The authors introduce CDMD, a simulation-free and data-free distillation framework for diffusion and flow models.
Data-free distillation of diffusion models with a single consistent objective
CDMD unifies generation and score estimation, achieving state-of-the-art few-step generation on ImageNet without external data or teacher simulations.
Top University
Yuxiang Fu · Qi Yan · Zike Wu · Yongxing Zhang · Purang Abolmaesumi · Lele Wang · +1 more
University of British Columbia · Vector Institute for AI · Canada CIFAR AI Chair
Research Digest··2 min read
The authors propose Consistent Distribution Matching Distillation (CDMD), a method that uses only a frozen teacher and a trainable student with one objective, avoiding distillation datasets, teacher rollouts, and auxiliary networks.
Why this paper
From University of British Columbia and 2 others
In one line
A single objective unifying sample generation and score estimation lets a student distill a frozen diffusion teacher data-free, with Wasserstein convergence and ImageNet FID 2.04 at one step.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ✓Limitations stated by the authors
- ✓Reports numbers on named benchmarks
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§