Recall@K
A core evaluation metric for information retrieval systems that measures the proportion of truly relevant documents captured within the top-K returned results. It is independent of answer generation quality assessment and is commonly used to evaluate the retrieval layer in Retrieval-Augmented Generation (RAG) systems.
Core Facts
Timeline (last 90 days)
The value of K should be determined in conjunction with the actual context window of the downstream generation stage; for example, if only the top 5 chunks are concatenated during generation, then evaluating Recall@5 is more meaningful than Recall@20.
The Recall@k metric is defined as the proportion of the top k results that contain known relevant documents.
Recall@K indicates what proportion of truly relevant documents were successfully found among the top K results returned by the system.
Recall@K measures whether relevant documents appear in the top K results, while MRR measures the ranking position of the first relevant result.
Set-based retrieval metrics correspond to the classical concepts of Precision@K and Recall@K, measuring the relevance and recall ratio of retrieved document sets.
The value of K in Top-K should be determined through offline evaluation by measuring Recall@K at different K values on an annotated test set and finding the inflection point where the recall curve begins to flatten.
All Facts (6)
The Recall@k metric is defined as the proportion of the top k results that contain known relevant documents.
65%UnverifiedThe value of K should be determined in conjunction with the actual context window of the downstream generation stage; for example, if only the top 5 chunks are concatenated during generation, then evaluating Recall@5 is more meaningful than Recall@20.
50%UnverifiedRecall@K indicates what proportion of truly relevant documents were successfully found among the top K results returned by the system.
50%UnverifiedRecall@K measures whether relevant documents appear in the top K results, while MRR measures the ranking position of the first relevant result.
50%UnverifiedSet-based retrieval metrics correspond to the classical concepts of Precision@K and Recall@K, measuring the relevance and recall ratio of retrieved document sets.
50%UnverifiedThe value of K in Top-K should be determined through offline evaluation by measuring Recall@K at different K values on an annotated test set and finding the inflection point where the recall curve begins to flatten.
50%