SubFlow: Sub-mode Conditioned Flow Matching for Diverse One-Step Generation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Lin, Yexiong, Shi, Jia, Ye, Shanshan, Wang, Wanyu, Yao, Yu, Liu, Tongliang
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913029328732160
author Lin, Yexiong
Shi, Jia
Ye, Shanshan
Wang, Wanyu
Yao, Yu
Liu, Tongliang
author_facet Lin, Yexiong
Shi, Jia
Ye, Shanshan
Wang, Wanyu
Yao, Yu
Liu, Tongliang
contents Flow matching has emerged as a powerful generative framework, with recent few-step methods achieving remarkable inference acceleration. However, we identify a critical yet overlooked limitation: these models suffer from severe diversity degradation, concentrating samples on dominant modes while neglecting rare but valid variations of the target distribution. We trace this degradation to averaging distortion: when trained with MSE objectives, class-conditional flows learn a frequency-weighted mean over intra-class sub-modes, causing the model to over-represent high-density modes while systematically neglecting low-density ones. To address this, we propose SubFlow, Sub-mode Conditioned Flow Matching, which eliminates averaging distortion by decomposing each class into fine-grained sub-modes via semantic clustering and conditioning the flow on sub-mode indices. Each conditioned sub-distribution is approximately unimodal, so the learned flow accurately targets individual modes with no averaging distortion, restoring full mode coverage in a single inference step. Crucially, SubFlow is entirely plug-and-play: it integrates seamlessly into existing one-step models such as MeanFlow and Shortcut Models without any architectural modifications. Extensive experiments on ImageNet-256 demonstrate that SubFlow yields substantial gains in generation diversity (Recall) while maintaining competitive image quality (FID), confirming its broad applicability across different one-step generation frameworks. Project page: https://yexionglin.github.io/subflow.
format Preprint
id arxiv_https___arxiv_org_abs_2604_12273
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle SubFlow: Sub-mode Conditioned Flow Matching for Diverse One-Step Generation
Lin, Yexiong
Shi, Jia
Ye, Shanshan
Wang, Wanyu
Yao, Yu
Liu, Tongliang
Machine Learning
Computer Vision and Pattern Recognition
Flow matching has emerged as a powerful generative framework, with recent few-step methods achieving remarkable inference acceleration. However, we identify a critical yet overlooked limitation: these models suffer from severe diversity degradation, concentrating samples on dominant modes while neglecting rare but valid variations of the target distribution. We trace this degradation to averaging distortion: when trained with MSE objectives, class-conditional flows learn a frequency-weighted mean over intra-class sub-modes, causing the model to over-represent high-density modes while systematically neglecting low-density ones. To address this, we propose SubFlow, Sub-mode Conditioned Flow Matching, which eliminates averaging distortion by decomposing each class into fine-grained sub-modes via semantic clustering and conditioning the flow on sub-mode indices. Each conditioned sub-distribution is approximately unimodal, so the learned flow accurately targets individual modes with no averaging distortion, restoring full mode coverage in a single inference step. Crucially, SubFlow is entirely plug-and-play: it integrates seamlessly into existing one-step models such as MeanFlow and Shortcut Models without any architectural modifications. Extensive experiments on ImageNet-256 demonstrate that SubFlow yields substantial gains in generation diversity (Recall) while maintaining competitive image quality (FID), confirming its broad applicability across different one-step generation frameworks. Project page: https://yexionglin.github.io/subflow.
title SubFlow: Sub-mode Conditioned Flow Matching for Diverse One-Step Generation
topic Machine Learning
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2604.12273