One-step Diffusion Models with Bregman Density Ratio Matching
By: Yuanzhi Zhu , Eleftherios Tsonis , Lucas Degeorge and more
Potential Business Impact:
Makes AI create pictures much faster.
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.
Similar Papers
Distillation of Discrete Diffusion by Exact Conditional Distribution Matching
Machine Learning (CS)
Makes AI create pictures much faster.
Revisiting Diffusion Models: From Generative Pre-training to One-Step Generation
Machine Learning (CS)
Makes AI create pictures much faster.
Learning Few-Step Diffusion Models by Trajectory Distribution Matching
CV and Pattern Recognition
Makes AI art and video creation much faster.