Valérie Costa
Project Summary: Sparse autoencoders are widely used in computer science to enhance the interpretability of large language models (LLMs). In this project, we aim to investigate their potential for understanding neural computation. We are using a recently...