Enhancing Automated Interpretability with Output-Centric Feature Descriptions Paper • 2501.08319 • Published 13 days ago • 10