@kindgracekind i'm thinking specifically of persona priming tokens at the beginning of a sequence. e.g. when you pin the emergent misalignment vector extremely negative you get:…

Posters: voooooogel. voooooogel

3 likes recorded by the source corpus.

Topics: AI & LLMs, AI safety

View Post: https://x.com/voooooogel/status/1997510601468453374