Mean Reciprocal Rank (MRR)
MRR measures how high up the first correct result appears in a ranked list, averaged over many queries. If the right answer is usually near the top, MRR is close to 1; if it's buried, MRR drops. It's a core retrieval and search metric.
Also known as: MRR
RAG & RetrievalAI Evaluation & Reliability
Mean Reciprocal Rank evaluates ranked results by one question: how far down is the first correct hit? For each query you take the reciprocal of that rank (1 if it’s first, 1/2 if second, 1/3 if third), then average across all queries. A high MRR means users — or an agent — usually find the right thing without scrolling.
It matters for RAG and search because retrieval quality caps everything downstream: if the relevant chunk ranks fifth and you only feed the top three to the model, the answer is built on the wrong context. MRR’s limit is that it only cares about the first correct result, so when multiple relevant results matter, metrics like NDCG or recall@k complement it.