What is BM25?

Why does BM25 matter?

It is a strong exact-term retrieval baseline and complements dense retrieval for identifiers, rare words, and domain-specific phrases.

BM25 in practice

Retrieve candidates with BM25 and dense search, combine their ranks, then evaluate the merged results before adding a more expensive reranker.

What is the common confusion about BM25?

BM25 does not understand semantic similarity directly, and its score has no universal meaning across different queries or index configurations.

Learn BM25 in the course

Lessons that name BM25 in a title or section

• Hybrid Retrieval with BM25 and Dense Embeddings

  Lexical and semantic retrieval fail on opposite query distributions. Hybrid retrieval with reciprocal rank fusion does not interpolate, it votes - and the vote wins on every query class.

  Phase 19: Capstone Projects

• Information Retrieval and Search

  BM25 is precise but brittle. Dense casts a wide net but misses keywords. Hybrid is the 2026 default. Everything else is tuning.

  Phase 05: NLP: Foundations to Advanced

• Advanced RAG (Chunking, Reranking, Hybrid Search)

  Basic RAG retrieves the top-k most similar chunks. That works for simple questions. It falls apart for multi-hop reasoning, ambiguous queries, and large corpora.

  Phase 11: LLM Engineering

Covered in Phase 05: NLP: Foundations to Advanced, Phase 11: LLM Engineering and Phase 19: Capstone Projects.

Related terms

• Hybrid RetrievalRetrieval that combines signals from different methods, commonly lexical matching and dense-vector similarity, before merging or reranking…
• Dense RetrievalFirst-stage retrieval that embeds queries and candidates into vector representations and ranks candidates by a similarity function.
• RerankerA second-stage model or scoring function that reorders a small candidate set using a richer comparison between the query and each candidate.
• RAG (Retrieval-Augmented Generation)A system pattern that retrieves evidence relevant to a request and supplies selected content to a generative model before it answers or…
• Reciprocal Rank Fusion (RRF)A rank-fusion method that combines several result lists by summing contributions that decrease with each item's rank in each list.

Sources

• The Probabilistic Relevance Framework: BM25 and Beyond

More terms in Retrieval & generation

• Approximate Nearest Neighbor (ANN)
• Chunking
• Dense Retrieval
• Grounding
• HNSW
• Hybrid Retrieval
• Maximum Marginal Relevance (MMR)
• RAG (Retrieval-Augmented Generation)
• Recall@K
• Reciprocal Rank Fusion (RRF)
• Reranker
• Semantic Search
• Vector Database

Open the Retrieval & generation list in the glossary