Sort Metrics by Required Evidence

Sort each metric by the extra evidence it most depends on. Reference answer needed Retrieved context needed Human labels needed Correctness against an approved pricing answer Context recall against expected source evidence Faithfulness of answer claims to retrieved chunks Relevance of returned chunks to the user question Agreement between the response and retrieved documents Judge-human agreement on borderline contract cases Calibration of pass, fail, and partial-support examples Small expert-labeled anchor set for automated judge correction

Sign up free — one personalized lesson every day, matched to your role and goals.

Already have an account? Sign in