← Back to Glossary
Faithfulness & Groundedness
Evaluation metrics checking whether every claim made in an LLM output is strictly supported by retrieved context chunks.
Why it exists
Ensures the generator doesn't inject outside hallucinations or unbacked facts into its answer.
How it works
Evaluator LLMs (like RAGAS) break response text into individual statements and verify if each statement is logically entailed by context chunks.
Related terms
Where you'll use this
roadmap