Reviewer Expectations and Rebuttals

Reviewed: 2026-10-01. This is a self-review aid, not a prediction of scores, acceptance, reviewer counts, or turnaround times. Obtain the actual review form for the venue, contribution type, and cycle.

Primary sources

• Nature editorial criteria: https://www.nature.com/nature/for-authors/editorial-criteria-and-processes
• Cell author instructions: https://www.cell.com/cell/authors
• NeurIPS 2026 handbook: https://neurips.cc/Conferences/2026/MainTrackHandbook
• ICML 2026 instructions: https://icml.cc/Conferences/2026/AuthorInstructions
• ICLR 2026 guide: https://iclr.cc/Conferences/2026/AuthorGuide
• CVPR 2026 guidelines: https://cvpr.thecvf.com/Conferences/2026/AuthorGuidelines
• CHI 2026 review process: https://chi2026.acm.org/papers-review-process/
• ARR current call: https://aclrollingreview.org/cfp
• Medical reporting resources: medical_journal_styles.md

Evaluate the contribution on its terms

Do not assign invented fixed “Critical/High/Moderate” official weights. Scientific soundness is not demonstrated by checklist completion, a large sample alone, a small P value, animal work, or a state-of-the-art benchmark win. Novel insight, negative findings, theoretical work, and other contributions may be in scope.

Common substantive concerns

• Claims exceed the design, analysis, data coverage, or uncertainty.
• Important controls, assumptions, alternative explanations, or baselines are missing.
• Data splitting, tuning, exclusions, missingness, or outcome selection compromise inference.
• Methods are insufficiently specified for interpretation or reproduction.
• Related work, overlapping submissions, or access restrictions are misrepresented.
• Ethical review, consent, disclosures, or privacy protections are inadequately reported.

A failure to beat a benchmark, a small numeric improvement, an older comparator, or one dataset is not by itself a universal rejection rule. Judge whether the evidence is adequate for the actual claim.

Rebuttal contract comes first

Check the response format, anonymity, word/page limit, permitted links, new-data policy, revision policy, and deadline before drafting. Examples:

• ICML 2026 permits author responses but no revised paper during author feedback; its instructions explicitly encourage prioritizing substantive concerns.
• CVPR 2026 uses a one-page anonymous rebuttal and restricts external links and new contributions or unrequested experimental results.
• ICLR 2026 allows documented revisions under its discussion rules.
• CHI 2026 has a specified revise-and-resubmit process for invited papers.

A general suggestion to “do every requested experiment now” must not override these rules or the authors' research scope.

Evidence-based response workflow

1. Separate factual misunderstandings, unclear exposition, methodological problems, and requests for expanded scope.
2. Identify what can be corrected, clarified, or tested within the allowed stage.
3. Give the evidence and exact manuscript location for each material response.
4. Label new analyses, planned work, and untested hypotheses distinctly.
5. State remaining limitations without promising nonexistent results.
6. Check all response and supplementary files for identity leaks when required.

Illustrative response pattern:

Concern: [reviewer's substantive point]
Response: [agreement, clarification, or justified disagreement]
Evidence: [existing analysis or actually completed allowed new analysis]
Change: [specific text/table/figure and location, if revisions are allowed]
Remaining limit: [what the evidence still does not establish]

For a precision concern, give the prespecified sample-size rationale and the observed confidence interval. Do not use observed/post hoc power or the favorable direction of a noisy estimate as proof of adequate evidence. For a method concern, show the relevant diagnostic or bounded correction; do not fabricate a rescue experiment, deployment, participant quotation, or benchmark comparison.

Human and policy responsibilities

Authors remain responsible for accurate responses and submission decisions. Follow the venue's rules on AI use and confidential material. This guide does not authorize an agent to submit a review, contact reviewers, or upload a manuscript. Journal-specific policies may restrict permitted assistance; see the medical reference for JAMA's current exceptions.