Boosting Maximum Entropy Reinforcement Learning via One-Step Flow Matching
This work addresses the problem of high inference latency in expressive policies for reinforcement learning practitioners, offering an incremental improvement by combining existing techniques in a novel way.
The paper tackles the challenge of integrating one-step Flow Matching into Maximum Entropy Reinforcement Learning by proposing FLAME, which bypasses partition function estimation and corrects bias in entropy estimation, resulting in performance matching multi-step diffusion policies with lower inference cost on MuJoCo benchmarks.
Diffusion policies are expressive yet incur high inference latency. Flow Matching (FM) enables one-step generation, but integrating it into Maximum Entropy Reinforcement Learning (MaxEnt RL) is challenging: the optimal policy is an intractable energy-based distribution, and the efficient log-likelihood estimation required to balance exploration and exploitation suffers from severe discretization bias. We propose \textbf{F}low-based \textbf{L}og-likelihood-\textbf{A}ware \textbf{M}aximum \textbf{E}ntropy RL (\textbf{FLAME}), a principled framework that addresses these challenges. First, we derive a Q-Reweighted FM objective that bypasses partition function estimation via importance reweighting. Second, we design a decoupled entropy estimator that rigorously corrects bias, which enables efficient exploration and brings the policy closer to the optimal MaxEnt policy. Third, we integrate the MeanFlow formulation to achieve expressive and efficient one-step control. Empirical results on MuJoCo show that FLAME outperforms Gaussian baselines and matches multi-step diffusion policies with significantly lower inference cost. Code is available at https://github.com/lzqw/FLAME.