ModelRefs / Sparse Retrieval — AI Glossary
Sparse Retrieval — AI Glossary
Retrieval using term-frequency representations (BM25, TF-IDF) that match on exact and overlapping vocabulary. Fast, interpretable, and requires no GPU.
Overview
Sparse retrieval assigns each document a high-dimensional sparse vector over the vocabulary. BM25 is the dominant algorithm: TF-IDF weighted by document length normalization. Fast, interpretable, and requires no GPU. Excels on exact-match, rare-word, and named-entity queries. Weakness: no understanding of synonyms or paraphrases.
Reference details
| Topic | rag |
|---|---|
| Also known as | BM25 retrieval, keyword retrieval, lexical retrieval |
| Last reviewed | 2026-06-24 |
Related terms
Commonly confused with
The category, of which BM25 is the standard implementation, named for the vectors: one dimension per vocabulary term, almost all zero. Dense retrieval is the counterpart, with a few hundred or thousand dimensions all populated. Learned sparse methods blur the line by producing sparse vectors from a model, which have the exact-match property with some semantic reach.
Continue your research
Use these connected ModelRefs sections to compare alternatives, inspect implementation paths, and review the evidence and governance boundaries relevant to Sparse Retrieval — AI Glossary.
Frequently asked questions
What is Sparse Retrieval?
Retrieval using term-frequency representations (BM25, TF-IDF) that match on exact and overlapping vocabulary.
Is Sparse Retrieval the same as BM25 retrieval?
Yes — BM25 retrieval, keyword retrieval, lexical retrieval are common aliases for Sparse Retrieval.
What concepts are related to Sparse Retrieval?
Closely related concepts include dense retrieval, fusion retrieval, ivf.