@MatthewJBar If your training process *directly trains the model to lie* in various ways it will lie. e.g. it's common practice to train RLHF models out of saying "I don't know", so they basically…

Posters: jd_pressman. jd_pressman

2 likes recorded by the source corpus.

Topics: AI & LLMs

View Post: https://x.com/jd_pressman/status/1726241282224496768