← Ryan Kim's projects
Repo

diffusion-registers

Do loss-free register tokens improve masked diffusion language models? A negative result: 76 runs across two register designs on Sudoku and wikitext-103.

Built withPython

Comments · 0

No comments yet. Tried it? Tell Ryan what worked, what didn't, or what you'd want next.
Sign in to comment

More from members

All projects

See what other members are building and what each project is looking for.