SAGE: An Agentic Explainer Framework for Interpreting SAE Features in Language Models.

Jiaojiao Han, Wujiang Xu, Mingyu Jin, Mengnan Du

Browse the full EACL paper archive.