However, I tried on several arXiv papers, and on all of them when clicking on references I get a "Sorry! We couldn't find the references for this paper. This paper might have been posted recently. Try again later."
What's the point of referencing arXiv? They're supposed to be preprints or non-peer reviewed papers. Other than for re-using some of the text/images, the citations to arXiv papers won't/shouldn't be accepted by peer-review committees.
Something on arXiv can have had more "peer-review" than something that made it past 3 reviewers for a normal publication if it has been discussed a lot.
Also, not everything cited has to be properly published. If you base an argument on a finding from a source, then yes, but e.g. but descriptions of approaches, ideas, ... are things that can be cite-worthy while not requiring the "fact-checking" peer review provides.
It's an interesting balance: On the one hand, it's bad if stuff that just was published as a preprint is treated as established (which seems to be partially the case in ML right now), on the other hand not referencing good material just because it "doesn't look scientific enough" also doesn't help anyone. (In some parts of computing, it seems like academia is toying with stuff industry has tried and discarded years ago, but that isn't acknowledged because it hasn't been published in a nice citation. It's fine to do work to validate that, but totally ignoring it is weird)
At least in math, that is just untrue. It can be years from when the first preprint comes out to when it's actually published. If everyone waited around for the official publication, nothing would ever get published. Journals are pretty pragmatic about it.
In addition to what others have said, there are lots of reasons to cite something that don't need that work to be correct. You cite works that inspire yours, even if they're not peer-reviewed. You cite works that ask questions you answer, as evidence that the problem is worth addressing. You cite papers that are works in progress but are related to what you are doing and are useful to give a reader background. Sometimes you intentionally cite mistaken work specifically so that you can correct it.
In CS, it also occasionally makes sense to cite an arXiv paper if it hasn't been accepted in a peer-reviewed publication. It's still helpful to acknowledge that you're aware of related work that exists regardless of what form it's been published in.
It is common practice to cite things other than peer-reviewed papers (at least in math). There are even citations for "private communication", usually meaning a personal email or conversation.
[+] [-] jmnicholson|8 years ago|reply
The arXiv of the future will not look like the arXiv https://www.authorea.com/users/3/articles/173764-the-arxiv-o...
[+] [-] alexeyza|8 years ago|reply
However, I tried on several arXiv papers, and on all of them when clicking on references I get a "Sorry! We couldn't find the references for this paper. This paper might have been posted recently. Try again later."
(on chrome 59.0.3071.115)
[+] [-] desku|8 years ago|reply
[+] [-] jesuslop|8 years ago|reply
[+] [-] mankash666|8 years ago|reply
[+] [-] detaro|8 years ago|reply
Also, not everything cited has to be properly published. If you base an argument on a finding from a source, then yes, but e.g. but descriptions of approaches, ideas, ... are things that can be cite-worthy while not requiring the "fact-checking" peer review provides.
It's an interesting balance: On the one hand, it's bad if stuff that just was published as a preprint is treated as established (which seems to be partially the case in ML right now), on the other hand not referencing good material just because it "doesn't look scientific enough" also doesn't help anyone. (In some parts of computing, it seems like academia is toying with stuff industry has tried and discarded years ago, but that isn't acknowledged because it hasn't been published in a nice citation. It's fine to do work to validate that, but totally ignoring it is weird)
[+] [-] tacomonstrous|8 years ago|reply
[+] [-] tgb|8 years ago|reply
[+] [-] michaelmior|8 years ago|reply
[+] [-] jrells|8 years ago|reply
[+] [-] BucketSort|8 years ago|reply
[+] [-] versteegen|8 years ago|reply