i've been banging on the drum for a while that behavior evals like summitbridge need to (at a minimum) run over a wide set of prompts - this is a big part of why. one experiment from a new paper on…

Posters: voooooogel. voooooogel

2 likes recorded by the source corpus.

Topics: AI safety

View Post: https://x.com/voooooogel/status/2032525103850991823