Fetching the paper…

Scale Alone Does not Improve Mechanistic Interpretability in Vision Models · Around