Inside the CVPR Rebuttal: The One-Page Reply That Shapes Computer Vision

This lightning talk demystifies the CVPR author rebuttal, a critical but often misunderstood stage of the peer-review process. We explore what this single page can and cannot do, how formatting constraints shape scientific argumentation, and why the rules around new experiments protect both authors and reviewers. By examining the structure and purpose of this procedural document, we reveal how a seemingly bureaucratic template actually defines the boundary between clarification and new contribution in one of computer vision's most competitive venues.
Script
After months of work and weeks of waiting, your computer vision paper receives reviews. You have exactly one page to respond. This is the CVPR rebuttal, and every margin, every column width, every font size is specified to the tenth of an inch.
The rebuttal exists to correct factual errors and clarify misunderstandings, not to present new science. Authors may not introduce new theorems, algorithms, or experimental results. This bright line separates responsive clarification from post-hoc contribution.
The format itself enforces discipline: a text area of 6 and 7 eighths by 8 and 7 eighths inches, two columns each 3 and a quarter inches wide, separated by 5 sixteenths of an inch. Authors work in 10-point Times, with references squeezed to 9-point. Every line counts.
Within that single page, authors may include one figure, a proof, or a comparison table built from already reported numbers. But reviewers are explicitly told not to request new experiments during rebuttal, and authors are protected from penalties for not running them.
The stakes are real but the scope is narrow. A well-crafted rebuttal can rescue a paper from misunderstanding or factual error. But it cannot fix a flawed experiment, and it should not try. The template acknowledges that peer review is iterative, but each iteration has boundaries.
This procedural template defines more than formatting. It codifies the negotiation between authors and reviewers at one of computer vision's most competitive venues, turning clarification into a constrained, high-stakes craft. To explore more research shaping the peer-review process, visit EmergentMind.com and create your own video breakdowns.