#ICLR2027 authors: if you have solid evidence that a reviewer is nitpicking, flag it to your AC. In rebuttal, be specific: quote the exact claim, then rebut it: "This is standard in my area, see [refs], pp. 12–14." "Feels AI-written" or "here is an AI score" is not enough.
@ManishGuptaMG1 Thank you @ManishGuptaMG1 for visiting @iitdelhi and engaging so patiently with our students. This was much needed to get them to think beyond short-term compensation objectives, and emphasizing the value of a PhD, deep skills, and fulfilling work.
ScAI is heading to #NeurIPS26 with 15 papers
Dialogue Systems, Reinforcement Learning, LLMs, graph ML, Materials, Scientific discovery, Recommendation System, AI for Science & Healthcare a pretty wide spread this year.
Congrats to all the authors.
Guaranteed registration for just one author per paper at @NeurIPSConf is regressive. Academic conferences are primarily meant for discussing research, not serving as an open-for-all AI trade show. The shift had been ongoing, but now it’s official.
@gmishne@iclr_conf I wonder if the biddable abstracts are updated live. Specifically, If I withdraw an abstract now, can someone still bid for it? @iclr_conf
@mariyaivasileva The conversion from abstract to submission would be much lower this year since both NeurIPS and AAAI first-phase decisions fall in between the abstract and full submission deadline. That was not the case last year.
@PaulGavrikov@andrewgwils No, it means work on the paper and submit to the next conference in line, which would be something like NeurIPS.
As things stand, each subsequent conference would continue to grow larger since all rejects from one conference goes to the next one in cycle.
@danish037 I believe most people do care about their reputation. The problem is reputation remains unaffected even when their papers get desk-rejected due to these decisions remaining behind firewalls. We need to make acceptance rates and desk-rejects per author public in openreview profile
@DrDatta_AIIMS Same for me. Didn't approve based my webpage (https://t.co/Lf1dJQ53z5). Not sure if there is a criteria or their verification agent is poorly optimized.
@abcde6485 I haven't computed that. In general >4 means accept w.h.p unless there are hallucinations. For an avg. of 4, the median and reviewer discussion matters. Accepts at <4 are cases where the AC overrruled the concerns of the negative reviewers. 4 out of 26 papers are in this category
Almost done with my NeurIPS pile. Some stats:
* No. of submissions: 116
* 40 withdrawn
* 26 accept recommendations
* Min(Avg. rating) | Accept: 3.33
* Max(Avg. rating)| Accept: 5
* No AR-based caps so far
Note: these decisions aren't final yet. PC-SAC calibration to happen next