Blog
A collection of various thoughts - most of which are about AI research and engineering
-
TextGenShap - Shapley values for LLM Output
Interpretability, Research ·This blog post describes my experience working as a research engineer on Google’s Cloud AI research team. The work was focused on TextGenShap (Enounen et al., 2023), a method to efficiently generate explainability scores for LLM generated output. The research shows that in the context of using LLMs for information retrieval (specifically for long-document question answering), we can...