Enhancing Automated Interpretability with Output-Centric Feature Descriptions Paper • 2501.08319 • Published 2 days ago • 10