Everyone talked about the Apollo eval of o1, no one seemed to talk about the METR eval. So far I'm finding the METR finding more interesting, given that they say 70% of failures are still spurious…

Posters: TheZvi. TheZvi

2 likes recorded by the source corpus.

Topics: AI & LLMs

View Post: https://x.com/TheZvi/status/1866534968362668485