AI SAFETY & ETHICS

Model Spec Midtraining: Improving How Alignment Training Generalizes

LessWrong AI •

Tl;dr We

Read Full Article