masyukunMay 15
Joseph Cox

ArXiv is going to ban researchers for a year if they are caught submitting AI slop. ArXiv told us it's a one strike rule: do it once, you're out.

https://www.404media.co/new-arxiv-rules-ai-generated-papers-ban/

ArXiv to Ban Researchers for a Year if They Submit AI Slop

The change comes as arXiv and others struggle to manage an influx of AI-generated materials masquerading as rigorous science.

404 Media
May 15 at 4:06pmWeb
Show threadFediThing May 15

@josephcox

As long as there's an appeals process for false positives, this is a great move!

Show threadOliver D. ReithmaierMay 15
@FediThing @josephcox a false positive being what exactly in this instance?
Show threadFediThing May 15

@odr_k4tana @josephcox

Someone who didn't use AI but was accused of using it 🙂

Show threadOliver D. ReithmaierMay 15
@FediThing @josephcox not how it works. Either your reference exists or it doesn't. If it doesn't, you either used AI for it, or you did an abysmal job curating your sources. Both cases are worth the punishment.
Show threadFediThing May 15

@odr_k4tana @josephcox

This is just meant to be a fallback in case whatever process is used to check whether references exist has failed to work properly.

I'm not defending AI or non-existent references in any way, they're horrifying to see in papers.

Show threadCoral (bleached era)May 15
@josephcox PLEASE. 10% odds if I click on something in the daily feed it's 135 pages long with no substance and badly broken tikiz
Show threadRue MohrMay 15
@josephcox the next step would be an article sighting ai articles
Show threadDavidMay 15

@josephcox 404media has paywalls. Here's the cited thread, without Xitter's login requirements either:

https://xcancel.com/tdietterich/status/2055000956144935055

Interestingly, the core is "If a submission contains incontrovertible evidence that the authors did not check the results of LLM generation," then the authors get booted. Subtler slop will slip through.

Thomas G. Dietterich (@tdietterich)

Attention @arxiv authors: 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. 1/

Nitter
Show threadJohn McLearMay 15
@josephcox I'm building tooling to help with these problems. Essentially empowering peer review when AI assistance is allowed but within constraints...
Show threadJoshua McNeillMay 15
@josephcox Interesting. I suspect that many fake AI papers would pass under the radar given their criteria being a very high bar, but that's certainly better than ignoring the issue.