OpenAI Engineer Says Labs View Conferences as Fraudulent
Keller Jordan states most big lab researchers read almost no papers from top AI venues.
Keller Jordan, a research engineer at OpenAI, posted that most big lab people now read almost zero papers and see ICLR, ICML, and NeurIPS as mainly full of overclaims and fraud, with occasional strong work. The statement prompted replies from academics and engineers including Jason Lee, Ravid Shwartz Ziv, Bharath Ramsundar, Dimitris Papailiopoulos, and area chairs who noted broken incentives, the value of ideas over numbers, and sadness over rising issues in the publication system. One reply suggested the trend might eventually reduce resume-driven work in academia.
PSA: Most biglab people now read almost zero papers and understand ICLR/ICML/NeurIPS to be mainly full of overclaims & fraud. (but there are a few diamonds in the rough of course)
看了篇ICLR2026的论文，怎么也想不到效果为什么那么好，正好代码开源了， 一看是用测试集标签选参数...这就是顶会吗:)
Reactions
34 postsKJ 
Keller Jordan@kellerjordan0^ This is obvious to everyone who has connections, but I think not necessarily obvious to the majority of students who aren’t as well-connected. Therefore it should be repeated every so often
- likes: 211
- replies: 4
- bookmarks: 14
- reposts: 3
KJ Keller Jordan@kellerjordan0Ok I do want to soften my statement a bit. I wouldn’t claim that the numerical majority of papers in these conferences are overclaiming or fraudulent. It’s more like, as a reader, if you select for the papers that seem most exciting, then those ones will be majority overclaims.
- likes: 241
- replies: 4
- bookmarks: 14
- reposts: 5
KJ Keller Jordan@kellerjordan0But to be honest I wouldn’t even really know anymore, because like I said, as a biglab person I hardly read papers now (*gets dunked on by a mob*)
- likes: 1
- replies: 0
- bookmarks: 0
- reposts: 0
WD will depue@willdepue@kellerjordan0 most papers feel like theyre just cargo culting research for hiring purposes, very disappointing
- likes: 89
- replies: 3
- bookmarks: 3
- reposts: 0
DP RS Ravid Shwartz Ziv@ziv_ravidMy take: No one takes the numbers in the papers as they are, but it's a great source for ideas that you can use and follow up on
- likes: 133
- replies: 4
- bookmarks: 23
- reposts: 4