EXPO: Stable Reinforcement Learning with Expressive Policies

TL;DR

This paper introduces EXPO, a novel reinforcement learning method that enhances training stability and sample efficiency for expressive policies by using an on-the-fly editing policy instead of direct value optimization.

Contribution

The paper proposes a new algorithm, EXPO, which improves stability and efficiency in training expressive policies by combining a base policy with an action editing policy.

Findings

Achieves 2-3x better sample efficiency than prior methods.

Effective in fine-tuning pretrained policies with offline data.

Improves stability in training expressive policies like diffusion and flow-matching.

Abstract

We study the problem of training and fine-tuning expressive policies with online reinforcement learning (RL) given an offline dataset. Training expressive policy classes with online RL present a unique challenge of stable value maximization. Unlike simpler Gaussian policies commonly used in online RL, expressive policies like diffusion and flow-matching policies are parameterized by a long denoising chain, which hinders stable gradient propagation from actions to policy parameters when optimizing against some value function. Our key insight is that we can address stable value maximization by avoiding direct optimization over value with the expressive policy and instead construct an on-the-fly RL policy to maximize Q-value. We propose Expressive Policy Optimization (EXPO), a sample-efficient online RL algorithm that utilizes an on-the-fly policy to maximize value with two parameterized policies -- a larger expressive base policy trained with a stable imitation learning objective and a light-weight Gaussian edit policy that edits the actions sampled from the base policy toward a higher value distribution. The on-the-fly policy optimizes the actions from the base policy with the learned edit policy and chooses the value maximizing action from the base and edited actions for both sampling and temporal-difference (TD) backup. Our approach yields up to 2-3x improvement in sample efficiency on average over prior methods both in the setting of fine-tuning a pretrained policy given offline data and in leveraging offline data to train online.