Hacker News Reader

Comment by mvdtnz | Hacker News Reader

From the research paper that the article is discussing.

https://www.tandfonline.com/doi/full/10.1080/08989621.2024.2...

> Namely, rates on T2 of up to 212 publications per year, and 5792 new coauthors per year, compared to up to 28.3 publications per year and 173 new coauthors per year on NL. This is despite that the NL list is weighted with more senior researchers, often at the pinnacle of their careers – and hence – likely to have higher rates.

View on HN

That is 212 papers, not 5000. 5000 is the number of new authors of those 200 papers.

Looks like the number is off, but is their point any different with 212 papers? I still think that it would be impossible to give any meaningful support to 212 papers in a year. That's a new paper every 1.72 days.

Those 200 papers correspond to 4 papers per week, or more than 1 paper every other day for the whole year, weekends and holidays included. Coauthored with thousands of people they'd never published with before. It's a rather incredible rate.

And that's before you have to try and actually write for grant funding, perform any actual substantial research (that took more than 1.5 days), serve as an editor or reviewer to do any of that "peer reviewed" part, go to a conference on your subject, deal with students, teach, or really anything else with academia.

Writing for grant funding alone has already been a big complaint about how long and difficult it can be with limited success rates in a lot of fields. Time constraints of being a reviewer and how much of your own academic time you have to give up has been a major complaint. They're all time consuming. And then a paper / 1.5 days.

Really, from having been in academia for a little while, putting out something well thought out once a month seems rather fast.

Edit: It would interesting as follow-on research to look at the specifics of suspicious cases. What kinds of methods are being used to "game the system." What kinds of exploits are being employed? Are there common patterns among the suspicious crowd?

Why publish 1x when you can publish the same 10x? Copy and paste data from other people's work? "Lorem ipsum" papers that nobody reads? Extensive citation circles where they all cite each other 50x? Graphs / tables that just look like gibberish data? Very obvious LLM use that has many of the common hallmarks and speaking patterns?

Yeah, you're listing some of the reasons I left academia.

im not a scientist but usually browse arXiv preprints, and it was interesting to see a feature that calls out how much similar text to past papers a submission has!

Did not realize that arXiv had that type of feature. May not actually be well known to many others on HN or arXiv. Thanks for noting the feature was even available.

Do you happen to know what they're using for the "similarity?" Might be it's own research subject in and of itself. just identifying similarity and paraphrasing substitution that's fundamentally the same paper. We changed a bunch of words with a paraphrasing tool. Used only as example cause it's the top search result for "paraphrasing text": https://quillbot.com/paraphrasing-tool They're apparently very easily available.

I've no idea and I cannot find any example now, If I recall correctly it was something along the lines of a custom -arXiv note: this paper has 56% text similarity with this other submission.

I checked the similar one, and was the same authors but different titles, gussued it was a case of a draft not getting a revision properly, but didnt double check for the author email beyond the name (it could also be someone impersonating the real authors with fake credentials, i guess)

but i think it was a revision!

if i find it again ill comment back here

Reply on news.ycombinator.com