4 ms·

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 thous

by aaplok 2y ago

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.

araes 2y ago
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?
- Yoric 2y ago
  Yeah, you're listing some of the reasons I left academia.
- tough 2y ago
  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!
  - araes 2y ago
    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 https://quillbot.com/paraphrasing-tool They're apparently very easily available.
    - tough 2y ago
      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

by aaplok 2y ago

araes 2y ago

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?

Yoric 2y ago
Yeah, you're listing some of the reasons I left academia.
tough 2y ago
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!
- araes 2y ago
  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 https://quillbot.com/paraphrasing-tool They're apparently very easily available.
  - tough 2y ago
    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