Y
Hacker News
new
|
ask
|
show
|
jobs
by
astrange
69 days 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.