Sparse Autoencoders Reveal Temporal Difference Learning in Large Language Models.

Can Demircan, Tankred Saanum, Akshay Kumar Jagadish, Marcel Binz, Eric Schulz

Browse the full ICLR paper archive.