End-to-end RAG evaluation

Evaluate the pipeline the user experiences: was the answer faithful to the retrieved context, and did it address the question.

What this lesson covers

• Score faithfulness to the retrieved context
• Score relevance to the original question
• Track both across releases to catch regressions

This lesson sits in Grounded Generation, part of Retrieval-Augmented Generation from Scratch. It assumes what came before it and leads directly into the next lesson in the module.