arXiv Metadata (OAI-PMH)
arXiv, Cornell University
Metadata for every arXiv preprint since 1991, updated daily, the raw record of physics, math, and AI.
arXiv offers bulk metadata for all its articles through OAI-PMH, which it calls the preferred way to keep an up-to-date copy. Each record is a preprint with title, authors, abstract, categories, versions, and license. Full-text PDFs and sources are available separately from Amazon S3 and Kaggle.
Access
- Cost
- FreeMetadata is free via OAI-PMH, the API, and a Kaggle snapshot; S3 full-text bulk access may incur AWS transfer charges.
- License
- Public domainMetadata CC0; full text under each article's own license
- Sign-up
- No sign-up
- How you get it
- API and Bulk downloadXML, JSON, PDF
What's in it
- One row is
- One record per preprint, with version history
- Covers
- Worldwide authors
- Years
- 1991-present
- Updated
- Daily
- Size
- More than 2 million preprints
Quality
Clean, authoritative, and complete for its scope, with daily updates.
- Author names are free text, not disambiguated
- Category labels are self-chosen by submitters
Connects to
- Semantic Scholar Academic Graph (S2AG): Citation counts for arXiv papers
- OpenAlex: Links preprints to later journal versions
Questions it could answer
- How fast did machine learning submissions grow relative to physics?
- How often are preprints revised, and how much do abstracts change?