CVPR 2027 · author response

The CVPR rebuttal: one page, no links, no surprises

CVPR gives you one page to answer every reviewer, and anything past that page is simply not read. CVPR 2027 reviews come out on Jan 25 and the rebuttal window closes on Feb 1. Here's how to use the page well.

Last checked against official sources on October 12, 2026.

Dates

Reviews released: Jan 25, 2027.
Rebuttal period: Jan 25 – Feb 1, 2027, Anywhere on Earth (stated in the 2027 Call for Papers).
Final decisions: Feb 25, 2027.

One week, and it overlaps with other deadlines. Read the reviews on day one, start any requested experiments immediately, and draft the page around them.

The rules

From the CVPR 2026 author guidelines and the official rebuttal template. The 2027 author guidelines aren't posted yet, so re-check them in January.

Length	One page, using the official rebuttal template. References and figures count toward the page. Longer responses, or ones with changed margins or formatting, are not reviewed.
New results	Avoid new contributions or experiments unless a reviewer specifically asked for them. Reviewers are told to disregard unrequested ones.
Links	No links to external material: no code, videos or project pages.
Anonymity	The rebuttal must stay anonymous, like the paper.
Area chair	A confidential comment to the area chair is allowed only in exceptional circumstances.

Fitting it on one page

Group concerns by topic, not by reviewer. One answer tagged (R1, R3) saves half the space of two separate answers.
Lead each answer with the conclusion in bold, then one line of evidence. Reviewers skim.
Cite the submission ("Table 3", "Sec. 4.2") instead of restating it.
If a reviewer asked for a number, give the number in a one-line table, not a paragraph.
Figures and references count toward the page. Use a figure only when a reviewer asked for something visual.
Put typo and clarity fixes in one closing line per reviewer.

What to answer first

CVPR reviewers give a recommendation with a justification that is meant to tell you what the rebuttal should address, so start there. Answer, in order:

Anything two or more reviewers raised.
The concern named in each borderline reviewer's justification.
Factual misreadings, with a pointer to where the paper already answers them.
Requested experiments you were able to run.
Everything else, briefly.

General method: triage, the four-beat answer and handling reviewer 2 are in how to write a rebuttal.

Skeleton for the official template

CVPR rebuttal skeleton (LaTeX body)

% Start from rebuttal.tex in the official CVPR author kit; keep its margins and fonts.
% One page total, INCLUDING references and figures.

We thank the reviewers. R1, R2 and R3 agree the paper is [shared strength].
The main concerns are (A) [concern raised by 2+ reviewers] and (B) [second concern].

\noindent\textbf{(A) [Concern A] (R1, R3).}
[Direct answer in one sentence.] [Evidence from the submission: Table/Fig/Sec.]
[Requested result, only if a reviewer asked for it: one line or one tiny table.]

\noindent\textbf{(B) [Concern B] (R2).}
[Answer.] [Evidence.] [What the final version will clarify.]

\noindent\textbf{R1-Q1 [short label].} [Answer in 1-3 lines.]
\noindent\textbf{R2-Q2 [short label].} [Answer.]
\noindent\textbf{R3 minor.} [Typos and clarifications we will fix: ...]

% References (count toward the page): [1] ... [2] ...

Mistakes that waste the page

Going over one page or shrinking the margins: the whole response is not reviewed.
Adding unrequested experiments: reviewers are told to ignore them.
A link to a project page or video, which isn't allowed.
Spending lines on thanks and praise for each reviewer.
Answering every minor point at length and the main concern in one line.

Before submission, see the CVPR 2027 submission checklist and the CVPR review form explainer.

Find the objections before January

The best rebuttal is the one with little to rebut. Run your CVPR draft past three simulated reviewers and an area chair, with every point linked to the passage it's about. First three runs free.

Review my CVPR paper, 3 free

Sources

reviewpaper.ai is not affiliated with CVPR, the CVF or IEEE or its organizers. Rules change between cycles; the official pages above always win.