More capable models are better at in-context scheming

 (apolloresearch.ai)

2 points | by JumpCrisscross 5 hours ago

0 comments