← All notes

Summary

Dense Passage Retrieval for Open-Domain Question Answering (2020), by Vladimir Karpukhin, Barlas Oguz, Sewon Min, Patrick Lewis, Ledell Wu, Sergey Edunov, Danqi Chen, and Wen-tau Yih, is at https://arxiv.org/abs/2004.04906. At the time, open-domain question answering usually relied on sparse retrievers such as TF-IDF or BM25. The authors show that retrieval can be done with dense vectors alone, learning embeddings from a small number of questions and passages in a simple dual-encoder framework. They report that this dense retriever outperforms a strong Lucene-BM25 system by 9% to 19% absolute on top-20 passage retrieval accuracy across a range of open-domain QA datasets, and that the improved retrieval helped their end-to-end QA system set new state-of-the-art results on multiple benchmarks. When evaluated on a wide range of open-domain QA datasets, the dense retriever outperforms a strong Lucene-BM25 system largely by 9% to 19% absolute in terms of top-20 passage retrieval accuracy, and helps their end-to-end QA system establish new state-of-the-art on multiple open-domain QA benchmarks. Title, authors, abstract, and comments were taken from the arXiv abs page. The version is v3 (30 Sep 2020), submitted 10 Apr 2020.

Claims

Pointer notes do not quote the paper. See Verification and Sources for the link and license.

Verification

• Format: pointer note only. The paper is not reproduced here.
• Why it matters: It matters because it established the dual-encoder recipe that most later neural retrievers still follow.
• Reproduced: not reproduced (2026-11-11).
• Related work: Related notes already in the library include RAG basics and Self-RAG.
• Limits: This is a pointer note only. The 9% to 19% top-20 gains are the authors' reported results versus Lucene-BM25 from the v3 paper (30 Sep 2020).
• License: arXiv default (non-exclusive distribution license) on the abs page (2026-11-11; http://arxiv.org/licenses/nonexclusive-distrib/1.0/). Submitted 10 Apr 2020 (v1); last revised 30 Sep 2020 (v3). Sources: [S1] https://arxiv.org/abs/2004.04906, v3 (30 Sep 2020). checked_on: 2026-11-11. Paper license: arXiv non-exclusive distribution; this note links to it and does not reproduce it. The note itself is CC BY 4.0.

When to apply

Use when you need the original Dense Passage Retrieval paper, the standard dual-encoder recipe behind many later neural retrievers.

Facts

• id: F1 claim: 'Dense Passage Retrieval learns embeddings from questions and passages with a simple dual-encoder framework, and shows that retrieval can be implemented using dense representations alone.' confidence: medium date: 2026-11-11 sources: [S1] license: CC-BY-4.0
• id: F2 claim: 'The authors report that the dense retriever outperforms a strong Lucene-BM25 system by 9% to 19% absolute in top-20 passage retrieval accuracy on a range of open-domain QA datasets.' confidence: medium date: 2026-11-11 sources: [S1] license: CC-BY-4.0
• id: F3 claim: 'They report that this retrieval helped their end-to-end QA system establish new state-of-the-art results on multiple open-domain QA benchmarks.' confidence: medium date: 2026-11-11 sources: [S1] license: CC-BY-4.0

Sources

• [S1] arXiv 2004.04906, https://arxiv.org/abs/2004.04906. Paper license: arXiv default (non-exclusive distribution license). v3 (30 Sep 2020). Accessed 2026-11-11. This note links to the paper and does not reproduce it.

License: this note is CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Pointer by iinventedlight (AI-assisted). Paper license: arXiv non-exclusive distribution; this note links to it and does not reproduce it.

@misc{iil_note_zxrkad6v,
  author = {{'iinventedlight (AI-assisted)'}},
  title = {{Dense Passage Retrieval for Open-Domain Question Answering}},
  howpublished = {iinventedlight Library},
  year = {2026},
  date = {2026-11-11},
  url = {https://iinventedlight.com/library/notes/2004-04906-dpr/},
  note = {Permanent ID iil:note:zxrkad6v. License: CC-BY-4.0.},
}

Markdown

Spotted a mistake? Use the MCP feedback tool or email hello@iinventedlight.com.

For language models and agents: start at /llms.txt | /agents.md | /.well-known/agent-library.json. This page in Markdown: /library/notes/2004-04906-dpr.md.