I'm writing a short (say around 7-10 page long) article on theoretical machine learning and I've established the main theoretical result with a complete mathematical proof. I've also written the other parts - the abstract, introduction, acknowledgements and bibliography. However, some computer simulations still need to be added.

I'm thinking of uploading the paper onto arXiv for my ongoing academic application purposes, so that potential employers can study the paper. I know an obvious question for it, being a machine learning paper, will be "Where are some illustrations?" So for the moment, if I remark in my paper along the line of "Illustrations are still to be added" and then finally add it to arXiv, will it be academically/professionally harmful for me?

The reason I want to add this instead of finishing it all the way is that currently I'm working on a different project and I feel it'd take me a while to go back to the abovementioned one, yet I'd like for the new employer to notice the result therein.

Q2: same question where only part of the theory has been established (albeit for a different paper).

Thank you!

  • 6
    It certainly looks much more professional if all your papers on arXiv are finished (i.e. are in a state that you would submit to a journal). Commented Nov 16, 2020 at 12:21
  • ‘ I know an obvious question for it, being a machine learning paper, will be “Where are some illustrations?” ’ — yeah that's a bit of an understatement! You might in fact say, ‘in the machine learning community, nobody gives a damn about theory / proofs that don't come with some example results on real data’. Commented Nov 16, 2020 at 21:39
  • How is that about Academia, rather than the technical details of ArXiV, please? Commented Nov 16, 2020 at 23:38
  • @RobbieGoodwin It's obviously about academia, it's not like arxiv will block a paper uploaded without images in it. Commented Nov 17, 2020 at 18:57

2 Answers 2


I suspect the relative merits of different approaches depends on discipline and maybe geography, but in my area (interdisciplinary with applied mathematics and business most prominent, North America), a preprint, on ArXiV or elsewhere, should be essentially complete within the scope the authors have chosen for it to have. So you could upload a preprint which just doesn't discuss those areas of your research you have not had time to complete, and which does not have all illustrations you might ideally want for the best possible exposition. A preprint may well end up significantly rewritten and expanded as part of a journal's peer review process, and it is only slightly disingenuous if you know fully well you're actually going to do that soon anyway.

However, importantly, your preprint should be complete enough you believe in good faith it could be publishable somewhere as-is, even if you are aiming for something better. We're not talking about a missing methododology or results section, or text that references still-nonexistent crucial figures or tables, in a way that can't be edited out without affecting the completeness of the paper.

If there are significant gaps which make this approach inappropriate, you have a work-in-progress draft, not a preprint yet. There is nothing precluding you from thoughtfully sharing that with colleagues, by email or even by say posting on your website, with all the pluses or minuses that entails. So if you want potential employers to have access, why not refer to it as work-in-progress or working draft and put a link to it stored on your research website for now?

Of course, in this case, you won't get credit for it as a "completed publication". But it is an option if it is unfinished but still interesting.


Such papers can certainly be accepted by the arXiv system (which does include some very minimal "refereeing" by volunteers, but only at the approve/disapprove level if not also for changing the listing for example from "general physics" to "chemical physics").

However it is frowned upon to submit incomplete work on arXiv. It is also terribly inconvenient for people who have printed and read your paper, only to find that it is significantly different from the final version of the paper.

Conlusion: You can do it, but it's better to be patient and submit a quality product that you're comfortable keeping up permanently (then make changes only in the hopefully rare event that you later spot a typo or error that needs to be corrected).

If you still want the material to be publicly available without the above problem, you can always post it in a GitHub repository, or Academia.edu or Mendeley, or ResearchGate, or elsewhere. This question from just 2 days ago is also related: Pre-print service like Arxiv but with private option, so I can correct my mistakes without publicly advertising my mistakes?.

You must log in to answer this question.

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