Three Acknowledged, One Silent: The Rebuttal Journey Continues

Navigating the ICML rebuttal process can be frustrating, especially when reviewers acknowledge your efforts without adjusting their scores.

3 min readMachine Learning

The peer review process is not functioning as a genuine mechanism for improvement when the outcome is predetermined and the rebuttal is a formality. Three of four reviewers have acknowledged the substantial work submitted, additional experiments, full proofs, direct responses to every raised point, and then simply chosen Option A without adjusting a single score. Their initial scores were already positive, so the exercise was never about whether the work could get better. It was about going through the motions.

This is what frustrates the researcher who shared this experience, and the frustration is justified. They did not submit a few clarifying sentences. They ran new experiments. They wrote proofs. They treated the reviewers' requests as legitimate paths to a stronger paper, and in return, they received the academic equivalent of a polite nod. The fourth reviewer, who gave a score of 3 and remains silent, only deepens the sense that the system rewards effort inconsistently, sometimes not at all.

Here is the practical reality for anyone navigating this process: you are not always being evaluated on the quality of your rebuttal. You are being evaluated on whether the reviewers feel like changing their minds, and that is a different thing entirely. Positive scores rarely move because moving them introduces risk. A reviewer who says "this is good" has little incentive to say "this is better" when the first statement already serves their purpose. The silent reviewer is not a mystery to solve; they are a variable you cannot control.

The takeaway is not that rebuttals are useless. It is that they are only as useful as the reviewers' willingness to engage with them honestly. If you receive three acknowledgements and no score changes, you have learned something about the process, not about your work. The same applies to the fourth silence: it tells you about the reviewer, not the validity of your proofs. Do not mistake politeness for progress. Do not mistake a response for a reconsideration. The system will let you run the experiments, write the proofs, and wait. The question is whether you will keep treating the process as a conversation when it has already told you it is a formality.

From Machine Learning

I've received 3 out of 4 acknowledgements, All of them basically are choosing Option A without changing their scores, because their initial scores were already positive. Meanwhile, the 4th reviewer had already given me a 3 and still hasn’t replied.

Read the original at Machine Learning