media creative tools
A Discriminator Replaced the Student's Extra Diffusion Model
DMAD matched distributions through log-density ratios and reached one-step ImageNet FID 1.04.
Summary
DMAD matched distributions through log-density ratios and reached one-step ImageNet FID 1.04.
DMAD learns the density-ratio signal for few-step visual distillation with two discriminator heads instead of maintaining an auxiliary student-score model. The authors report leading compared image, video and audio-video results, including one-step ImageNet-64 FID 1.04.
Why it matters
DMAD matched distributions through log-density ratios and reached one-step ImageNet FID 1.04.
Limits and context
No additional limitation was separately recorded.
Key claims
DMAD matched distributions through log-density ratios and reached one-step ImageNet FID 1.04.
Evidence: source-2026-10-02-019
Sources
- arXiv preprint 2610.02188arXiv · primary research
Corrections
No corrections have been recorded for this story.