Interpreting and Steering LLMs with Mutual Information-based Explanations on Sparse Autoencoders

LLMSV
Re-assign community