Slacker News Slacker News logo featuring a lazy sloth with a folded newspaper hat
  • top
  • new
  • show
  • ask
  • jobs
Library
← Back to context

Comment by vman512

1 day ago

Sounds right. The policy for rejection can depend on what you want - you might accept the top K highest probability tokens or top P probability mass. Or you can do something like importance sampling and probabilistically reject based on the ratio of likelihoods

0 comments

vman512

Reply

No comments yet

Contribute on Hacker News ↗

Slacker News

Product

  • API Reference
  • Hacker News RSS
  • Source on GitHub

Community

  • Support Ukraine
  • Equal Justice Initiative
  • GiveWell Charities