retrojustification in lm sim is underrated, you can just write any vague hints and there's a path where the lm will justify it later in a way you hadn't expected
Posters: Shoalst0ne. Shoalst0ne
3 likes recorded by the source corpus.
Topics: AI & LLMs
View Post: https://x.com/Shoalst0ne/status/1831387379485307127