Looks like the AAAI-26 rollercoaster isn’t slowing down anytime soon.
[image: 1757459547907-685cb90de9969beg.jpeg]
Both the latest notices confirm what many of us have been feeling: the review timeline is slipping yet again. Originally, Phase 1 notifications were due Sept 8. Then the OpenReview crash pushed it to Sept 12. Now the official word is Sept 15.
August 4, 2025
Supplementary material and code due by 11:59 PM UTC-12
September 15, 2025
Notification of Phase 1 rejections
October 7–13, 2025
Author feedback window
November 8, 2025
Notification of final acceptance or rejection (Main Technical Track)
That three-day “extra wait” might not sound like much, but for authors it’s brutal. It means:
Suspended in mid-air: checking inbox + OpenReview every morning, still no word.
Rebuttal delayed: the author response window got pushed past the October holidays — which makes sense only if reviews themselves aren’t ready.
Compressed transfer time: folks planning to bounce to ICLR 2026 if rejected are losing precious prep days. With ICLR abstract due Sept 19 and full paper Sept 24, every delay cuts deep.
To add spice, the program chairs hinted the Phase 1 rejection rate could hit 50–67%. That means only ~1/3 of submissions will survive past the first cut. With nearly 29,000+ papers in the system — more than double last year — the scale is unprecedented .
The bigger picture
Emergency “last-minute reviewers” are being pulled in to cover gaps.
Other conferences are also bending: NeurIPS’s “dual-city” experiment saw accepted papers later force-rejected due to quota caps.
The pattern is clear: our current peer review model is hitting a breaking point. Technical crashes, reviewer overload, rebuttals turning into vent sessions — all signs of strain.
Open questions for us as a community
Do we just accept longer waits and higher rejection odds as the new normal?
Should AAAI (and other big A* conferences) move toward dynamic, rolling review models rather than single-shot deadlines?
Or do we need to rethink reciprocal review obligations more fundamentally — to balance load without roulette-style assignments?
For now, all we can do is hang tight until Sept 15 ( no more extensions). But honestly, given the trajectory, I wouldn’t be surprised if “Sept 15” becomes “Sept whenever.”
Anyone here already prepping ICLR as a fallback? Or are you holding out for the rebuttal round?
6
1
1
1
1
1
1
