kandpal-2023-bm25-recall-range
IN premise — summaries/2026/08/24/kandpal-2023-long-tail-knowledge-s6-conclusion-and-future-work.md
Created 2026-08-25T02:58:07+00:00
BM25 top-k recall varies only mildly (approximately 0.3 to 0.8) as the number of relevant documents in the corpus ranges from 10^0 to 10^4.
Summary
BM25's ability to pull the right documents into its top results stays fairly steady, hovering between about 0.3 and 0.8, even as the pool of truly relevant documents in the corpus grows from a handful to tens of thousands. This means the system can rely on consistent retrieval quality at scale without needing to re-tune the search as the corpus expands.