Mean Flows for One-step Generative Modeling

TL;DR

This paper introduces the MeanFlow model, a one-step generative approach using average velocity for flow fields, achieving state-of-the-art results on ImageNet 256x256 without pre-training.

Contribution

It presents a novel framework based on average velocity, providing a self-contained method that outperforms existing one-step models and narrows the gap with multi-step approaches.

Findings

Achieves an FID of 3.43 on ImageNet 256x256 with 1-NFE.

Requires no pre-training, distillation, or curriculum learning.

Significantly outperforms previous one-step diffusion/flow models.

Abstract

We propose a principled and effective framework for one-step generative modeling. We introduce the notion of average velocity to characterize flow fields, in contrast to instantaneous velocity modeled by Flow Matching methods. A well-defined identity between average and instantaneous velocities is derived and used to guide neural network training. Our method, termed the MeanFlow model, is self-contained and requires no pre-training, distillation, or curriculum learning. MeanFlow demonstrates strong empirical performance: it achieves an FID of 3.43 with a single function evaluation (1-NFE) on ImageNet 256x256 trained from scratch, significantly outperforming previous state-of-the-art one-step diffusion/flow models. Our study substantially narrows the gap between one-step diffusion/flow models and their multi-step predecessors, and we hope it will motivate future research to revisit the foundations of these powerful models.