The rebuttal process

Most of my papers have been borderline submissions, because I like to write papers that are slightly outside the Overton window. One consequence is that writing rebuttals is a very important part of my publishing process.
I mostly publish at machine learning conferences, but I suspect these ideas generalize. Here’s some advice and my process:
Copy the reviews into a document that is shareable and supports comments, like Google Docs or a document on Althea. Share this with my co-authors.
Read the reviews without taking notes. When this is done, go on a walk (or run) to let go of my emotions.
Triage the stated weaknesses - by taking a linear combination of:
- (a) how likely Reviewer 2 is to increase their score,
- (b) how screwed we are if Reviewer 2 does not increase their score, and
- (c) how much work it takes to get Reviewer 2 to increase their score.
Work with my co-authors to resolve the most important weaknesses via further experimental work.
After time has passed and emotions have cooled, write a graceful reply to the kind volunteer academic who has spent their time with my work.
Now, given my ICLR reviews, I'm going to take a walk and try to find gratitude and understanding.
First posted on LinkedIn, November 6, 2026.


