Building a RAG pipeline is easy; making it robust is hard. How do you know if your retrieved context is relevant, or if the LLM is hallucinating based on good context?
Module 1: The Triad of Metrics
Core Evaluation Metrics
- Context Relevance: Does the retrieved document contain the answer?
- Faithfulness (Groundedness): Is the answer strictly derived from the context?
- Answer Relevance: Does the generated answer address the user's original query?