NeurIPS 2026 Reviews Are Out Today (22 July, AoE) - Discussion Thread [D]

Opening

Reviews drop today. This thread is for reactions, celebrations, commiserations, and anything useful in between.

Celebrate Wins

First: if you got good reviews, say so. There’s a norm in these threads where only the bad news gets aired, and it skews everyone’s sense of what’s normal. Post your wins.

The Noise of the Review Process

Second, the thing worth repeating every cycle: the review process is noisy, and that noise is measured, not folklore. The NeurIPS consistency experiments (2014, repeated 2021) found that a large fraction of accepted papers would have been rejected by an independent second committee. Reviewer assignment, load, and luck of the draw account for a lot. A score is a weak signal about your work and a strong signal about the process that produced it. That cuts both ways.

It’s not a license to dismiss every criticism as noise - it’s a reason to weight reviews by the quality of the argument rather than the number attached to them. The reviewer who found a real hole in your evaluation did you a favor, even if the tone was rough. The one who clearly skimmed did not, regardless of the score.

What to Do Next

So: prioritize the reviews that make the paper better. Fix what’s fixable, contest what’s genuinely wrong, and concede the rest gracefully in the rebuttal.

Things Worth Discussing

• Reviews that caught something you’d missed
• Rebuttal strategy - what’s worth contesting vs. conceding, and when new experiments actually shift a score
• Patterns you’re seeing this cycle (missing baselines, compute comparisons, ablation depth, reproducibility asks)
• Framing a response when a reviewer has clearly misread the submission
• Backup plans: ICLR, AISTATS, workshops

Please paraphrase rather than paste review text, and no speculation about reviewer or AC identities.

A Final Word

To anyone who got bad news: this doesn’t define your research impact. Plenty of heavily-cited work took two or three cycles to land somewhere. Rejection is a scheduling problem.

How did everyone do?