Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

GPTZero finds 100 new hallucinations in NeurIPS 2025 accepted papers

gptzero.me

Post reply on HN
Print view View original

501–510 of 528 posts

Re: GPTZero finds 100 new hallucinations in NeurIPS 2025 accepted papers

#501

Post by j2kun »

Lerc wrote: post #448
j2kun wrote: post #440

Earlier quoted context omitted.

They counted multiple hallucinations in a single paper toward the 100, and explicitly call out one paper with 13 incorrect citations that are claimed (reasonably, IMO) to be hallucinated.

So you are saying their claim of > GPTZero's analysis 4841 papers accepted by NeurIPS 2025 show there are at least 100 with confirmed hallucinations Is not true. [Edit - that sounds a bit harsh making it seem like you are accusing them, it's more that this is a logical conclusion of your(imo reasonable) interpretation.

I think it is true and intentionally vague for marketing purposes. And FWIW, I support the effort writ large

Re: GPTZero finds 100 new hallucinations in NeurIPS 2025 accepted papers

#502

Post by j2kun »

bjourne wrote: post #486
j2kun wrote: post #430

Earlier quoted context omitted.

I'm not blaming anything on anything, because I did not (nor did the authors) confirm the cause of any of these errors. > I don't share your view that hallucinated citations are less damaging in background section. Who exactly is damaged in this particular instance?

Trust is damaged. I cannot verify that the evidence is correct only that the conclusions follow from the evidence. I have to rely on the authors to truthfully present their evidence. If they for whatever reason add hallucinated citations to their background that trust is 100% gone.

You are speaking in the abstract. Did you read this paper? I suspect you did not.

Re: GPTZero finds 100 new hallucinations in NeurIPS 2025 accepted papers

#503

Post by piyh »

neilv wrote: post #339
i_am_proteus wrote: post #268

Earlier quoted context omitted.

>this error does make me pause to wonder how much of the rest of the paper used AI assistance And this is what's operative here. The error spotted, the entire class of error spotted, is easily checked/verified by a non-domain expert. These are the errors we can confirm readily, with obvious and unmistakable signature of hallucination. If these are the only errors, we are not troubled. However: we do not know if these…

> If these are the only errors, we are not troubled. However: we do not know if these are the only errors, they are merely a signature that the paper was submitted without being thoroughly checked for hallucinations. They are a signature that some LLM was used to generate parts of the paper and the responsible authors used this LLM without care. I am troubled by people using an LLM at all to write academic research p…

>I am troubled by people using an LLM at all to write academic research papers.

I'm an outsider to the academic system. I have cool projects that I feel push some niche application to SOTA in my tiny little domain, which is publishable based on many of the papers I've read.

If I can build a system that does a thing, I can benchmark and prove it's better than previous papers, my main blocker is getting all my work and information into the "Arxiv PDF" format and tone. Seems like a good use of LLMs to me.

Re: GPTZero finds 100 new hallucinations in NeurIPS 2025 accepted papers

#504

Post by the__alchemist »

thomasahle wrote: post #444
neilv wrote: post #339

Earlier quoted context omitted.

> If these are the only errors, we are not troubled. However: we do not know if these are the only errors, they are merely a signature that the paper was submitted without being thoroughly checked for hallucinations. They are a signature that some LLM was used to generate parts of the paper and the responsible authors used this LLM without care. I am troubled by people using an LLM at all to write academic research p…

> And also plagiarism, when you claim authorship of it. I don't actually mind putting Claude as a co-author on my github commits. But for papers there are usually so many tools involved. It would be crowded to include each of Claude, Gemini, Codex, Mathematica, Grammarly, Translate etc. as co-authors, even though I used all of them for some parts. Maybe just having a "tools used" section could work?

I suspect the parent post was concerned about plagiarizing the author of training data; not software tools.

Re: GPTZero finds 100 new hallucinations in NeurIPS 2025 accepted papers

#505

Post by jama211 »

techpression wrote: post #436
jama211 wrote: post #409

Earlier quoted context omitted.

Yup, and no matter how flimsy an anti-ai article is, it will skyrocket to the top of HN because of it. It makes sense though, HN users are the most likely to feel threatened by LLMs, and therefore are more likely to be anxious about them. I don’t love ai either, but that’s the truth.

Strange, I find it quite the opposite, especially ”pro-ai” comments are often top of the list.

I think there’s a bit of both, with a valley in the middle

Re: GPTZero finds 100 new hallucinations in NeurIPS 2025 accepted papers

#506

Post by jama211 »

kortilla wrote: post #418
jama211 wrote: post #408

Earlier quoted context omitted.

Graphic design is a completely different kettle of fish. Comparing it to academic paper writing is disingenuous.

The thread is about not knowing anyone at all who thinks AI is plagiarizing.

Yes, plagiarising text based content. No one in this thread meant graphics.

Re: GPTZero finds 100 new hallucinations in NeurIPS 2025 accepted papers

#507

Post by aydyn »

tsimionescu wrote: post #499
aydyn wrote: post #495

Earlier quoted context omitted.

What do you mean how is that relevant? Its a vast majority opinion in society that using ai to help you write is fine. Calling it "plagiarism" is a tiny minority online opinion.

First of all, the very fact that companies need to encourage it shows that it is not already a majority opinion in society, it is a majority opinion among company management, which is often extremely unethical. Secondly, even if it is true that it is a majority opinion in society doesn't mean it's right. Society at large often misunderstands how technology works and what risks it brings and what are its inevitable do…

> Secondly, even if it is true that it is a majority opinion in society doesn't mean it's right. Society at large often misunderstands how technology works and what risks it brings and what are its inevitable downstream effects. It was a majority opinion in society for decades or centuries that smoking is neutral to your health - that doesn't mean they were right.

That its a majority opinion instead of a tiny minority opinion is a strong signal that its more likely to be correct. For example its a majority opinion that murder is bad; this has held true for millennia.

Heres a simpler explanation: toaster frickers tend to seek out other toaster frickers online in niche communities. Occams razor.

Re: GPTZero finds 100 new hallucinations in NeurIPS 2025 accepted papers

#508

Post by Proziam »

Der_Einzige wrote: post #329
Proziam wrote: post #15

Earlier quoted context omitted.

All 3 of these should be categorized as fraud, and punished criminally.

Only when we can arrest people who say dumb stuff on the internet too. Much like how trump and bubba (bill Clinton) should share a jail cell, those who pontificate about what they don’t know about (I.e non academics critiquing academia) can sit in the same jail cell as the supposed criminal academics. You gotta horse trade if you want to win. Take one for the team or get out of the way.

Non-academics can definitely offer valid critiques of academia.

You don't need to be in academia to understand that scientific progress depends on trust. If you don't trust the results people are publishing, you can't then build upon them. Reproducibility has been a known issue for a long time[0], and is widely agreed upon to be a 'crisis' by academics[1].

The advent of an easier way to publish correct-looking papers, or to plagiarize and synthesize other works without actually validating anything is only going to further diminish trust.

[0] https://www.nature.com/articles/533452a#citeas

[1] https://journals.plos.org/plosbiology/article?id=10.1371/jou...

Re: GPTZero finds 100 new hallucinations in NeurIPS 2025 accepted papers

#509

Post by Proziam »

freedomben wrote: post #106
Proziam wrote: post #31

Earlier quoted context omitted.

False equivalence. This isn't about "using AI" it's about having an AI pretend to do your job. What people are pissed about is the fact their tax dollars fund fake research. It's just fraud, pure and simple. And fraud should be punished brutally, especially in these cases, because the long tail of negative effects produces enormous damage.

I was originally thinking you were being way too harsh with your "punish criminally" take, but I must admit, you're winning me over. I think we would need to be careful to ensure we never (or realistically, very rarely) convict an innocent person, but this is in many cases outright theft/fraud when someone is making money or being "compensated" for producing work that is fraudulent. For people who think this is too h…

Thank you for the comment!

I think the negative reaction people have comes from fear of punishment for human error, but fraud (meaning the real legal term, not colloquially) requires knowledge and intent.

That legal standard means that the risk of ruinous consequences for a 'lazy kid' who took a foolish shortcut is very low. It also requires that a prosecutor look at the circumstances and come to the conclusion that they can meet this standard in a courtroom. The bar is pretty high.

That said, it's very important to note that fraud has a pretty high rearrest (not just did it, but got arrested for it) rate between 35-50%. So when it gets to the point that someone has taken that step, a slap on the wrist simply isn't going to work. Ultimately, when that happens every piece of work they've touched, and every piece of work that depended on their work, gets called into question. The dependency graph affected by a single fraudster can be enormous.

Re: GPTZero finds 100 new hallucinations in NeurIPS 2025 accepted papers

#510

Post by AntonioEritas »

"I got my €220 back (ouch that's a lot of money for this kind of service, thanks capitalism)"

220 is actually quite the deal. In fact, heavy usage means Anthropic loses money on you. Do you have any idea how much compute cost to offer these kind of services?

Post reply on HN