Jialin Zhu, Xing Liu, Feixiang He, and He Wang theoretically analyzed the relationship between Distribution Matching Distillation (DMD) and Drifting Models.
Distribution matching distillation and drifting models are equivalent under finite-sample analysis
The authors show that training a Drifting Model is naturally equivalent to Distribution Matching Distillation, and propose Multi-Bandwidth DMD for improved one-step generation.
Chinese Tech
Jialin Zhu · Xing Liu · Feixiang He · He Wang
Baidu Inc. · Central South University · University College London
Research Digest··2 min read
The authors prove an exact finite-sample correspondence between Distribution Matching Distillation (DMD) and Drifting Models, two approaches for one-step generative modeling.
Why this paper
From Baidu Inc. and 2 others
In one line
Distribution Matching Distillation is equivalent to Drifting Models, and multi-bandwidth training improves DMD convergence and quality.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ·No stated limitations found
- ·No benchmark numbers found
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§