arXiv：cs.LG· Jialin Zhu, Xing Liu, Feixiang He, He Wang·· 4 小时前
多带宽分布匹配蒸馏：论分布匹配蒸馏与 Drifting Models 的等价性
Multi-Bandwidth Distribution Matching Distillation: On the Equivalence of Distribution Matching Distillation and Drifting Models
AI 导读
该论文证明，将预训练 Diffusion & Flow Style Generative Models 的速度场/噪声场转换为 Drifting Models 的吸引力场、并从生成分布估计排斥力场后，训练 Drifting Model 与分布匹配蒸馏（DMD/DMD2）天然等价。
正文
Abstract:Researchers are exploring effective one-step generative model continuously, and, Drifting Models (Deng et al., 2026), demonstrate great potential in one-step generation recently. There are works that reveal the connection between Diffusion & Flow Style Generative Models (DFSGMs) (Ho et al., 2020; Song et al., 2020a;b; Lipman et al., 2022; Liu et al., 2022) and Drifting Models (Li & Zhu, 2026; Lai et al., 2026; Turan et al., 2026). But no one has yet established a precise correspondence between the Drifting Model and the widely used distillation method- Distribution Matching Distillation (DMD/DMD2) (Yin et al., 2024b;a) to the best of our knowledge, even though their optimization objective formulas are virtually identical. In this paper, we prove that by converting the velocity-field / noise-field from the pre-trained DFSGMs into the attraction force field in Drifting Models and estimating the repulsion force field from the generative distribution, training the Drifting Model is naturally equivalent to the Distribution Matching Distillation. With this equivalent concept, we propose an improved method based on DMD from the Drifting Model's perspective- Multi-Bandwidth Distribution Matching Distillation (MBDMD).
| Subjects: | Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV) |
| Cite as: | arXiv:2610.10989 [cs.LG] |
| (or arXiv:2610.10989v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2610.10989 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Jialin Zhu [view email]
[v1]
Wed, 7 Oct 2026 23:21:55 UTC (469 KB)
来源：arXiv：cs.LG · arxiv.org