Faithfulness measures whether every claim in a generated response can be factually inferred from the retrieved context. It evaluates grounding rather than general correctness, helping detect hallucinated information in a RAG pipeline. section 6.5.1 section 6.5.2
Use faithfulness to inspect the generation component after evaluating retrieval quality:
- Check whether the response remains within the information supplied by the retrieved documents. section 6.5.1
- Treat unsupported claims as evidence that the response is not faithful, even when the answer appears relevant to the query. section 6.5.1
- Use the metric alongside context precision, context recall, and answer relevancy for broader RAG evaluation. section 6.5.2
Figure 6.7 RAGAS Core Metric extensively covers LLM Grounding and Response (Source)
Use the grounding-and-response figure to place faithfulness within the RAGAS evaluation view. section 6.5.2










