Our work on bilinear autoencoders got a spotlight at the Mechanistic interpretability workshop (NeurIPS 2025)!
