Uncovering Conceptual Blindspots in Generative Image Models Using Sparse Autoencoders

Open in new window