[论文] Mean Flow Distillation: Robust and Stable Distillation for Flow Matchi...

论文概要

研究领域: CV
作者: An Zhao, Shengyuan Zhang, Zhongjian Sun, Yixiang Zhou, Zejian Li, Ling Yang, Tianrun Chen, Lingyun Sun
发布时间: 2026-06-09
arXiv: 2606.11155

中文摘要

Mean Flow Distillation（MFD）是专为流匹配模型设计的蒸馏框架。理论上证明MFD作为时间低通滤波器，有效抑制VSD固有的高频优化噪声，同时确保全局轨迹一致性。证明均值流匹配定理：匹配期望平均速度足以实现严格分布对齐。在4D占用预测和文生图等高维流形任务上，MFD实现SOTA性能，支持高保真单步生成。

原文摘要

Flow Matching models have demonstrated strong performance across a wide range of generative tasks. However, their reliance on ODE-based iterative sampling incurs substantial computational overhead in inference, which limits their applicability in real-time scenes. While distillation is a promising solution, existing approaches largely borrow from diffusion-based score matching, often failing to exploit the intrinsic geometric structure of flows and suffering from training instability, high variance, and degraded generation quality. In this paper, we propose Mean Flow Distillation (MFD), a novel distillation framework tailored for flow matching models. We theoretically demonstrate that MFD acts as a temporal低通滤波器, effectively suppressing the high-frequency optimization noise inherent in var...

自动采集于 2026-06-11

#论文 #arXiv #CV #小凯