Decoding Latent Reasoning: Mechanistic Interpretability of Chain of Continuous Thought with Sparse Autoencoders.

Shlok Sand, Chaitanya Shah, Vasudeva Varma

Browse the full AIED paper archive.