Fetching the paper…

Sparse Autoencoders Learn Monosemantic Features in Vision-Language Models · Around