@xwang_lk Seems to be common at all conferences now. Here's my group's 777 reject at COLM

@fredsala Insane. Sorry to hear it.

@xwang_lk I really hope not! The biggest problem with this would be that it would discourage people from raising scores in the interactive review process in the future. The new discussion format was a big improvement this year, I hope they don't destroy it.

@xwang_lk The system is tragically broken.

@xwang_lk Last two years I had NeurIPS papers rejected at: 87765 and 7666. Decided to never submit again. ML conferences not worth it anymore. Solve hard problems and build software people can use!

@xwang_lk I had a ridiculous experience at NeurIPS this year. Score reduced to strongly reject just for not citing a June paper. LLM reviews citing non-existing works. It killed any prestige publishing had for me, and I’m leaning toward only using arXiv from now on.

@xwang_lk Academia is a broken system with broken incentives.

@xwang_lk Reviews don’t matter. Reviewers don’t matter. Very unfortunate.

@xwang_lk This is really not a great thing. I had a Neurips paper last year with 77766

@xwang_lk The system is rigged. When you prioritize metrics over merit, genuine innovation suffers. Just focus on solving real problems.

@xwang_lk Is it allowed for ACs to control acceptance rates like that? What’s official policy?

@xwang_lk I'd say when all reviewers say yes to a paper, it should be accepted, save the tough choices for oral/spotlight slots, and if you overrule, it'd be nice to give a brief public reason!

@xwang_lk Rejecting because you used decimal instead of % is insane

@xwang_lk Ouch :( I've had that experience too where unanimous accept papers were rejected with weak arguments for overturning the consensus. In general I think unanimous accepts should be accepted unless there are clear crucial errors that the reviewers missed.

@xwang_lk this already happened to me 2 yeard ago for neurips🥹

@xwang_lk The scores are not even fully open to authors yet… 😅

@xwang_lk Could not agree more. We can shorten the presentation time length for some posters to include more papers for the conference.

@xwang_lk How can all reviewers accept be rejected in the end. Makes no sense 😕

@xwang_lk It's very sad to know

@xwang_lk Yeah, I feel you, I think the meta review should be more transparent though. I get why they're doing it, it's a tough balancing act, especially if you have a ton of papers but I agree with you that the whole thing can be done better. I think it also reflects the world we live in

@xwang_lk Really.. where did you get this? I hope that it is not happening unless there is a flaw in the paper that the reviewer missed..

@xwang_lk This is really crazy

@xwang_lk Seems very common in many conference. Same thing happened with us in ICCV.

@xwang_lk As usual, never understood this, like cmon guys just give them a space, that’s not this hard. Sad academic reality

@xwang_lk Is there an understanding of what 5444 is in the old scale? Like 7666? These ones were going through most of the time.

@xwang_lk gulp

@xwang_lk the whole system feels arbitrary at this point. when all reviews are positive, just accept the paper already, right? fairness should be the standard, not these acceptance games.

@xwang_lk What is this post about?