Safety & Alignment

Prompt injection, jailbreaks, Constitutional AI, reward hacking and mechanistic interpretability.

25concepts
408flashcards
202minutes of reading

No beginner concepts in this track. Show all.