Vision-and-Language Training Helps Deploy Taxonomic Knowledge but Does Not Fundamentally Alter It

Sparse Autoencoders Learn Monosemantic Features in Vision-Language Models