Q-Flow: Stable and Expressive Reinforcement Learning with Flow-Based Policy

TL;DR

Q-Flow introduces a novel reinforcement learning framework that uses flow-based policies to achieve stable and expressive decision-making, overcoming previous optimization challenges.

Contribution

It leverages flow dynamics to propagate value information, enabling stable gradient-based optimization without unrolling solvers, balancing stability and expressivity.

Findings

Q-Flow outperforms state-of-the-art baselines by 10.6 percentage points on OGBench.

It enables stable online adaptation within the same framework.

The method effectively bridges the gap between policy stability and expressivity.

Abstract

There is growing interest in utilizing flow-based models as decision-making policies in reinforcement learning due to their high expressive capacity. However, effectively leveraging this expressivity for value maximization remains challenging, as naive gradient-based optimization requires backpropagating through numerical solvers and often leads to instability. Existing approaches typically address this issue by restricting the expressive capacity of flow-based policies, resulting in a trade-off between optimization stability and representational flexibility. To resolve this, we introduce Q-Flow, a framework that leverages the deterministic nature of flow dynamics to explicitly propagate terminal trajectory value to intermediate latent states along the policy-induced flow. This formulation enables stable policy optimization using intermediate value gradients without unrolling the numerical solver, effectively bridging the gap between stability and expressivity. We evaluate Q-Flow in the offline learning setting on the challenging OGBench suite, where it consistently outperforms state-of-the-art baselines by an average of 10.6 percentage points, while also enabling stable online adaptation within the same framework.