Inducing self-other overlap with SFT reduces deception at scale, but generaliza

(lesswrong.com)

1 points | by joozio 11 hours ago

No comments yet.