I read in the instructions for one conference:

Please do not use arXiv references for papers that have been published at official venues such as conferences, journals or workshops. Please make the effort to identify these at least for all papers prior to 2017.

What are the downsides of using arXiv references instead of conference/journal references?

  • 7
    As a general comment for those advocating that arxiv isn't paywalled, I usually google the name of the paper I'm looking for if I can't access it by other means, leading to arxiv/researchgate/archives-ouvertes/etc... Commented Feb 12, 2017 at 0:52
  • 4
    When you say "somewhere", what kind of somewhere is that?
    – E.P.
    Commented Feb 12, 2017 at 14:30
  • 1
    @E.P. Instructions for camera ready version of papers at EACL 2017 sent by email to authors. Commented Feb 12, 2017 at 17:23
  • 1
    @FábioDias: Even though the reader may be able to find open-access copies of a cited paper, there are valid reasons to directly cite the arXiv paper. At least: (1.) saving reader effort, in particular if you put a clickable hyperlink to the arXiv version; (2.) making it clear that references to figures/theorems/tables/etc. are relative to the arXiv version and not the publisher version (or some other online version), in case they differ; (3.) not give the impression to readers that they are encouraged to pay for the publisher version of the paper.
    – a3nm
    Commented Feb 13, 2017 at 0:43
  • 1
    @FábioDias you can cite a specific version of an arXiv article. Commented Feb 13, 2017 at 17:01

5 Answers 5


Using arXiv instead of a conference / journal reference has an obvious disadvantage: readers won't know that the paper eventually passed peer review and was published. This is important context, especially if they want to cite the paper themselves later.

Using arXiv in addition to a conference / journal reference might make sense, but some might consider it redundant.

  • 43
    readers won't know that the paper eventually passed peer review and was published. Indeed, +1. They also won't know where it was published, which can be just as important a piece of context (e.g., Annals of Mathematics vs Predatory Journal of Mickey Mouse Mathematics).
    – Dan Romik
    Commented Feb 11, 2017 at 21:57
  • 2
    +1 for do both. That's good practice, and not redundant as arxiv may contain a newer version, extra material and is open access.
    – innisfree
    Commented Feb 12, 2017 at 5:19
  • 3
    @innisfree newer version than the journal version? Extra material over the journal version? Are you kidding? It would be extremely bad practice to add material and keep updating the paper on arxiv after it has already published, except maybe to correct some egregious error, and even then an erratum should be issued, since many people would go to the journal as the official source for the paper rather than arxiv.
    – Dan Romik
    Commented Feb 12, 2017 at 6:18
  • 11
    Yep, extra material, that was cut from published version for brevity or as it was e.g., speculative. And updates for typos, or even updates in light of new experimental data. Not considered bad practice in physics.
    – innisfree
    Commented Feb 12, 2017 at 6:32
  • 5
    Journals often ask that any archiving (eg on the arXiv) has a reference and a link (eg by doi) to the published version. I've started including the arXiv number (inspired by Scott Morrison) in the abstract of the published paper in a reversal of this demand. eg dx.doi.org/10.1007/s11225-015-9603-6, dx.doi.org/10.4310/HHA.2015.v17.n2.a4, dx.doi.org/10.1007/s10485-015-9400-4 Thus if people leave off the arXiv number in their reference to my paper, then anyone looking up the journal version will have it in front of them with no effort at all, and ideally won't waste $£€ Commented Feb 12, 2017 at 10:06

One reason I haven't seen mentioned is indexing of citations. Some services that create citation statistics might not correctly associate the preprint with the published paper.

But linking to the arXiv version is always a good idea:

  • It's a lot easier for the author to update an arXiv preprint than the version in a Journal. E.g one of my papers has some (minor) errors in the published version which are fixed on arXiv.

  • Not everyone has access to paywalled journals, especially not all the time (e.g. from a mobile device).


Venues are important for judging if a paper is worth reading.

ArXiv is a mixed bag, as it is not peer reviewed, and may even have been changed!

If I look at references, and they come from some predatory publisher or fake conference, I don't bother reading them. Experience tells me they are at most a rip-off of some other work, and in the worst case simply copy&paste from Wikipedia.

Now if your work is based on these things, I would be rather skeptical if it really works. Even many papers published in top conferences are nonsense and not reproducible.

So please, indicate which papers have at least passed some peer review, and in which venue - to at least hint at their quality.

  • 6
    "Venues are important for judging if a paper is worth reading. ArXiV [sic] is a mixed bag" -- compare to the largest STM academic publisher: en.wikipedia.org/wiki/Elsevier#Criticism_and_controversies Commented Feb 12, 2017 at 10:09
  • 3
    "and may even have been changed" Non-issue, because you should cite arXiv preprints with their version number!! Commented Feb 12, 2017 at 20:40
  • 1
    The "not peer reviewed" thing is not as important as you seem to make it. Usually, authors retract their arXiv preprints if the referee finds a serious error (although authors don't always correct minor errors found by referees). Commented Feb 12, 2017 at 20:41
  • 4
    @darijgrinberg: Lots of arXiv preprints have not yet undergone peer review, and many never will. Commented Feb 13, 2017 at 5:03
  • @darijgrinberg you should does not imply that people do it right... people also should cite Wikipedia only with a revision number, yet nobody does. And I've seen many - even prominent, not only crap - unreviewed papers on arXiv. Commented Feb 13, 2017 at 19:24

The arXiv version is not always updated to match the published version, which is the official "version of record" (although these days it is rarer and rarer that the arXiv version is not updated to match the published version).

  • 1
    This is true, but not always: sometimes it may also happen that the arXiv version is updated after publication to fix errors that were found since then. It's up to everyone to decide what they consider to be the "official version of record".
    – a3nm
    Commented Feb 13, 2017 at 0:40

The main concern should be to assist your future readers as much as possible, including to enable them to track down the source article. If you can direct them to the latest version and provide a DOI for them to track it easily, then it would be beneficial to do so.

If that version happens to be a reputable peer-reviewed journal, then all the better to get your readers to trust your claims. If submitting to a journal they will also use your references to determine whether your topic is relevant for their publication, they expect you to have cited publications in the field including their journal.

Disclosure of the version you've based your claims on is a valid concern. In my experience bioRxiv updates manuscripts to the published version but it may be worth checking that the aspects you've based your citation on have not changed, particularly if they've been retracted. However, it does look good to show that you have included the latest findings in your field in your literature review. Perhaps it may be suitable to cite the published version and include a sourced at arXiv note or url? Similar to how you would give the date accessed (version) for a website source.

  • The optimal way to assist readers is to link both the journal and the arXiv version, possibly even pointing out whether the two versions substantially deviate (if you know it). I tend to do exactly that in my preprints. Unfortunately, this kind of information maps badly onto standard bibliography formats, which is why I think not many people do it. (I write my bibliography manually, so I don't have troubles doing it.) Commented Feb 12, 2017 at 20:39
  • @darijgrinberg Ideally (as I've suggested) but where only one can be given, I'd recommend the (more credible) journal if possible. Giving both versions if may not be accepted by many journals in the interests of space. Even a note such as "sourced on arXiv" should be enough for most readers to trace it down on google if they had the journal reference.
    – Tom Kelly
    Commented Feb 12, 2017 at 22:29

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .