The ICML 2026 rebuttal phase has become a pressure cooker, and the question of what reviewers are actually seeing in their batches is one that deserves a straight answer. Based on the discussion, the reality is that post-rebuttal scores are not a monolith. Some reviewers hold firm, others shift meaningfully, and the aggregate picture is far messier than any single thread can capture. That is not a failure of the process; it is the process, and pretending otherwise does no one any favors.
For authors refreshing their dashboards, the practical takeaway is this: a tracker like Paper Copilot gives you a useful temperature check, not a crystal ball. It aggregates self-reported scores from a self-selecting group, which means the distribution you see is a sample of the loudest and most anxious voices, not a random draw of the full reviewer pool. The tool reflects a slice of reality, but it is a slice that has been filtered through human anxiety, selective participation, and the natural skew of who bothers to log in and share. Treat it as a directional signal, not a verdict.
What this means for you is straightforward. If your score sits below the threshold you hoped for, the rebuttal was never about flipping every reviewer to a 7. It was about moving the needle on the ones who are genuinely undecided, the ones whose critiques were substantive rather than dismissive. A reviewer who raises a real methodological concern is not your enemy; they are the one giving you a chance to respond with precision. The authors who gain the most from this phase are those who read the reviews as a map of objections, not a list of grievances, and who use the rebuttal to clear up confusion rather than to argue for the sake of winning.
The deeper issue here is that the community has outsized faith in the precision of a process that is inherently human. Reviewers bring their own expertise, biases, and time constraints to every submission. A tracker cannot capture that texture, and neither can a Reddit thread. What you can do is use the data to calibrate your expectations, then focus your energy on the work itself. The score you see after rebuttal is a snapshot, not a sentence. The paper you submit next year will be judged on its own merits, and the tools you use to track it should inform your strategy, not define your worth.