arXiv identifiers explained: format, versions and arXiv DOIs

Last reviewed on October 3, 2026

Short answer: an arXiv ID (arXiv identifier) is the permanent name arXiv gives every paper it hosts, for example arXiv:2401.12345. The first four digits are the year and month the paper was submitted (here January 2024), and the number after the dot is its sequence number in that month. A version suffix such as v2 points to a specific revision. Any arXiv ID can be opened at https://arxiv.org/abs/ followed by the ID.

The two formats

In new-style IDs the subject category is not part of the identifier. You often see it next to the ID, for example arXiv:1706.03762 [cs.CL], but [cs.CL] is only a label. In old-style IDs the archive name (such as hep-th or math) is part of the identifier and cannot be dropped.

What a valid arXiv number looks like

A quick validity check:

• New style: ^\d{4}\.\d{4,5}(v\d+)?$, where the first two digits are the year, the next two a month from 01 to 12, and the number has 4 digits for 0704–1412 and 5 digits from 1501 on.
• Old style: ^[a-z-]+(\.[A-Z]{2})?/\d{7}(v\d+)?$, for example cond-mat/0207270.
• The prefix arXiv: is optional when you store the ID, but it is conventional in citations.

The final test is simply whether https://arxiv.org/abs/<id> opens a paper.

Versions

Authors can replace a paper with a revised version. Each revision keeps the same ID and gets a new suffix: v1, v2, and so on. An ID without a suffix always resolves to the latest version. Cite with a version when the exact text matters (for example when quoting page numbers or results that changed between versions); otherwise the bare ID is fine.

arXiv IDs and DOIs

Since 2022 arXiv registers a DOI for every paper through DataCite, including older ones. It is built directly from the arXiv ID: 10.48550/arXiv.2401.12345. It is a real, resolvable DOI, so reference managers and citation styles that expect a DOI will accept it. You do not need to request or create it; it exists automatically.

That arXiv DOI identifies the preprint. If the paper was later published in a journal or proceedings, the published version has its own, different DOI from the publisher. arXiv shows it on the abstract page when the authors have added it. When citing a published paper, prefer the publisher's DOI; when only the preprint exists, cite the arXiv ID or the arXiv DOI.

How to find the arXiv ID of a paper

• In the PDF: arXiv-generated PDFs carry a vertical stamp in the left margin of the first page, such as arXiv:1810.04805v2 [cs.CL] 24 May 2019.
• In the URL: it is the last part of arxiv.org/abs/… or arxiv.org/pdf/….
• By title: search on arxiv.org/search and copy the ID from the result.
• From a PDF you already have: drop it into the in-browser extractor; it reports the arXiv ID (old or new style) if it appears in the first pages.

arXiv IDs in reference managers

Zotero's "Add Item by Identifier" accepts an arXiv ID and fetches the full record. Mendeley's document details include an arXiv ID field alongside DOI and PMID, and can look the record up from it. In BibTeX (biblatex) the usual fields are eprint = {2401.12345} with eprinttype = {arxiv} (or archivePrefix = {arXiv} in natbib-style files) and optionally primaryClass.

arXiv author identifiers

Authors on arXiv can also have a public author identifier, which gives a stable listing of their papers at an address like arxiv.org/a/lastname_f_1. It identifies a person, not a paper, and can be linked to an ORCID iD.