Scar Literature from Reviewing CVPR (Continued) - On the Metaphysics of Rebuttal (Gemini 2.5 Pro Translated Version)

Seeing a few private messages asking me how to write a rebuttal, I might as well ride this wave of popularity. I don’t want to think too deeply, so I’ll just casually touch on a few points.

Note that my use of “metaphysics” here is not the metaphysics of Western philosophy, but rather in the sense of “viewing the world from an isolated, static, and one-sided perspective, believing all things are isolated and unchanging.”

1. AC’s Workload: An Area Chair (AC) handles 20-30 papers (I heard it was up to 35 this year). Their goal, according to the generally understood acceptance rate, is to “reject most papers.” No matter how capable an AC is, it’s unlikely they can give serious attention to such a large number of papers and reviews, especially considering most ACs are “busy people” in real life. So, if a paper’s average score is low from the start—though it’s hard to say—there’s really no need to “waste each other’s time,” because the AC won’t even “guide the discussion,” and reviewers genuinely won’t bother to meticulously change their scores. The effort you put into a rebuttal is essentially betting that “the AC is not just a score averager,” which is an extremely low-probability event. Although I’ve encountered very serious and responsible ACs, the proportion of excellent ACs in a (crappy) conference like CVPR is relatively low.
2. Reviewer’s Memory: Seven seconds. I’ve seen many rebuttal strategies that address review comments point by point. However, from my own experience, I might have long forgotten why I gave a particular review comment. If you reply to me, I’d have to go back and look at the paper, recall my thought process at that time. So, if you feel that the discussion after rebuttal or the meta-review is like a “chicken talking to a duck” (i.e., talking past each other), it’s because the reviewer has forgotten; their only memory of your paper is “I want to reject it.” The effort you put into a rebuttal is essentially betting that “the reviewer is someone willing to spend time re-reading your paper.” Honestly, this probability isn’t high either. So, it might be best to tell the complete story directly in your rebuttal. Reduce the gamble wherever possible.
3. Tell Them What to Do: Following up on the previous point, whether it’s an AC or a reviewer, people have an inherent inertia when doing things. Being too lazy to re-read a paper/supplementary material/review (that they think should be rejected) is a very realistic situation. Even academic giants far more accomplished than myself are the same; your paper is far less interesting than Zelda. So, directly telling them what to do isn’t out of the question. Using myself as an example again, if a rebuttal asks me to re-read a specific part of the paper, I always will. Because, frankly, I don’t know what else to do; the author has pointed me to a clear path, so I’ll instinctively go look. Of course, the prerequisite here is that your objective must be clear, like the young fellow in this example: Please Give Me One Million Dollars (Note: This link is likely a humorous placeholder). Otherwise, if the reviewer follows your request and achieves nothing, you might end up “trying to steal a chicken but losing the rice used as bait” (i.e., your efforts backfire).

4. To Flatter or to Confront: That is the question. I’m always asked this, but I have a very different perspective on it, starting with the definition of the word “rebuttal.”

   “The word ‘rebuttal’ comes from the practice of law and debate, referring to the act of refuting or disproving an argument or evidence. Its origin can be traced back to the Old French word ‘rebouter,’ meaning ‘to thrust back,’ which in turn is composed of the prefix ‘re-‘ (meaning ‘back’) and ‘bouter’ (meaning ‘to push’ or ‘to strike’). This word evolved into the English ‘rebut,’ meaning to contradict or oppose, and then the suffix ‘-al’ was added to form ‘rebuttal,’ transforming the verb into a noun describing the act of refuting or the argument made in refutation.”

   What this means is that if a reviewer correctly understands the meaning of the rebuttal stage (the refutation stage), they will rationally treat points that confront them. Conversely, if a reviewer thinks rebuttal is just a process to “try to get me to change my score,” then confrontation is absolutely not an option; you must flatter. Only when they are happily flattered will they bother with you. Although it’s a bit comical (or racist), based on my long-term observation of various review processes, reviewers who are native English speakers (or use English as their working language long-term) tend to be the former, while non-native speakers (who don’t understand the word’s etymology) tend to be the latter. And the easiest way to distinguish between these two types is to look at the number of grammatical errors in their review.

It’s hard to say if these views have stirred up a hornet’s nest again, especially the fourth point. I accept all criticism. Also, please have your advisor review your rebuttal, and don’t blindly apply my views. You can flame me in the comments section, but please avoid provocative statements; any such comments will be collapsed. Thank you.