arXiv Preprints: What They Are and How to Use Them

arXiv (pronounced “archive,” styled with a capital X for the Greek chi) is a free, open-access repository where researchers post preprints and, in some fields, accepted manuscripts, before or alongside formal journal publication. It hosts more than three million scholarly articles across physics, mathematics, computer science, and several other fields, and is one of the most heavily used pieces of open research infrastructure in existence. This guide covers what arXiv actually is, how its subject categories and moderation system work, how it differs from peer-reviewed publication, and how to cite an arXiv preprint correctly. For the specific mechanics of getting a first submission approved, see the companion term on the arXiv endorsement system; for the general concept this guide builds on, see preprint.

What arXiv Is (and Isn’t)

arXiv was founded in 1991 by physicist Paul Ginsparg, originally hosted at Los Alamos National Laboratory as a way for physicists to circulate papers faster than journal publication allowed. It relocated to Cornell University in 2001, where Cornell University Library operated it for 25 years. On July 1, 2026, arXiv completed a long-planned spin-out and now operates as an independent nonprofit organization with its own Board of Directors, with Cornell University and the Simons Foundation as founding members and funders alongside a broader base of member institutions and donors. arXiv remains free to read and free to submit to.

What arXiv is not: a journal, a peer-review venue, or a publisher in the traditional sense. Nothing posted to arXiv has been through editorial peer review as a condition of appearing there. Submissions go through a lighter, distinct process, endorsement and moderation, covered below, that checks scope and baseline scholarly standards rather than correctness or significance.

Subject Categories

Every arXiv submission is assigned to at least one subject category (also called an “archive” or, at a finer grain, a discipline-specific taxonomy), which determines where it’s listed, who moderates it, and which endorsers can approve it. arXiv organizes content into eight top-level subject areas:

• Physics — by far the largest and oldest, itself split into many sub-categories such as astro-ph (astrophysics), cond-mat (condensed matter), gr-qc (general relativity and quantum cosmology), hep-th/hep-ph/hep-ex/hep-lat (high energy physics, theory/phenomenology/experiment/lattice), math-ph, nucl-ex/nucl-th, and quant-ph, among others.
• Mathematics (math.*) — subdivided by area, e.g. math.AG (algebraic geometry), math.PR (probability), math.NT (number theory).
• Computer Science (cs.*) — subdivided by area, e.g. cs.LG (machine learning), cs.CL (computation and language/NLP), cs.CV (computer vision).
• Quantitative Biology (q-bio.*)
• Quantitative Finance (q-fin.*)
• Statistics (stat.*)
• Electrical Engineering and Systems Science (eess.*)
• Economics (econ.*)

A submission has one primary category but can be cross-listed into secondary categories where it’s also relevant, which is how, for example, a machine-learning paper with a statistics angle can appear in both cs.LG and stat.ML listings. Choosing an appropriate primary category matters beyond discoverability: it determines which moderators review the submission and which established authors are eligible to endorse a first-time submitter in that category.

How Endorsement and Moderation Work

arXiv uses two separate, sequential gatekeeping mechanisms, and conflating them is a common mistake:

Endorsement happens before a first submission to a given subject category is accepted. A researcher submitting to a category for the first time generally needs endorsement from an existing arXiv author in good standing who has a sufficient publication history in that same category. Endorsement checks only whether the submission is topically appropriate for the category — it is explicitly not a judgment of scientific quality or correctness, and it is not peer review. Full mechanics, including who qualifies as an endorser, are covered in the arXiv endorsement system entry.

Moderation is separate, ongoing, and applies to every submission, not just first-timers. arXiv moderators are volunteer subject-matter experts, generally holding a terminal degree in the relevant field, approved by discipline-level advisory committees and arXiv staff. Moderators check for baseline scholarly standards (proper formatting, references, professional language), scholarly and topical appropriateness, and policy compliance (no plagiarism, correct rights to license the work, disclosure of generative-AI tool use where applicable). They can reclassify a submission to a more appropriate category, decline a submission outright, or, after the fact, reclassify or withdraw an already-announced paper found to violate policy. None of this is peer review either: moderators are not assessing whether the paper’s conclusions are correct, only whether it meets arXiv’s baseline standards and belongs where it’s been placed.

How arXiv Differs From Peer-Reviewed Publication

The distinction matters for how a research administrator should treat an arXiv posting in a CV, grant report, or tenure file:

• No peer review. Endorsement and moderation are quality-control layers, but neither substitutes for the independent expert review a journal or conference typically requires before acceptance. A paper being “on arXiv” says nothing about whether it has been, or will be, formally peer reviewed.
• Versioned, and mutable. Authors can post revised versions (v1, v2, v3…) of the same submission as they revise the work, including in response to informal feedback or a parallel peer-review process elsewhere. Earlier versions remain permanently accessible, but the “current” version by default is the latest one — which is why precise citation should specify a version when the distinction matters (see below).
• No formal retraction process in the journal sense. An author can withdraw a submission, which marks it as withdrawn while preserving the record, but arXiv is not the venue that issues a “retraction” in the sense used for a published journal article; that determination belongs to the eventual publishing journal, if any.
• Coexists with, rather than replaces, journal publication. Most fields with strong arXiv cultures (physics, math, CS) still treat the peer-reviewed journal or conference version as the version of record for tenure and credit purposes, with the arXiv posting serving as the freely accessible, rapidly disseminated counterpart. See how to know if an article is peer reviewed for how to check whether a specific arXiv-hosted paper has also cleared formal peer review elsewhere.

How to Cite an arXiv Preprint

Since January 2022, arXiv has registered a DOI (via DataCite) for essentially every article in its corpus, in the format 10.48550/arXiv.[identifier]. That DOI is the most durable, style-guide-friendly way to cite an arXiv posting, and most current citation-style guidance (APA 7th, for example) now recommends citing the DOI directly rather than a bare arXiv URL where one is available.

Where no DOI is cited, or where the version matters, use the arXiv identifier itself, in the format arXiv:YYMM.NNNNN (year and month of first submission, followed by a sequence number), optionally suffixed with a version number (e.g. arXiv:2301.04567v2) when it’s important to cite the specific version referenced, since content can change between versions. A typical reference-list citation includes: author(s), year, title, “arXiv” as the source, the identifier or DOI, and, where required by the citation style, “preprint” as a descriptor to make clear the work has not necessarily been peer reviewed. For worked examples across citation styles and the same logic applied to other preprint servers, see how to cite a preprint.

When arXiv Is (and Isn’t) the Right Venue

Posting to arXiv is generally appropriate for rapidly disseminating results ahead of or alongside formal peer review, establishing a timestamped priority claim on a method or result, gathering informal community feedback before or during journal submission, and meeting funder or institutional preprint-sharing expectations. Most major journals in arXiv-heavy fields (physics, math, CS in particular) explicitly permit prior arXiv posting and do not treat it as prior publication that would disqualify a later submission — but this varies by field and specific journal, and it is worth checking a target journal’s preprint policy before assuming it’s compatible, particularly outside disciplines where preprinting is already the norm.

arXiv posting is not a substitute for formal publication where a venue, funder, or evaluation process specifically requires peer-reviewed output, and it should not be represented as such. Where genuine confidentiality is required (e.g. patent-sensitive results before filing, or embargoed clinical findings), posting to a public preprint server before those protections are in place is generally the wrong choice.

arXiv vs. Other Preprint Servers

arXiv is a subject-based, moderated repository dating to 1991 and rooted in physical sciences, math, and computer science — distinct in scope, governance, and moderation model from field-specific servers that emerged later. For direct side-by-side comparisons, see arXiv vs. bioRxiv (life sciences preprints), arXiv vs. Research Square, and arXiv vs. SSRN (social sciences and humanities).

Frequently Asked Questions

Is arXiv peer reviewed?

No. arXiv submissions go through endorsement (a topical-fit check for first-time category submitters) and moderation (a baseline scholarly-standards and policy check), neither of which is peer review. A paper’s presence on arXiv is not evidence that it has undergone independent expert review.

Do I need permission to post my paper on arXiv?

You need the legal right to license the work under arXiv’s terms — this can matter if a journal you plan to submit to has a restrictive prior-publication policy, or if a co-author or funder has separate requirements. Check the target journal’s preprint policy and any funder mandate before posting, particularly outside fields where preprinting is already standard practice.

What does the “v2” on an arXiv paper mean?

It’s the second version of that submission; authors can post revisions after the original v1, and all prior versions remain accessible. Cite the specific version if the distinction between versions matters to your point.

Is arXiv free to use?

Yes, both reading and submitting are free. arXiv is funded by member institutions, the Simons Foundation, and other donors rather than by author or reader fees.

Can I withdraw a paper from arXiv?

Yes, authors can withdraw a submission, which marks it as withdrawn while preserving the record rather than deleting it outright. This is a different mechanism from a journal retraction; see arXiv’s own policies for the current withdrawal process before assuming a withdrawn arXiv posting behaves like a retracted journal article.