A Few-Step Generative Model on Cumulative Flow Maps
Open Access
- 3 July 2026
- journal article
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Graphics
- Vol. 45 (4) , 1-13
Abstract
We propose a unified, few-step generative modeling framework based on 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 cumulative parameterization that applies broadly to existing diffusion- and flow-based models without being tied to a specific prediction 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.Keywords
All Related Versions
Funding Information
- National Science Foundation (IIS #2433322)
- National Science Foundation (ECCS #2318814)
- National Science Foundation (CAREER #2433307)
- National Science Foundation (IIS #2106733)
- National Science Foundation (OISE #2433313)
- National Science Foundation (CNS #1919647)
This publication has 39 references indexed in Scilit:
- Functional Diffusion2024
- On Distillation of Guided Diffusion Models2023
- High-Resolution Image Synthesis with Latent Diffusion Models2022
- Diffusion Probabilistic Models for 3D Point Cloud Generation2021
- Neural Geometric Level of Detail: Real-time Rendering with Implicit 3D Shapes2021
- Learning Gradient Fields for Shape Generation2020
- Predicting Animation Skeletons for 3D Articulated Models via Volumetric Nets2019
- DeepSDF: Learning Continuous Signed Distance Functions for Shape Representation2019
- Occupancy Networks: Learning 3D Reconstruction in Function Space2019
- Multiscale structural similarity for image quality assessment2004