I’m excited to share our recent paper on AI Control through Mechanistic Interpretability approaches. You can read the full paper here.