recall@5
A retrieval evaluation metric that measures whether the top five retrieved results include the relevant document, and is sensitive to the baseline capability of a retrieval system.
Timeline (last 90 days)
recall@5 is defined as: for each query, check whether the top 5 retrieval results contain at least one relevant document (or the proportion of all relevant documents that were recalled), then average over all queries.
CodebaseRAG's CI gate measures average recall@5, MRR and nDCG@5 on the golden set, and the build fails when average recall@5 falls below the committed baseline of 0.409.
The gate uses only recall@5 as a hard threshold, with MRR and nDCG@5 as observational metrics, to avoid ranking fluctuations causing unstable builds.
recall@5 measures whether the top 5 chunks returned by retrieval include the golden chunk that can answer the question, and it directly reflects whether the chunk boundaries are correct.
When recall@5 on adversarial questions is low, the correct action is to go back and check the chunking strategy; only after recall@5 meets the bar is it meaningful to invest in optimizing the embedding model or reranker.
recall@5 refers to whether the correct answer is contained in the top 5 results returned by retrieval, and strict mode requires an exact match rather than partial relevance.
All Facts (6)
recall@5 is defined as: for each query, check whether the top 5 retrieval results contain at least one relevant document (or the proportion of all relevant documents that were recalled), then average over all queries.
50%UnverifiedCodebaseRAG's CI gate measures average recall@5, MRR and nDCG@5 on the golden set, and the build fails when average recall@5 falls below the committed baseline of 0.409.
50%UnverifiedThe gate uses only recall@5 as a hard threshold, with MRR and nDCG@5 as observational metrics, to avoid ranking fluctuations causing unstable builds.
50%Unverifiedrecall@5 measures whether the top 5 chunks returned by retrieval include the golden chunk that can answer the question, and it directly reflects whether the chunk boundaries are correct.
50%UnverifiedWhen recall@5 on adversarial questions is low, the correct action is to go back and check the chunking strategy; only after recall@5 meets the bar is it meaningful to invest in optimizing the embedding model or reranker.
50%Unverifiedrecall@5 refers to whether the correct answer is contained in the top 5 results returned by retrieval, and strict mode requires an exact match rather than partial relevance.
50%