Score Distillation of Flow Matching Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhou, Mingyuan, Gu, Yi, Zheng, Huangjie, Song, Liangchen, He, Guande, Zhang, Yizhe, Hu, Wenze, Yang, Yinfei
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908689685807104
author Zhou, Mingyuan
Gu, Yi
Zheng, Huangjie
Song, Liangchen
He, Guande
Zhang, Yizhe
Hu, Wenze
Yang, Yinfei
author_facet Zhou, Mingyuan
Gu, Yi
Zheng, Huangjie
Song, Liangchen
He, Guande
Zhang, Yizhe
Hu, Wenze
Yang, Yinfei
contents Diffusion models achieve high-quality image generation but are limited by slow iterative sampling. Distillation methods alleviate this by enabling one- or few-step generation. Flow matching, originally introduced as a distinct framework, has since been shown to be theoretically equivalent to diffusion under Gaussian assumptions, raising the question of whether distillation techniques such as score distillation transfer directly. We provide a simple derivation -- based on Bayes' rule and conditional expectations -- that unifies Gaussian diffusion and flow matching without relying on ODE/SDE formulations. Building on this view, we extend Score identity Distillation (SiD) to pretrained text-to-image flow-matching models, including SANA, SD3-Medium, SD3.5-Medium/Large, and FLUX.1-dev, all with DiT backbones. Experiments show that, with only modest flow-matching- and DiT-specific adjustments, SiD works out of the box across these models, in both data-free and data-aided settings, without requiring teacher finetuning or architectural changes. This provides the first systematic evidence that score distillation applies broadly to text-to-image flow matching models, resolving prior concerns about stability and soundness and unifying acceleration techniques across diffusion- and flow-based generators. A project page is available at https://yigu1008.github.io/SiD-DiT.
format Preprint
id arxiv_https___arxiv_org_abs_2509_25127
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Score Distillation of Flow Matching Models
Zhou, Mingyuan
Gu, Yi
Zheng, Huangjie
Song, Liangchen
He, Guande
Zhang, Yizhe
Hu, Wenze
Yang, Yinfei
Computer Vision and Pattern Recognition
Artificial Intelligence
Machine Learning
Diffusion models achieve high-quality image generation but are limited by slow iterative sampling. Distillation methods alleviate this by enabling one- or few-step generation. Flow matching, originally introduced as a distinct framework, has since been shown to be theoretically equivalent to diffusion under Gaussian assumptions, raising the question of whether distillation techniques such as score distillation transfer directly. We provide a simple derivation -- based on Bayes' rule and conditional expectations -- that unifies Gaussian diffusion and flow matching without relying on ODE/SDE formulations. Building on this view, we extend Score identity Distillation (SiD) to pretrained text-to-image flow-matching models, including SANA, SD3-Medium, SD3.5-Medium/Large, and FLUX.1-dev, all with DiT backbones. Experiments show that, with only modest flow-matching- and DiT-specific adjustments, SiD works out of the box across these models, in both data-free and data-aided settings, without requiring teacher finetuning or architectural changes. This provides the first systematic evidence that score distillation applies broadly to text-to-image flow matching models, resolving prior concerns about stability and soundness and unifying acceleration techniques across diffusion- and flow-based generators. A project page is available at https://yigu1008.github.io/SiD-DiT.
title Score Distillation of Flow Matching Models
topic Computer Vision and Pattern Recognition
Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2509.25127