▲
Hacker News
new
|
best
|
ask
|
show
|
jobs
points by
steve1977
2 days ago | 0 comments
I'm pretty sure the typical "Claudisms" are influenced by RLHF as well, so if it can have that effect, why not these...
[–]
twoodfin
2 days ago
Right: The humans are training the models on much more than the particulars of the tasks being trained.
Hacker News
[–] twoodfin 2 days ago