PaperLens
紙
Home
Close reading
Textbook
Learn
More
My Page
All-access
Sign in with Google
Mode
Students
Professional
Language
JA
EN
Theme
◐
Sign out
New
Lab
Landscape
Contributors
Glossary
Home
Close reading
Textbook
Learn
More
#dmd
1 articles
01
2026-09-11
·
★ MEMBER
·
PAPER
·
12 min read
Paper Deep Dive: Mask Forcing — Fixing Autoregressive Video Diffusion Distillation with a Dual-Noise Mask
Distilling a bidirectional video diffusion teacher into a real-time autoregressive student tends to produce over-saturated, over-smoothed video. Mask Forcing eases this mode collapse by randomly swapping some tokens in the student's own rollout for slightly cleaner ones — no real data, no extra post-training stage.