Y
Hacker News
new
|
ask
|
show
|
jobs
by
fridder
47 days ago
Yeah. I usually do this by telling it to be adversarial and find gaps and holes. Not fool proof but it does seem to increase the quality. It has helped when using local models in particular.
1 comments
SubiculumCode
47 days ago
Yeah, you have to shortcut the RL-trained people pleasing
link