12

I gave a talk on moving from mechanistic to compositional interpretability at the Interpretable Deep Learning seminar!