Hacker News new | ask | show | jobs
Speculative sampling: LLMs writing a lot faster using smaller LLMs (blog.dust.tt)
4 points by spolu 1103 days ago