OpenAI: Investigating the consequences of accidentally grading CoT during RL

 (alignment.openai.com)

2 points | by pretext 2 hours ago

0 comments