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.
| Comments: | 11 pages, 12 figures |
| Subjects: | Machine Learning (cs.LG); Graphics (cs.GR) |
| Cite as: | arXiv:2605.03623 [cs.LG] |
| (or arXiv:2605.03623v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2605.03623 arXiv-issued DOI via DataCite (pending registration) |
|
| Related DOI: | https://doi.org/10.1145/3811380
DOI(s) linking to related resources |
From: Zhiqi Li [view email]
[v1]
Tue, 5 May 2026 10:51:40 UTC (23,338 KB)
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。