Flow Q-Learning
This addresses a key bottleneck in offline RL for researchers and practitioners by enabling more stable and efficient training of complex action distributions.
The paper tackles the challenge of training expressive flow-matching policies in offline reinforcement learning by introducing flow Q-learning (FQL), which uses a one-step policy to avoid unstable recursive backpropagation and costly iterative action generation, resulting in strong performance across 73 challenging tasks.
We present flow Q-learning (FQL), a simple and performant offline reinforcement learning (RL) method that leverages an expressive flow-matching policy to model arbitrarily complex action distributions in data. Training a flow policy with RL is a tricky problem, due to the iterative nature of the action generation process. We address this challenge by training an expressive one-step policy with RL, rather than directly guiding an iterative flow policy to maximize values. This way, we can completely avoid unstable recursive backpropagation, eliminate costly iterative action generation at test time, yet still mostly maintain expressivity. We experimentally show that FQL leads to strong performance across 73 challenging state- and pixel-based OGBench and D4RL tasks in offline RL and offline-to-online RL. Project page: https://seohong.me/projects/fql/