Faithfulness

Faithfulness is a RAG evaluation metric that checks whether every claim in a generated answer is supported by the retrieved context. A low faithfulness score means the model added information that was not in its sources, which is a common signal of hallucination.

How faithfulness is scored

Most tools split the answer into individual claims, then check each claim against the retrieved context, often using an LLM as the judge. The score is the share of claims the context supports, from 0 to 1. RAGAS and DeepEval both offer a faithfulness metric.

Faithfulness is not the same as correctness

An answer can be faithful to a wrong document, or correct but not supported by the context. Pair faithfulness with checks on retrieval quality and with a golden dataset of known answers.

Tip: agree a minimum score with your team and fail the CI build when answers drop below it.

Read the full guide: What Is RAGAS — Testing RAG Pipelines Explained for QA Engineers