This was fairly obviously a sampling bug and the people claiming it was nonzero evidence about the stability of RLHF were telling on themselves ngl. https://t.co/PjXMy0PByL

Posters: jd_pressman. jd_pressman

7 likes recorded by the source corpus.

Topics: AI & LLMs, AI safety

View Post: https://x.com/jd_pressman/status/1760528881319346675