faithfulness
An eval metric for RAG: the share of claims in an answer that the retrieved context supports.
In more detail
Ragas computes it by splitting the answer into claims and checking each against the context. It measures the generator alone: an answer can be faithful to the wrong passages, which context precision and recall catch.
Related terms
- context precision — An eval metric for RAG retrieval: whether the relevant chunks were ranked above the irrelevant ones in what was retrieved.
- context recall — An eval metric for RAG retrieval: the share of the facts a reference answer needs that appear in the retrieved context.
- grounding — Making a model answer only from supplied sources, cite them, and refuse when they do not contain the answer.
- Ragas — An open-source library of evaluation metrics for RAG pipelines and agents, such as faithfulness, context precision, context recall and tool call accuracy.
Taught in
- Grounded Answers With Citations — AI Engineering: LLM Apps, RAG & Agents
- Model-Graded Evals — AI Engineering: LLM Apps, RAG & Agents
- Evaluating RAG And Agents — AI Engineering: LLM Apps, RAG & Agents