leogao comments on Sparse Autoencoders Find Highly Interpretable Directions in Language Models

[leogao]

Awesome work! I like the autoencoder approach a lot.