How LLMs Learn: Tracing Internal Representations with Sparse Autoencoders

Re-assign community