OM2P: Offline Multi-Agent Mean-Flow Policy

TL;DR

OM2P introduces an efficient offline multi-agent reinforcement learning algorithm using mean-flow models, significantly reducing memory and training time while maintaining superior performance in benchmark tasks.

Contribution

It is the first to successfully integrate mean-flow generative models into offline MARL, enabling one-step action sampling and improved efficiency.

Findings

Achieves up to 3.8x reduction in GPU memory usage.

Realizes up to 10.8x faster training times.

Demonstrates superior performance on Multi-Agent Particle and MuJoCo benchmarks.

Abstract

Generative models, especially diffusion and flow-based models, have been promising in offline multi-agent reinforcement learning. However, integrating powerful generative models into this framework poses unique challenges. In particular, diffusion and flow-based policies suffer from low sampling efficiency due to their iterative generation processes, making them impractical in time-sensitive or resource-constrained settings. To tackle these difficulties, we propose OM2P (Offline Multi-Agent Mean-Flow Policy), a novel offline MARL algorithm to achieve efficient one-step action sampling. To address the misalignment between generative objectives and reward maximization, we introduce a reward-aware optimization scheme that integrates a carefully-designed mean-flow matching loss with Q-function supervision. Additionally, we design a generalized timestep distribution and a derivative-free estimation strategy to reduce memory overhead and improve training stability. Empirical evaluations on Multi-Agent Particle and MuJoCo benchmarks demonstrate that OM2P achieves superior performance, with up to a 3.8x reduction in GPU memory usage and up to a 10.8x speed-up in training time. Our approach represents the first to successfully integrate mean-flow model into offline MARL, paving the way for practical and scalable generative policies in cooperative multi-agent settings.