SmartFAQs.ai
Back to Learn
Intermediate

MRR

Mean Reciprocal Rank (MRR) is a retrieval evaluation metric that calculates the average of the reciprocal ranks of the first relevant document across a set of queries. In RAG pipelines, it measures the retriever's effectiveness at placing the most pertinent ground-truth context at the very top of the result list, where a rank of 1 yields a score of 1.0 and a rank of 10 yields 0.1.

Definition

Mean Reciprocal Rank (MRR) is a retrieval evaluation metric that calculates the average of the reciprocal ranks of the first relevant document across a set of queries. In RAG pipelines, it measures the retriever's effectiveness at placing the most pertinent ground-truth context at the very top of the result list, where a rank of 1 yields a score of 1.0 and a rank of 10 yields 0.1.

Disambiguation

A search relevance metric for RAG evaluation, not a SaaS business metric for revenue.

Visual Metaphor

"A target where only your best-placed arrow counts; a bullseye is a perfect score, but the value drops sharply the further that single best arrow is from the center."

Key Tools
RagasTruLensDeepEvalPyTrec_Evalscikit-learn
Related Connections

Conceptual Overview

Mean Reciprocal Rank (MRR) is a retrieval evaluation metric that calculates the average of the reciprocal ranks of the first relevant document across a set of queries. In RAG pipelines, it measures the retriever's effectiveness at placing the most pertinent ground-truth context at the very top of the result list, where a rank of 1 yields a score of 1.0 and a rank of 10 yields 0.1.

Disambiguation

A search relevance metric for RAG evaluation, not a SaaS business metric for revenue.

Visual Analog

A target where only your best-placed arrow counts; a bullseye is a perfect score, but the value drops sharply the further that single best arrow is from the center.

Related Articles