SuperFlow: Training Flow Matching Models with RL on the Fly

TL;DR

SuperFlow introduces an RL training framework for flow-based models that improves efficiency and performance in text-to-image tasks by adaptive sampling and step-level advantage computation, reducing training time and enhancing quality.

Contribution

It proposes variance-aware sampling and continuous-time advantage estimation for flow models, addressing inefficiencies in existing RL training methods.

Findings

Reduces training steps by up to 56.3%.

Decreases training time by up to 16.7%.

Improves performance on text-to-image tasks by up to 47.2%.

Abstract

Recent progress in flow-based generative models and reinforcement learning (RL) has improved text-image alignment and visual quality. However, current RL training for flow models still has two main problems: (i) GRPO-style fixed per-prompt group sizes ignore variation in sampling importance across prompts, which leads to inefficient sampling and slower training; and (ii) trajectory-level advantages are reused as per-step estimates, which biases credit assignment along the flow. We propose SuperFlow, an RL training framework for flow-based models that adjusts group sizes with variance-aware sampling and computes step-level advantages in a way that is consistent with continuous-time flow dynamics. Empirically, SuperFlow reaches promising performance while using only 5.4% to 56.3% of the original training steps and reduces training time by 5.2% to 16.7% without any architectural changes. On standard text-to-image (T2I) tasks, including text rendering, compositional image generation, and human preference alignment, SuperFlow improves over SD3.5-M by 4.6% to 47.2%, and over Flow-GRPO by 1.7% to 16.0%.