Hacker News new | past | comments | ask | show | jobs | submit login

Mechanistic interpretability is just getting started, but there's been some interesting research on induction heads. Here's an explanation:

https://www.lesswrong.com/posts/TvrfY4c9eaGLeyDkE/induction-...




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

Search: