HN
Hacker News
Top
New
Best
Ask
Show
Jobs
Comment by guywithabowtie | Hacker News Reader
Parent
Full thread
guywithabowtie
·
Does it need RLHF to make sense?
View on HN
hkab
·
No, RLHF only helps the model to align to human preference
theemathas
·
RLHF makes it possible to get good results with a bad prompt.
Reply on news.ycombinator.com