Achieving Logarithmic Regret in KL-Regularized Zero-Sum Markov Games

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Nayak, Anupam, Yang, Tong, Yagan, Osman, Joshi, Gauri, Chi, Yuejie
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866910010802438144
author Nayak, Anupam
Yang, Tong
Yagan, Osman
Joshi, Gauri
Chi, Yuejie
author_facet Nayak, Anupam
Yang, Tong
Yagan, Osman
Joshi, Gauri
Chi, Yuejie
contents Reverse Kullback-Leibler (KL) divergence-based regularization with respect to a fixed reference policy is widely used in modern reinforcement learning to preserve the desired traits of the reference policy and sometimes to promote exploration (using uniform reference policy, known as entropy regularization). Beyond serving as a mere anchor, the reference policy can also be interpreted as encoding prior knowledge about good actions in the environment. In the context of alignment, recent game-theoretic approaches have leveraged KL regularization with pretrained language models as reference policies, achieving notable empirical success in self-play methods. Despite these advances, the theoretical benefits of KL regularization in game-theoretic settings remain poorly understood. In this work, we develop and analyze algorithms that provably achieve improved sample efficiency under KL regularization. We study both two-player zero-sum matrix games and Markov games: for matrix games, we propose OMG, an algorithm based on best response sampling with optimistic bonuses, and extend this idea to Markov games through the algorithm SOMG, which also uses best response sampling and a novel concept of superoptimistic bonuses. Both algorithms achieve a logarithmic regret in $T$ that scales inversely with the KL regularization strength $β$ in addition to the traditional $\widetilde{\mathcal{O}}(\sqrt{T})$ regret without the $β^{-1}$ dependence.
format Preprint
id arxiv_https___arxiv_org_abs_2510_13060
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Achieving Logarithmic Regret in KL-Regularized Zero-Sum Markov Games
Nayak, Anupam
Yang, Tong
Yagan, Osman
Joshi, Gauri
Chi, Yuejie
Machine Learning
Computer Science and Game Theory
Optimization and Control
Reverse Kullback-Leibler (KL) divergence-based regularization with respect to a fixed reference policy is widely used in modern reinforcement learning to preserve the desired traits of the reference policy and sometimes to promote exploration (using uniform reference policy, known as entropy regularization). Beyond serving as a mere anchor, the reference policy can also be interpreted as encoding prior knowledge about good actions in the environment. In the context of alignment, recent game-theoretic approaches have leveraged KL regularization with pretrained language models as reference policies, achieving notable empirical success in self-play methods. Despite these advances, the theoretical benefits of KL regularization in game-theoretic settings remain poorly understood. In this work, we develop and analyze algorithms that provably achieve improved sample efficiency under KL regularization. We study both two-player zero-sum matrix games and Markov games: for matrix games, we propose OMG, an algorithm based on best response sampling with optimistic bonuses, and extend this idea to Markov games through the algorithm SOMG, which also uses best response sampling and a novel concept of superoptimistic bonuses. Both algorithms achieve a logarithmic regret in $T$ that scales inversely with the KL regularization strength $β$ in addition to the traditional $\widetilde{\mathcal{O}}(\sqrt{T})$ regret without the $β^{-1}$ dependence.
title Achieving Logarithmic Regret in KL-Regularized Zero-Sum Markov Games
topic Machine Learning
Computer Science and Game Theory
Optimization and Control
url https://arxiv.org/abs/2510.13060