用数学统一框架加速扩散模型采样，一步生成质量更高。
One-step Diffusion Models with Bregman Density Ratio Matching
- 将扩散模型压缩为一步采样，基于贝格曼散度匹配密度比。
- 在CIFAR-10上比反KL蒸馏的FID降低18.3，视觉质量接近教师模型。
- 理论严谨且实用，适合追求高效生成的研究者与工程师。
扩散模型和流模型虽生成质量高，但因采样步骤多而计算成本高。现有蒸馏方法虽能加速生成，但多数目标缺乏统一理论基础。本文提出Di-Bregman，将扩散蒸馏建模为基于贝格曼散度的密度比匹配，提供一个凸分析视角，统一了多种已有目标。在CIFAR-10和文本到图像生成任务上的实验表明，Di-Bregman相比反KL蒸馏实现更低的一步采样FID（提升18.3），同时保持与教师模型相当的视觉保真度。结果表明，贝格曼密度比匹配是一种兼具理论基础与实际效果的高效一步扩散生成路径。
原文摘要 · Abstract (English)
Diffusion and flow models achieve high generative quality but remain computationally expensive due to slow multi-step sampling. Distillation methods accelerate them by training fast student generators, yet most existing objectives lack a unified theoretical foundation. In this work, we propose Di-Bregman, a compact framework that formulates diffusion distillation as Bregman divergence-based density-ratio matching. This convex-analytic view connects several existing objectives through a common lens. Experiments on CIFAR-10 and text-to-image generation demonstrate that Di-Bregman achieves improved one-step FID over reverse-KL distillation and maintains high visual fidelity compared to the teacher model. Our results highlight Bregman density-ratio matching as a practical and theoretically-grounded route toward efficient one-step diffusion generation.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。