Jun Li
š¤ SpeakerAppearances Over Time
Podcast Appearances
Yes.
So this is this is sort of been bubbling for a little while now because both sides of the research process has sort of just been inundated with all kinds of.
research papers, non-peer reviewed, but hasn't even... So normally there's a multi-step process, right?
When you're presenting your paper, you release it to journals or certain publications or organizations that will take the time to review it.
check it for any obvious big errors and say, yay or nay, whether it should move on.
This research is done well enough to be considered for the next step and so on and so forth.
And with the advent of AI being used in this research process, whether being the subject of the research or in the writing process or the development process,
the that pipeline of how it has been done has been overloaded both from how many uh how relatively easy it is to get a research paper out there and then of course the reviewers themselves there's not a ton of uh resources for uh doing the proper um
proper process.
And because of that, a lot of times reviewers will then help themselves by having AI also review some of these papers.
So the pipeline has been fairly clogged.
Pargives said its computer science category will only accept review slash survey articles and position papers that have already passed peer review at a journal or conference.
A preprint is a paper shared before formal peer review.
The update requires authors to provide documentation of acceptance.
Archive framed this as a return to policy, noting those content types were never core and were previously allowed at moderator discretion.
The change follows a surge of low-quality submissions that moderators say include AI-generated slop.
Independent coverage reports the shift sparked debate.
Some researchers warn that it may limit early synthesis work and first-time authors, while others welcome relief for moderators and readers.
Differentiator here is targeting specific paper types most vulnerable to low-effort AI output.
while leaving research articles otherwise unaffected.