Y
Hacker News
new
|
ask
|
show
|
jobs
by
researchers
509 days ago
Tuning for qualitative outcomes is pretty much solved via RLHF/DPO (what this post calls "preference tuning"). Right?