English

Near-Optimal Regret for Policy Optimization in Contextual MDPs with General Offline Function Approximation

Machine Learning 2026-02-17 v1

Abstract

We introduce \texttt{OPO-CMDP}, the first policy optimization algorithm for stochastic Contextual Markov Decision Process (CMDPs) under general offline function approximation. Our approach achieves a high probability regret bound of O~(H4TSAlog(FP)),\widetilde{O}(H^4\sqrt{T|S||A|\log(|\mathcal{F}||\mathcal{P}|)}), where SS and AA denote the state and action spaces, HH the horizon length, TT the number of episodes, and F,P\mathcal{F}, \mathcal{P} the finite function classes used to approximate the losses and dynamics, respectively. This is the first regret bound with optimal dependence on S|S| and A|A|, directly improving the current state-of-the-art (Qian, Hu, and Simchi-Levi, 2024). These results demonstrate that optimistic policy optimization provides a natural, computationally superior and theoretically near-optimal path for solving CMDPs.

Keywords

Cite

@article{arxiv.2602.13706,
  title  = {Near-Optimal Regret for Policy Optimization in Contextual MDPs with General Offline Function Approximation},
  author = {Orin Levy and Aviv Rosenberg and Alon Cohen and Yishay Mansour},
  journal= {arXiv preprint arXiv:2602.13706},
  year   = {2026}
}
R2 v1 2026-07-01T10:36:43.840Z