Home / Venues and deadlines

Submitting to CVPR 2027

The dates from the official call, the 2027 author guidelines and what gets a paper rejected without review, and what to check in your PDF before you upload it.

DeskReady · Facts checked against the official CVPR pages on 22 October 2026

Deadlines

All times are Anywhere on Earth (UTC-12), from the CVPR 2027 call for papers.

What the 2027 call asks of authors

• An OpenReview profile for every co-author. The call requires an up-to-date profile for all co-authors, and a new profile without an institutional email goes through moderation that "can take up to two weeks". If a co-author has no profile yet, this is the first thing to do, well before 10 November.
• No similar paper elsewhere during review. By submitting, authors confirm that no substantially similar paper has been or will be submitted to another conference or workshop between 16 November 2026 and 22 February 2027.
• Every author reviews. All authors are added to the reviewer pool unless exempt; exemptions are granted only in exceptional circumstances and must be approved by the program chairs in advance. Highly irresponsible reviewing can get the reviewer's own CVPR 2027 submissions rejected.
• Contemporaneous work. Papers that appeared online after 15 September 2026 are generally treated as contemporaneous: not comparing against them is generally not a reason to reject, but they should still be cited and discussed.
• The author list is fixed early. After 10 November authors can't be added or removed, only reordered, and the order is final after 16 November. Every author needs a complete OpenReview profile by 10 November, or the paper is rejected without review.

The 2027 author guidelines

The CVPR 2027 author guidelines were published on 22 October 2026. What changed from last year and what gets a paper rejected without review:

• New: a mandatory LLM Usage Statement. After the main text and before the references, under the title "LLM Usage Statement": which LLM or generative AI tools were used in the research or the writing (editing and grammar included), which parts of the paper they affected and to what extent. If none were used, the guidelines give the sentence: "No LLMs or generative AI tools were used in this work". It does not count toward the page limit, must stay anonymous, and a paper without it is rejected without review.
• Eight pages, including figures and tables, in the CVPR 2027 template. Extra pages may contain only cited references and the LLM Usage Statement.
• New: the reciprocal-reviewer quota. Each author may co-author at most one paper on which no author is an eligible reciprocal reviewer, meaning someone with an accepted main-track conference paper or full journal article at one of the listed venues by 10 November. In practice: if none of you has such a paper, you can submit one paper. The 25-papers-per-author limit continues.
• Rejected without review: more than eight pages, a missing or altered template, no LLM Usage Statement, a paper that is not properly anonymised, incomplete author enrollment or inaccurate conflicts, external links that expand the paper, dual submission, plagiarism, prompt injection (hidden or visible instructions aimed at reviewers or their tools), and fabricated content, "including citations to non-existent work".
• Anonymity: no names, co-workers or grant numbers in the acknowledgments, no links that identify the authors, and no citation of your public code repository (you may say the code will be released).
• No external links that expand the content, in the paper, the supplementary material or the rebuttal.
• Rebuttal: one page in the rebuttal template, without new contributions or results.

Before you upload: general things to check

Beyond the formal rules, a few things worth a last look in any paper, easy to miss after many rounds of edits. In computer-vision papers they often look like this:

• Numbers in the text that the tables contradict. "Our method improves mAP by 2.3 points" when the table shows 1.9, or a result quoted from an older run of the experiments.
• Bold that marks the wrong cell. The best result per column is bold and the second best underlined, and after the last round of experiments the marks no longer match the numbers.
• Claims the results don't show. "Best on every benchmark" when one row disproves it, or an ablation the text describes differently from the table.
• Figures that disagree with their captions and the text. A legend naming the wrong method, a plotted point that doesn't match its table, a trend the text describes and the curve doesn't show.
• References that lead nowhere. The 2027 guidelines name citations of non-existent work as a reason to reject without review. Check every entry in a real database; the free reference checker does it for a PDF or a pasted list.

The method for each: when your text and tables disagree, numbers that don't add up and the pre-submission checklist.

Sources

• CVPR 2027 call for papers (dates, OpenReview, dual submission, contemporaneous work)
• CVPR 2027 author guidelines (page limit, LLM Usage Statement, quotas, desk rejection, anonymity, links, reviewing, rebuttal)

DeskReady is not affiliated with CVPR or the Computer Vision Foundation. Always confirm dates and rules on the official pages before you submit.