Hacker News
new
past
comments
ask
show
jobs
points
by
tarsinge
19 hours ago
|
comments
by
astrange
10 hours ago
|
[-]
No, that's how base model pretraining works. Claude's behavior is more based on its constitution and RLVR feedback, because that's the most recent thing that happened to it.
reply