Scale Alone Does not Improve Mechanistic Interpretability in Vision Models
This study investigates whether the recent trend of scaling up neural networks in terms of dataset and model size has improved our understanding of their internal workings, specifically in the…
Continue reading