Yoav Gur-Arieh Roy Mayan Chen Agassy Atticus Geiger Mor Geva Enhancing Automated Interpretability with Output-Centric Feature Descriptions. 2025 January abs/2501.08319 CoRR https://doi.org/10.48550/arXiv.2501.08319 db/journals/corr/corr2501.html#abs-2501-08319 streams/journals/corr