Saved in:
Bibliographic Details
Main Authors: Huang, Yushi, Zhou, Xiangxin, Wang, Ruoyu, Zhang, Chi, Zhang, Jun, Pang, Tianyu
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2605.26108
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916060817522688
author Huang, Yushi
Zhou, Xiangxin
Wang, Ruoyu
Zhang, Chi
Zhang, Jun
Pang, Tianyu
author_facet Huang, Yushi
Zhou, Xiangxin
Wang, Ruoyu
Zhang, Chi
Zhang, Jun
Pang, Tianyu
contents Recent advances in few-step diffusion distillation have enabled efficient image generation, yet aligning these models with human preferences remains challenging. We propose Reward-Tilted Distribution Matching Distillation (RTDMD), a two-stage framework that unifies distribution matching distillation with reward-guided reinforcement learning for few-step flow generators. We show that minimizing the KL divergence to a reward-tilted teacher distribution naturally decomposes into a distribution matching term and a reward maximization term. In the first stage, we introduce Ambient-Consistent Distribution Matching Distillation (AC-DMD), which performs subinterval-wise distribution matching and augments the fake score objective with a consistency regularizer to help the fake score model track the shifting generator distribution under limited updates. In the second stage, we jointly optimize both terms: for the reward maximization term, we derive a hybrid policy gradient that combines a GRPO-style estimator for the stochastic intermediate transitions with direct reward backpropagation through the deterministic final step, and further introduce step-subset GRPO (SubGRPO) to reduce variance. Experiments on SD3, SD3.5, and FLUX.2 demonstrate that RTDMD establishes new state-of-the-art results across preference, aesthetic, and compositional metrics with only 4 inference steps, outperforming previous few-step text-to-image generation methods. Code and models are available at https://github.com/Harahan/RTDMD.
format Preprint
id arxiv_https___arxiv_org_abs_2605_26108
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Reinforcing Few-step Generators via Reward-Tilted Distribution Matching
Huang, Yushi
Zhou, Xiangxin
Wang, Ruoyu
Zhang, Chi
Zhang, Jun
Pang, Tianyu
Computer Vision and Pattern Recognition
Recent advances in few-step diffusion distillation have enabled efficient image generation, yet aligning these models with human preferences remains challenging. We propose Reward-Tilted Distribution Matching Distillation (RTDMD), a two-stage framework that unifies distribution matching distillation with reward-guided reinforcement learning for few-step flow generators. We show that minimizing the KL divergence to a reward-tilted teacher distribution naturally decomposes into a distribution matching term and a reward maximization term. In the first stage, we introduce Ambient-Consistent Distribution Matching Distillation (AC-DMD), which performs subinterval-wise distribution matching and augments the fake score objective with a consistency regularizer to help the fake score model track the shifting generator distribution under limited updates. In the second stage, we jointly optimize both terms: for the reward maximization term, we derive a hybrid policy gradient that combines a GRPO-style estimator for the stochastic intermediate transitions with direct reward backpropagation through the deterministic final step, and further introduce step-subset GRPO (SubGRPO) to reduce variance. Experiments on SD3, SD3.5, and FLUX.2 demonstrate that RTDMD establishes new state-of-the-art results across preference, aesthetic, and compositional metrics with only 4 inference steps, outperforming previous few-step text-to-image generation methods. Code and models are available at https://github.com/Harahan/RTDMD.
title Reinforcing Few-step Generators via Reward-Tilted Distribution Matching
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2605.26108