mechanistic interpretability, sparse autoencoders, vision transformers, representation learning
No public activity