The CVPR 2026 review form, explained

How CVPR reviewers read your paper

CVPR's training material asks for a summary in the reviewer's own words, specific evidence-based strengths and weaknesses (novelty concerns must cite the specific prior work), and a recommendation with a justification that tells the authors what the one-page rebuttal should address. Do not reject solely for missing SOTA, missing comparisons to arXiv-only work, easily fixable errors, or a missing limitations section. Refer to "the paper", not "you" or "the authors".

• Refer to "the paper" — never "you" or "the authors".
• Professional, courteous tone with concrete suggestions.
• Short or generic reviews are flagged as irresponsible; each weakness needs specifics.
• Do not request new experiments as a condition for the rebuttal (the rebuttal cannot add new experiments or contributions).

Where the bar sits. CVPR 2026 accepted about 25% of 16,092 submissions. No score statistics are public; on this 1–5 mapping, a mean of ≥ 3.5 (mostly Weak Accept or better) is a rough proxy for acceptance.

The reviewer form, field by field

In the order reviewers see it, with every score and its official wording.

1. 01

   Summary

   Free text

   2–4 sentences in your own words: what the paper does and claims.

2. 02

   Strengths

   List of points

   Specific and evidence-based. Avoid generic praise such as "the paper is interesting".

3. 03

   Weaknesses

   List of points

   Specific and detailed — bullet lists of short phrases are not a detailed review. A novelty concern must cite the specific prior work and explain how closely it relates. Do not reject solely for not beating SOTA, for lacking comparisons to arXiv-only work, or for easily fixable errors.

4. 04

   Preliminary Recommendation

   Score · 1–5

   Strong Accept: significant contribution with strong evidence and broad interest. Weak Accept: sound and well motivated but incremental or not best on every benchmark. Borderline: reasonable but with gaps in evaluation or positioning. Weak Reject: limited novelty relative to closely related prior work. Reject: fundamental technical issues.

   1. 5Strong Accept
   2. 4Weak Accept
   3. 3Borderline
   4. 2Weak Reject
   5. 1Reject
5. 05

   Justification For Recommendation And Suggestions For Rebuttal

   Free text

   Connect the rating to the assessment above, and state precisely what the rebuttal should address.

6. 06

   Confidence Level

   Score · 1–5

   How certain you are about your assessment.

   Score	Meaning
   5	You are absolutely certain about your assessment. You are very familiar with the related work and checked the math/other details carefully.
   4	You are confident in your assessment, but not absolutely certain. It is unlikely, but not impossible, that you did not understand some parts of the submission or that you are unfamiliar with some pieces of related work.
   3	You are fairly confident in your assessment. It is possible that you did not understand some parts of the submission or that you are unfamiliar with some pieces of related work. Math/other details were not carefully checked.
   2	You are willing to defend your assessment, but it is quite likely that you did not understand the central parts of the submission or that you are unfamiliar with some pieces of related work. Math/other details were not carefully checked.
   1	Your assessment is an educated guess. The submission is not in your area or the submission was difficult to understand. Math/other details were not carefully checked.

7. 07

   Dataset Attribution

   Free text

   Are all data assets properly cited? Flag any use of withdrawn or deprecated datasets. "No concerns" otherwise.

8. 08

   Ethics Flag

   Free text

   "No" or a brief description of the ethical concern (e.g. human-subjects data without IRB statement).

Area chair meta-review

CVPR ACs work in triplets; the meta-review must justify the recommendation from the reviews and discussion.

• Award candidate
• Oral
• Highlight
• Poster
• Findings
• Reject

CVPR author requirements

Reviewers check these. Gaps are easy points to lose.

1. Main text within 8 pages plus references
2. Anonymized submission
3. Negative societal impact discussed where relevant
4. IRB approval stated for human-subjects data
5. All datasets cited; no withdrawn datasets used
6. No hallucinated or unverifiable citations

Policy on LLMs in reviewing