust RL

RL: Reinforcement Learning

Ask about this topic

Overview

Distributionally robust reinforcement learning seeks policies that retain good performance when the true dynamics, rewards, or deployment tasks differ from those seen during training. Common approaches define ambiguity sets using Wasserstein distance, total variation, or divergence measures, then optimize worst-case Bellman values with robust Q-learning, actor-critic, or planning methods. Work spans tabular and function-approximation settings, average-reward control, safety, offline learning, multi-agent games, and reinforcement learning from human feedback. A central tension is that larger uncertainty sets can improve protection against distribution shift but cause excessive pessimism and higher sample costs; support mismatch can make reliable learning impossible without additional assumptions or useful side information.

Latest papers

CardsList

34 papers0 papers in the last four weeks, down 100% on the four weeks before. 0.0% of all new papers.

  1. Distributionally Robust Deep Q-Learning

    May 25, 2025Chung I Lu, Julian Sester, Aijia ZhangRobust Markov Decision ProcessesDeep Q-Learning

    National University of Singapore, Department of Mathematics, 21 Lower Kent Ridge Road, 119077.

  2. Model-Free Robust Average-Reward Reinforcement Learning with Sample Complexity Analysis

    May 18, 2025Zachary Roch, George Atia, Yue WangReinforcement LearningAverage-Reward RL

    Department of Electrical and Computer Engineering, University of Central Florida, Orlando, Florida, USA. · Department of Computer Science, University of Central Florida, Orlando, Florida, USA.

  3. Distributionally Robust Reinforcement Learning with Human Feedback

    Mar 1, 2025Debmalya Mandal, Paulius Sasnauskas, Goran RadanovicDirect Preference OptimizationLLM Fine-Tuning

    Dept. of Computer Science University of Warwick, UK · Department of Computing Science University of Alberta, Edmonton, Canada · Max-Planck Institute for Software Systems, Germany

  4. Distributionally Robust Reinforcement Learning with Interactive Data Collection: Fundamental Hardness and Near-Optimal Algorithms

    Apr 4, 2024Miao Lu, Han Zhong, Tong Zhang +1Distributionally Robust RLRobust RL

    †Department of Management Science and Engineering, Stanford University. · ‡Center for Data Science, Peking University. · §Department of Computer Science, University of Illinois Urbana-Champaign.