x
Mechanistic interpretability through clustering — AI Alignment Forum