26 July 2024
DeepMind makes big jump toward interpreting LLMs with sparse autoencoders - 2024-07-26 15:04:18Z
Title:DeepMind makes big jump toward interpreting LLMs with sparse autoencoders Summary: A new research by Google DeepMind shows how sparse autoencoders (SAEs) with special JumpReLU activation can help interpret LLMs. Link: DeepMind makes big jump toward interpreting LLMs with sparse autoencoders
