Our paper “Toward Faithful Retrieval-Augmented Generation with Sparse Autoencoders” has been accepted to ICLR2026!
