Wait, why wouldn’t RLHF influence word choices? | Hacker News

Hacker Newsnew | past | comments | ask | show | jobs | submit

		skywhopper on Aug 12, 2024 \| parent \| context \| favorite \| on: Ask HN: Why does ChatGPT love the word "eager" so ... Wait, why wouldn’t RLHF influence word choices?

llm_nerd on Aug 12, 2024 [–]

I didn't say it wouldn't (or rather couldn't), I said it was unlikely for the selected hypothesis given standard training data vs RLHF iterations.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact