ABCDE

[ABCDE @abcde6485]

@tsoding Only power glitches reboot mine

[Тsфdiиg @tsoding]

Linux has the highest uptimes among other OSes because Linux people are just afraid to reboot.

[ABCDE @abcde6485]

@lpachter They should rent the gpu n invest the savings in nvidia for the kid’s alpha school tuition.

[Bojan Tunguz @tunguz]

Mathematics has now become Math-as-a-Service: MaaS.

[Fuxin Li @fuxinli2]

As organizers of CVPR 2025, we were one of the first to introduce authorship limits with a limit of 25 per author. In light of the excessive amount of submissions of ICLR 2027, many here were proposing stricter authorship limits. I want to offer my 2 cents: I don't think authorship limits are useful to reduce the number of submissions. Authorship limit is a tool to limit people putting their names into too many papers that they may have not contributed enough, which has served its role already with the current limits.

In CVPR 25, more than 90% of the first-authors only have 1 first-authored submission, and more than 99% of the first authors have at most 2 first-authored submissions. With this in mind, rate limits for submissions are likely not going to help reducing the number of submissions much, if at all. The problem is rather that everyone who wants to be an author nowadays can be one with AI-generated papers, not necessarily a few authors spamming the system, and I would assume it is more likely the case right now, although the proportions may have shifted a bit.

Area chair desk rejections (maybe with automatic flagging before manual checking) before the paper going through the reviewing process might be a viable approach right now, and I think it's worth testing that in upcoming conferences. But submission fees and authorship limits are not necessarily going to help with the problem because the problem is too many people submitting a few papers, rather than the other way around.

[Andrew Gordon Wilson @andrewgwils]

I'm a Program Chair for ICML 2027. We'd love any creative suggestions for a great conference! How to manage the explosion of AI slop and insane submission growth? How to manage reviewing? How to incentivize high quality creative work? How to reduce bureaucracy and overhead?

[KC @ScarletKc]

GPT-6 Astra 用下来，我真的开始怀疑 OpenAI 的方向走错了。

它很擅长目标明确、有固定正确答案的任务。但碰到开放、发散、没有标准答案的事情，就特别别扭。自由探索一个想法、写点有意思的东西，或者面对模糊的问题自己判断值得往哪走，这些才是我觉得它明显欠缺的能力。从 GPT-5 开始，我就有这种感觉。

很多任务开始时，连最终想要��么都没确定，需要在探索过程中逐渐发现。GPT 却总像在等你把题目出完整，把验收标准列清楚。可如果这些都得我先想好，最需要智能的那部分工作，我已经自己做完了。

我怀疑这也解释了它为什么那么爱堆防御性代码、写一大堆测试。测试能给出明确的通过或失败，它很容易一直围着这些可验证的结果打转。至于方向有没有价值、有没有更好的可能，就需要另一种判断力了。

Claude Fable 在这类任务上给我的感觉就好得多。它会从理解你想做的事情出发，顺着不完整的想法继续展开，给你一些原本没想到的可能性。即使没有固定答案，也能把事情往前推进。

这让我怀疑 OpenAI 是不是太执着于能判对错、能计分的能力了。真实世界有太多任务根本没有标准答案。在这种自由探索和开放任务上，我觉得 Anthropic 才是真正领先的，光看表面跑分看不出来。

[Douglas Gray @de3ug]

Since everyone seems to be having a panic attack about ICLR hitting 60k abstracts submitted, let me add one more crazy suggestion to the pile.

Promotion/relegation.

Your paper has to start at a regional conference. Only if it’s accepted there can you submit to a top tier venue.

Rejected papers need to be accepted at a workshop before they can be resubmitted anywhere.

2000+ papers is just too many for one venue, and nobody seems to care enough about 2nd tier or local conferences. We should change that.

[BURKOV @burkov]

All these brute force proofs remind me of an infinite number of monkeys using an infinite number of typewriters, which would type the whole Shakespeare collection with probability one.

[Ashutosh Kumar @ashu_1069]

Why is everyone surprised by the ICLR submission numbers? We wanted to democratize technology and knowledge in the first place, right?

Literature reviews are now done with ease, condensed ideas get out faster, and what’s that called again? Oh yes, “autoresearch.” I’ve got the agent high on that. Here’s my paper. Now I'm ready to review 4 papers with the "autoreview" readme file. It's also a loop, I believe. It happened with code, and now it's time for academic research. This is what we're working towards with AI development, I suppose, getting everything at 2x or 5x. Why this fuss over the output of that progress? Automate everything. LFG

[XY Han @XYHan_]

there are neurips workshops that got more than 500 this year

[Anshul Kundaje @anshulkundaje]

If it's isn't obvious already, it is the end of papers as a measure of productivity, expertise or accomplishment. The odds of publishing in "glam" conferences & journals also very quickly approaches random & the time for gate keeping review increasing exponentially. 1/

[Anshul Kundaje @anshulkundaje]

And please let's not try create worse systems that try to mimic the past by using AI. New systems need to realign incentives as well. That goal of doing science should be to make progress on solving problems, not decorating CVs with number of papers & citations & glam venues. 5/

[ABCDE @abcde6485]

I hate to say it but still the best orchestrator/communicator is opus 4.7. Astra is a good mindless employee at best. You even need to ask it to "think" so disappointing. Of course, opus 5 is worse than the two and Fable is somewhere else on the pareto frontier

[Sariel Har-Peled 🆗 @sarielhp]

They are so behind the times - there is no problem making a video using AI. The only solution to ensure real human submission, by the real authors, is to demand submissions printed using as ink the blood of the authors and verify it using DNA sequencing...

[T. Y. Hayashi @lianda_edu]

This is basically discriminating against non-native English scholars. Back then, their writing would just be called poor English. Now it’ll be called AI-generated.

[𝖬𝖺𝗁𝖽𝗂 𝖢𝗁. (bluesky:@mahdi.ch) @mahdi_tcs_]

Anybody who's tried AI in math knows that, unless you're an AI company, you currently can't just push a button and have AI solve sufficiently advanced problems. You have to join the AI in all kinds of rabbit holes and navigate the problem, much like collaborating with a human.

[ABCDE @abcde6485]

What does a paper signal? Truth or the ability of the author? If it adds to knowledge by $1mil spent on token, is it bad? For ability just do oral n chalk talk. To handle the volume do 1st pass ai review. Problem solved.

[Transactions on Machine Learning Research @TmlrOrg]

TMLR has faced a deluge of submissions, necessitating stricter desk rejection policies due to limited reviewer capacity

Co-EiC Nihar Shah reached out to authors of 10 papers slated for desk reject. Could they answer questions about their *own* submission?
https://t.co/vhX5w6gcx2

[AISTATS Conference @aistats_conf]

#AISTATS 2027 Call for Paper is out, and submissions are open!

This year's edition is taking place in Montreal on May 3-6th, 2027 🇨🇦

https://t.co/ey4xEbAuHr
New this year: AI review, and more!
Abstract deadline: September 29, 2026
Paper deadline: October 6, 2026

[Quentin Berthet @qberthet]

Submit to #AISTATS 2027!

co-chaired this year by Aymeric Dieuleveut and myself

AI reviews this year!