Announcement_5

New preprint about extracting hidden knowledge from LLMs with mechanistic interpretability is out!