Large Language Model explainability is a hot topic in ML research. I will detail why LLM evaluation methods allow us to better understand their internal behavior, and how they work.