New arXiv policy: 1-year ban for hallucinated references

(twitter.com)

231 points | by gjuggler 3 hours ago

10 comments

  • btown 3 hours ago
    > The penalty is a 1-year ban from arXiv followed by the requirement that subsequent arXiv submissions must first be accepted at a reputable peer-reviewed venue.

    This is incredibly good for science. arXiv is free, but it's a privilege not a right!

    I'm not seeing this clearly listed on https://info.arxiv.org/help/policies/index.html so it's possible this is planned but not live yet - or perhaps I'm not digging deeply enough?

    As a certain doctor once said: the whole point of the doomsday machine is lost if you keep it a secret!

    • paulpauper 11 minutes ago
      No, this is incredibly bad.

      Arxiv doesn't even check the submission closely, so how can they know?

      subsequent arXiv submissions must first be accepted at a reputable peer-reviewed venue.

      Second, there is such a huuuuge gap between publishing on arvix and peer review. I can attest personally that it's not even close. I've gotten probably dozen rejections from peer review and no problems publishing in arxiv math. This is because peer review checks not just for if something is new or correct, but also if it's of "interest to math community," which is inherently subjective, but also makes peer review many magnitudes harder than publishing on arxiv Even when a well-known professor in number theory praised the paper when I got an endorsement and a second emailed me and and encouraged me to publish it, it still got rejected 3 times and still waiting.

      • helterskelter 8 minutes ago
        You could at least filter out hallucinated references which simply don't exist pretty trivially, I'd imagine.
    • dataflow 1 hour ago
      > This is incredibly good for science.

      I disagree. It's just one darn hallucinated citation for heaven's sake, not fraud or something. It doesn't account for the substance or quality of their work at all. A one-year ban seems plenty sufficient for a minor first time mistake like this. People make mistakes and a good fraction of them can learn from those mistakes. There's no need to permanently cripple someone's ability to progress their life or contribute to humanity just because an AI hallucinated a reference one time in their life. That's punitive instead of rehabilitative.

      • toast0 1 hour ago
        > It's just one darn hallucinated citation for heaven's sake, not fraud or something.

        It is fraud.

        > It doesn't account for the substance or quality of their work at all.

        References are part of the work. If you're making up the references, what else are you making up?

        > People make mistakes and a good fraction of them can learn from those mistakes. There's no need to permanently cripple someone's ability to progress their life or contribute to humanity just because an AI hallucinated a reference one time in their life.

        A one year ban is not permanent. Having a negative consequence for making poor decisions seems like an inducement to learn from the mistake?

        In an ideal world, one would be keeping notes on references used while doing the research that lead to writing the paper. Choosing not to do that is one poor decision.

        Having a positive outlook, if asking an AI to provide references that may have been missed, one should at least verify the references exist and are relevant. Choosing not to do that is also a poor decision, even if one did take notes on references used while researching.

        • dataflow 1 hour ago
          > It is fraud.

          No, it is emphatically not. Fraud requires intent to deceive.

          > A one year ban is not permanent.

          ...what text are you reading? Nobody was calling the one-year ban permanent, or even against it. I was literally in favor of it in my comment. I explicitly said it is already plenty sufficient. What I said is there's no need to go beyond that. My entire gripe was that they very much are going beyond that with a permanent penalty. Did you completely miss where they said "...followed by the requirement that subsequent arXiv submissions must first be accepted at a reputable peer-reviewed venue"?

          • LPisGood 42 minutes ago
            Fraud requires intent to deceive _or_ reckless disregard, sometimes called, “conscious indifference”for the veracity of the statement asserted.
            • dataflow 34 minutes ago
              No. One single hallucinated citation on a document with you as an author is not evidence of your reckless disregard for anything. These exaggerations are crazy and you would absolutely deny such accusations if you missed your co-author's AI hallucinating a citation on your manuscript too. At best it would be careless, if you really relish extrapolating from one data point and smearing people's character based on that. Not reckless. It's quite literally the difference between going five miles per hour over the speed limit versus fifty.
              • algorithmsRcool 2 minutes ago
                Allowing hallucinated content or citations into your work is an act of carelessness and disregard for the time of people that are going to read your paper and it should be policed as such.

                And flatly, if a person can't be bothered to check their damn work before uploading it, why should anyone else invest their time in reading it seriously?

              • toast0 22 minutes ago
                If your co-author inserted the fradulent reference, I agree that you may not have committed fraud. But your co-author did, and you didn't check their work.

                You didn't pick your co-author very well, but arXiv lacks investigative powers to determine which co-author did the bad, so they all get the consequence.

          • zeusdclxvi 50 minutes ago
            If you are using AI-hallucinated references in scientific papers then there is some obvious intent to deceive there
          • NiloCK 43 minutes ago
            > No, it is emphatically not. D Fraud requires intent to deceive.

            I'm about as pro AI-as-a-research--and-writing-assistant and anti AI-witchhunt as they come, but I simply cannot parse what I've quoted here.

            Posting slop to arxiv is blatant deception. Posting an article is an attestation that the article is a genuine engagement with the literature. If you're posting things to arxiv that are not sincere engagements with the literature, you are attempting to deceive.

            • protocolture 1 minute ago
              >I'm about as pro AI-as-a-research--and-writing-assistant and anti AI-witchhunt as they come, but I simply cannot parse what I've quoted here.

              Ditto. And its only 1 year. Like its about the most reasonable thing they could have done.

          • toast0 27 minutes ago
            > followed by the requirement that subsequent arXiv submissions must first be accepted at a reputable peer-reviewed venue"?

            This part seemed reasonable too. I'm not in academia, but my understanding is most people writing papers intend for them to be accepted by reputable peer-reviewed venues, but post to arXiv because those venues don't always allow for simple distribution.

            If your papers aren't going to be accepted at reputable venues and you posted slop to arXiv before (and they noticed it!), seems reasonable that they only want reputable stuff from you in the future?

        • blazespin 1 hour ago
          it's very silly, but not a big deal. Arxiv is becoming irrelevant these days anyways.

          In fact would be better if they just banned AI, so we could just get off the luddite platforms.

          Automated research is the future, end of story. And really it couldn't have come out at a better time, given the increasingly diminishing returns on human powered research.

          • andrepd 1 hour ago
            Poe's law striking hard.
      • wrs 1 hour ago
        A "mistake" would be a typo in a real citation. A hallucinated citation is evidence of just plain laziness and negligence, which taints the entire submission.
        • dataflow 1 hour ago
          No it is not. Seriously. All you need for this to happen is for your lab partner to ask AI to add a missing citation that they are already familiar with at the last minute before a midnight submission deadline, and for the AI to hallucinate something else, and for them to honestly miss this. It does not even imply any involvement on your part, let alone that either of you were lazy or negligent on the actual research or substance of the paper. The lack of any sympathy or imagination here is astounding.
          • applfanboysbgon 3 minutes ago
            Your constructed hypothetical makes it even worse. If there are 2+ people in this scenario who have good intentions, this should especially never happen. When you sign your name on a paper, you are nonetheless vouching for everything written in it, including the things you didn't personally write. You should absolutely be checking every single reference your co-author included and verifying that it says what your co-author claims it says. This is something you should have been doing completely independent of LLMs existing.
          • asdff 1 hour ago
            There are no deadlines for journal submissions. Even if you felt you were running close to your revisions being due, an email to an editor will probably fix this for you. And what you described is still negligent, not verifying the garbage output bot did not in fact output garbage.
          • bigstrat2003 59 minutes ago
            The lack of understanding that you are responsible for the content you create, no matter what tools you use, is what's astounding.
      • goolz 1 hour ago
        If you cannot be bothered to check your references when writing academic quality papers then you have no place writing them in the first place. The punishment is not chopping off a finger, it is a polite reminder to do the bare minimum.
        • slashdave 46 minutes ago
          Well, in the good old days, when we have refereed journals, it would be part of the publishing process.
      • ajkjk 1 hour ago
        It's not the kind of mistake that is possible unless you're engaging in fraud anyway.
        • dataflow 1 hour ago
          > It's not the kind of mistake that is possible unless you're engaging in fraud anyway.

          Seriously? You can't fathom an honest researcher asking for AI to find a citation they know exists, and the AI inserting or modifying a citation incorrectly without them realizing?

          If you find evidence of fraud by all means lay down the hammer. Using a single hallucinated citation like it's some kind of ironclad proxy just because you think they must be committing fraud is insane.

          • mixtureoftakes 1 hour ago
            if you're not checking citations in the paper youre publishing AND trusting a non SOTA, hallucination prone ai model to come up with sources for it, its probably for the best of everyone that this paper isn't published.

            yes there will be rare exceptions but in general i feel like this is a really good addition.

          • asdff 1 hour ago
            Why would you ask the ai to find a citation you know exists? Just reach for that citation.
          • BonoboIO 35 minutes ago
            Verifying that the reference you cite actually exists is the absolute minimum standard for academic work. It is not optional, not something to skip because of a deadline, and not something to outsource blindly to hallucination-prone AI.

            If someone cannot meet that bar, they have no business publishing research papers. I have written academic papers myself, and I find it astonishing that people are trying to justify this as if it were some understandable workflow mistake. At that point it is simply slop with academic formatting. Post it on a blog or somewhere else, but do not put it into the scientific record.

            A one-year ban is not a lifetime ban. Maybe six months would also have been enough, but the author can use that time to think about whether they should verify references next time — and to manually check every other citation.

          • andrepd 57 minutes ago
            If you are citing a work you paste a citation to that work. If you are bullshitting you ask an AI to come up with a citation. Jesus, there is zero reason to ever "generate a citation" if you are not, in fact, commiting fraud.
            • galaxyLogic 34 minutes ago
              I much agree. But I wonder shouldn't the citations all be hyperlinks and thus easy to verify?
          • BonoboIO 39 minutes ago
            I would not necessarily go as far as calling it fraud, but if you cannot even verify that the reference you are citing actually exists, you are not ready to publish research papers.

            Deadlines are not an excuse here. Checking whether a cited book, paper, or passage exists is the absolute minimum standard for scientific work, not an optional extra. I have written academic papers myself, and I find it astonishing that people are trying to justify this as if it were some understandable workflow mistake. At that point it is simply slop with academic formatting.

            A one-year ban is not a lifetime ban. Maybe six months would also have been enough, but the point is that the author gets time to think about whether they should verify references next time. They can also use that time to manually check every other citation.

      • patcon 1 hour ago
        A citation is where you derived knowledge... If you haven't checked it and you are submitting something that should represent a ton of labour (and which will consume labour to review), you don't understand what you're doing. It is not just crossing T's and dotting I'd.

        Your being set behind is less important than the fact that your publishing is setting everyone else behind.

        Such a banned person is being helped to "step out of the way", and someone more competent will assuredly step forward to consume the limited maintenance labour more thoughtfully

        • dataflow 1 hour ago
          > Your being set behind is less important than the fact that your publishing is setting everyone else behind

          One hallucinated citation does not in any way imply anyone is being left behind. All it means nobody is checked that particular line of the manuscript after it was written. The rest of the paper could still be solid and treated accordingly. If you find evidence of the contrary, of course treat it accordingly, but this is so obviously not that.

          • dumpsterdiver 40 minutes ago
            > One hallucinated citation does not in any way imply anyone is being left behind.

            The parent said “setting” others behind, which refers to lost time.

            Being “left” behind implies a degraded trajectory, which is defined not by time lost, but by the final destination.

            Different but related things (e.g. lost time can indeed affect your final destination, for instance, after growing old correcting a scourge of hallucinated citations - which should have been table stakes all along).

          • some_furry 1 hour ago
            No. It's fraud.
          • andrepd 58 minutes ago
            You clearly misunderstand. You cite a work in your paper because you have read that work, and build upon it or want to refer to it to back up a specific claim. Generating references is fraud period, because you are implying that you have read a work when in fact you just asked an AI "please insert some reference-shaped text here" to make it look like a proper paper. It is sadly not a necessary, but certainly a VERY sufficient, reason to conclude a paper is fraudulent.
      • conartist6 1 hour ago
        Yes, it is fraud
      • themafia 8 minutes ago
        > There's no need to permanently cripple someone's ability to progress their life or contribute to humanity

        I don't think you need to publish on arXive to contribute meaningfully to humanity.

        > That's punitive instead of rehabilitative.

        Unfortunately science is competitive. Yours is a race to the bottom where the people who can afford the most expensive models and who are least concerned with the truth can publish the most papers and benefit financially and professionally by doing so. This is not a zero sum arena, grant money and opportunities will possibly be rewarded to them, and not to another team who is producing more careful and genuine output.

      • Loughla 1 hour ago
        Don't use AI? Problem solved?
      • mianos 1 hour ago
        You are being ironic right?
  • imenani 3 hours ago
    • JumpCrisscross 2 hours ago
      > Our Code of Conduct states that by signing your name as an author of a paper, each author takes full responsibility for all its contents, irrespective of how the contents were generated (Dieterrich, T. G.)
  • rgmerk 7 minutes ago
    Good.

    If it’s not worth your time to check the output of your LLM carefully, it’s not worth my time to read it.

  • noobermin 30 minutes ago
    Seeing the usual LLM hypers angry replying to this on twitter is such a tell. Just like the comments on the LLM poisoning articles, some people just can't accept that some people don't like LLMs and get upset when you put any amount of hindrance to their rapid acceptance.
  • MinimalAction 1 hour ago
    There needs be to a careful vetting before such adverse actions. If somebody includes a name and pushed it without express permission, does everyone get the ban? I agree that implemented the right way, this is good.
  • jszymborski 23 minutes ago
    Should be more harsh in my opinion.
  • bigfishrunning 3 hours ago
    Good; academic literature is in crisis because of all of the slop. Forcing some consequences on easily-detectable hallucinations can only be a good thing
    • tengwar2 2 hours ago
      It's not just AI, though. I did a doctorate in physics about 40 years back, and bad references were a problem back then.
      • dualvariable 1 hour ago
        Doesn't matter if it is AI hallucinations or entirely human scientific fraud, the problem is the same, and the solution works fine for both cases.

        If you can't validate that your bibliography is full of real articles, you shouldn't get published.

        LLMs have just poured gasoline on the fire.

      • lucb1e 1 hour ago
        In what way? Surely something like the source not quite saying what was cited, or mixing up citations, rather than inventing them outright?
        • tengwar2 1 minute ago
          In physics, references which just didn't exist. That could be that the author made it up, but often it's because they transcribed the reference from another paper without reading it - we know because a few people have deliberately introduced fake references to trace how far they would go. The reasons are not the same as for AI, but the problem they produce is the same.

          References which don't accurately reflect the quoted material seem more common in other subjects.

        • miki123211 1 hour ago
          That, and mixing reference details from multiple sources and messing it up.

          Let's say you read a paper on Arxiv but cite the version that was submitted to a journal or conference, without realizing that the authors made changes to the version they submitted and forgot to upload them to Arxiv.

      • wrs 1 hour ago
        "Bad", like, you literally just made them up? I hope that would have been a problem.
      • add-sub-mul-div 1 hour ago
        Yes and ffs arrows kill people too but we don't bring that up every time we talk about what to do with guns.
      • asdff 1 hour ago
        Imagine how bad they are now then.
  • squirrelon 1 hour ago
    Had a colleague submit a paper with literal AI slop left in the text, got hit with a nasty revision request. Check your drafts before you submit, people. The reviewers will find it.
    • miki123211 1 hour ago
      Also check your LaTeX comments, Arxiv makes those publicly visible!!!

      I'm a screen reader user and usually read papers as raw TeX. I've seen everything: slurs, demeaning comments towards reviewers and professors, admissions of fraud, instructions to coauthors to commit further fraud before paper submission to mask the earlier fraud... it's all there. There's far less of it than I would think, definitely <1% of papers, but it's there.

      I think it would be useful to run an LLM anti-fraud pass on the TeX source of all new arxiv papers. It wouldn't catch everything, but it would catch some of the dumbest fraudsters.

      On the positive side, you can also find stronger claims that didn't survive review, additional explanations that didn't make the cut due to the conference's page limit, as well as experimental results that the authors felt weren't really worth including. Those need to be approached with an abundance of caution, but are genuinely useful sometimes.

    • SchemaLoad 1 hour ago
      Sad the suggestion here is to just disguise the slop to make it harder for reviewers to spot rather than not submitting slop to begin with.
  • random3 3 hours ago
    It seems a good idea to ban cheating, but how hard is it, especially in new reasoning/agents contexts to validate references?

    The deeper question is whether legitimate AI generated results are allowed or not? Test - In the extreme - think proof of Riemann Hypothesis autonomously generated (end to end) formally proven - is it allowed or not?

    • Ifkaluva 2 hours ago
      This is not about banning cheating, it’s about banning inaccurate information.
    • Retric 2 hours ago
      You don’t need to solve everything, catching a few thousand non existent citations with such a policy is on its own a net benefit.
    • pointlessone 2 hours ago
      It is allowed as long as it’s verified.

      The thread specifically points out that if authors can’t be arsed to simply proofread their text the rest can not be trusted either.

      It’s a simple heuristic against low quality submissions, not an anti-ai measure.

    • beloch 1 hour ago
      There already exists multiple tools for automatically verifying references. This measure will likely only filter out the laziest and most incompetent of AI slop submissions. It's a very modest raising of the bar, but comes at zero cost to honest researchers.

      I expect arXiv will still have problems with slop submissions but, at least, their references should actually exist going forward.

    • lionkor 2 hours ago
      If you use AI correctly, nobody should be able to tell that it was used at all.
    • llm_nerd 1 hour ago
      It isn't "cheating" they're concerned with, it's sloppiness. This dictum isn't some sort of AI ban, but instead simply that if there is evidence that it was so low effort that the work includes such blatant problems, it's just adding noise.
    • pinkmuffinere 2 hours ago
      > think proof of Riemann Hypothesis autonomously generated (end to end) formally proven - is it allowed or not?

      Sorry to be rude, but this seems like a dumb question. I want science to progress. A primary purpose of these journals is to progress science. A full proof of the Riemann Hypothesis progresses science. I don't care how it was produced, if Hitler is coauthor, etc, I just care that it is correct. Whether the authors should be rewarded for whatever methods they used can be a separate question.

      • kingstnap 1 hour ago
        Terence Tao had a nice talk from the Future of Mathematics conference posted yesterday [0] that shapes a lot of my own feelings on this matter.

        The short of it is he argues how first to correctness shouldn't be the only goal / isn't a great optimisation incentive. Presentation and digestibility of correct results is a missing 1/3 when you've finished generation and verification. I completely agree with him. You don't just need an AI generated proof of the Reimann Hypothesis. You would really like it to be intentional and structured for others to understand.

        A really beautiful quote I learned of in the talk is this:

        > "We are not trying to meet some abstract production quota of definitions, theorems, and proofs. The measure of our success is whether what we do enables people to understand and think more clearly and effectively about math." - William Thurston

        [0] https://www.youtube.com/watch?v=Uc2zt198U_U