Modular MeanFlow: Towards Stable and Scalable One-Step Generative Modeling

TL;DR

Modular MeanFlow (MMF) is a novel, theoretically grounded approach for one-step generative modeling that offers stable training, high-quality sample generation, and strong generalization, unifying and extending existing methods.

Contribution

We introduce Modular MeanFlow, a flexible framework for learning time-averaged velocity fields that unifies and generalizes existing methods while avoiding costly derivatives.

Findings

Achieves competitive sample quality in image synthesis and trajectory modeling.

Demonstrates robust convergence and generalization, especially with limited or out-of-distribution data.

Provides a stable training mechanism with a curriculum warmup schedule.

Abstract

One-step generative modeling seeks to generate high-quality data samples in a single function evaluation, significantly improving efficiency over traditional diffusion or flow-based models. In this work, we introduce Modular MeanFlow (MMF), a flexible and theoretically grounded approach for learning time-averaged velocity fields. Our method derives a family of loss functions based on a differential identity linking instantaneous and average velocities, and incorporates a gradient modulation mechanism that enables stable training without sacrificing expressiveness. We further propose a curriculum-style warmup schedule to smoothly transition from coarse supervision to fully differentiable training. The MMF formulation unifies and generalizes existing consistency-based and flow-matching methods, while avoiding expensive higher-order derivatives. Empirical results across image synthesis and trajectory modeling tasks demonstrate that MMF achieves competitive sample quality, robust convergence, and strong generalization, particularly under low-data or out-of-distribution settings.