Fine-tuning Flow Matching Generative Models with Intermediate Feedback

TL;DR

This paper introduces AC-Flow, an actor-critic framework for stable and effective fine-tuning of flow-based text-to-image models using intermediate feedback, overcoming previous training instabilities.

Contribution

The paper proposes a novel AC-Flow method with reward shaping, dual-stability, and critic weighting to improve fine-tuning stability and performance of flow models.

Findings

Achieves state-of-the-art text-to-image alignment on Stable Diffusion 3

Demonstrates robustness to unseen human preference models

Maintains generative quality and diversity during fine-tuning

Abstract

Flow-based generative models have shown remarkable success in text-to-image generation, yet fine-tuning them with intermediate feedback remains challenging, especially for continuous-time flow matching models. Most existing approaches solely learn from outcome rewards, struggling with the credit assignment problem. Alternative methods that attempt to learn a critic via direct regression on cumulative rewards often face training instabilities and model collapse in online settings. We present AC-Flow, a robust actor-critic framework that addresses these challenges through three key innovations: (1) reward shaping that provides well-normalized learning signals to enable stable intermediate value learning and gradient control, (2) a novel dual-stability mechanism that combines advantage clipping to prevent destructive policy updates with a warm-up phase that allows the critic to mature before influencing the actor, and (3) a scalable generalized critic weighting scheme that extends traditional reward-weighted methods while preserving model diversity through Wasserstein regularization. Through extensive experiments on Stable Diffusion 3, we demonstrate that AC-Flow achieves state-of-the-art performance in text-to-image alignment tasks and generalization to unseen human preference models. Our results demonstrate that even with a computationally efficient critic model, we can robustly finetune flow models without compromising generative quality, diversity, or stability.