Bilinear MLPs enable weight-based mechanistic interpretability

Add code
Oct 10, 2024

Share this with someone who'll enjoy it:

View paper onarxiv icon

Share this with someone who'll enjoy it: