A Few-Step Generative Model on Cumulative Flow Maps

TL;DR

This paper introduces a unified generative modeling framework using cumulative flow maps that enables efficient, few-step or one-step generation across various tasks without increasing model complexity.

Contribution

It presents a novel cumulative-flow abstraction that connects local updates with finite-time transport, applicable to diffusion and flow-based models, supporting minimal-step generation.

Findings

Effective across diverse tasks including image and shape generation.

Supports few-step and one-step generation with minimal changes.

Reduces inference cost while maintaining quality.

Abstract

We propose a unified, few-step generative modeling framework based on \emph{cumulative flow maps} for long-range transport in probability space, inspired by flow-map techniques for physical transport and dynamics. At its core is a cumulative-flow abstraction that connects local, instantaneous updates with finite-time transport, enabling generative models to reason about global state transitions. This perspective yields a unified few-step framework built on cumulative transport and \revise{cumulative} parameterization that applies broadly to existing diffusion- and flow-based models without being tied to a specific prediction \revise{instantiation}. Our formulation supports few-step and even one-step generation while preserving synthesis quality, requiring only minimal changes to time embeddings and training objectives, and no increase in model capacity. We demonstrate its effectiveness across diverse tasks, including image generation, geometric distribution modeling, joint prediction, and SDF generation, with reduced inference cost.