Definition
A ranking metric used to evaluate RAG retrieval performance by calculating the average of the reciprocal ranks of the first relevant document for a set of queries. It quantifies how effectively a system places the single most relevant chunk at the top of the search results, directly impacting the quality of the LLM's context window.
Measures only the rank of the first relevant result, unlike nDCG which considers the relevance of the entire list.
"A searchlight that loses half its brightness for every meter it travels; if the target is at 1 meter, visibility is 1 (1/1), but if it's at 2 meters, visibility drops to 0.5 (1/2)."
- nDCG (Normalized Discounted Cumulative Gain)(Alternative Metric)
- Precision@K(Component)
- Vector Database Retrieval(Prerequisite)
- Hit Rate(Related Metric)
Conceptual Overview
A ranking metric used to evaluate RAG retrieval performance by calculating the average of the reciprocal ranks of the first relevant document for a set of queries. It quantifies how effectively a system places the single most relevant chunk at the top of the search results, directly impacting the quality of the LLM's context window.
Disambiguation
Measures only the rank of the first relevant result, unlike nDCG which considers the relevance of the entire list.
Visual Analog
A searchlight that loses half its brightness for every meter it travels; if the target is at 1 meter, visibility is 1 (1/1), but if it's at 2 meters, visibility drops to 0.5 (1/2).