Exactly.
I think, we are over-using the term "AI Slop". If AI writes something, and a human checks n verifies or modifies it to make it correct, it is 100% fine with me.
The problem is if the AI content is not verfied or updated properly before submission/use, that's "AI Slop".
I fear research presentations turning into AI witch-hunts. The ultimate question for science is whether the claim is true, not how the answer was produced.
@doubleunplussed@harryjwang So, now, AI detectors should dictate how humans should write? or, our writing style?
We better change our scientific name from "Homo Sapiens" to "Homo AIphobic" from now on too? 🤣
This year, the NeurIPS 2026 Position Paper Track made the decision to require that all papers be substantially human-written, with AI used for only copy-editing or similar peripheral changes to the main text!
For more details, please check our blogpost: https://t.co/wrWuMQJwrx
How are best papers distributed across ML's top venues?
I mapped every best paper (including runners-up) at NeurIPS, ICML & ICLR, 2016–2024, to its region and affiliation. Main results below:
The US alone = 65%.
Industry is everywhere, like Google (14.2%), Meta(4.4%), Microsoft(3.6%), totalling 32.83%, more than half of the US figure (65%).
@DobrikG Nice point!
I feel it is kinda reciprocal too, in a sense. If I feel the authors tried their best and actually worked, I always try to engage; and my reviews sound like discussion then.
Otherwise, I usually point out errors, methodological gaps and writing suggestions.
I'm very disappointed with my NeurIPS review batch. People are literally submitting anything! Class projects, vibe coded things, anything.
Only 1 out of 4 papers used 9 pages fully. Among them, 1 paper has with only 4.5 pages! Figures, tables and depth: very bad.
@j_dekoninck I'm sorry, but it is never an insta-reject. It also doesn’t decrease model performances in all cases. And, I think, there are good logics on both sides (keeping zero or not).
Please have an open mind while reviewing :)
@HildeKuehne Thanks a lot!!!
In terms of networking for grad schools (or, as you said previously, personal recommendations matters more now), do you have any suggestions or tips? Especially considering my weird background problem?
@HildeKuehne Last part is very true. :3
My bachelors is in Industrial Engineering from an unknown low-ranked uni.; and I work on GenAI and healthcare applications.
So, the admission season went very very very bad, despite having good research experience.
Any tips? :)
Reviewers & ACs for #ICML2026 have been recognized for their service!
- Reviewers: 4439 Gold (free registration), 4437 Silver. 17749 total reviewers were assigned >= 1 paper
- ACs: 1647 receive free registration, out of 1691 who were assigned >= 1 paper
TY for your hard work!
@HildeKuehne Yes, I agree.
The problem is, in the research/publication world, we always assumed everyone will act fairly and properly (authors, reviewers, ACs/Editors). It was a "goodwill"-based system, with no other incentive/punishment. And, now, we're realizing that it’s not enough anymore
@albertlawman What if the idea is AI slop? Also, clarity and presentation is always included as a metric everywhere, along with novelty ('idea'), soundness and rigor.
You can have a great idea, but if you can't make people understand, its useless.