BM25

BM25 is a ranking function for sparse bag-of-words retrieval. It saturates term frequency and length-normalizes documents, usually beating plain TF-IDF in practice.

Still a default first-stage retriever and a strong complement to embeddings.

← Back to AI Glossary