The rebuttal is one of the most consequential documents you will write as an academic researcher — and one of the least taught. Most PhD programmes offer no formal instruction on how to write one, leaving researchers to learn by trial, error, and occasionally painful rejection. This guide gives you the framework to write rebuttals that actually move scores.
When Rebuttals Are Offered
Not all conferences include a rebuttal phase, but many of the most competitive venues do. NeurIPS, ICML, ICLR, ACL, EMNLP, CVPR, ICCV, and ECCV all offer structured author response periods, typically lasting between three and seven days after reviews are released. During this window, authors can submit a response of limited length — often 500 words, sometimes up to one page — that reviewers and area chairs read before making their final recommendations. Some venues also allow a back-and-forth discussion phase between authors and reviewers. Knowing whether a venue offers rebuttals before you submit — information readily available on conference pages indexed through platforms like latestconferences.com — should factor into your submission timing and preparation strategy.
The Rebuttal's Purpose: Clarify, Do Not Extend
The single most important principle of effective rebuttals is understanding what they are for: clarification of existing content, not addition of new results. Reviewers and programme committees are not evaluating a revised or extended paper — they are evaluating the paper that was submitted. A rebuttal that introduces new experiments, new theoretical results, or substantively new claims that were absent from the original submission will typically be discounted or may even be penalised. Your job is to demonstrate that the paper already answers the reviewer's concerns, or that their concern rests on a misreading, or — where the concern is valid — to show that you understand it and commit to a specific, credible fix in the camera-ready version.
Addressing Factual Errors vs. Opinion Differences
Distinguish sharply between two categories of reviewer concern. Factual errors — where the reviewer has misread a table, confused your method with a baseline, or made a claim about your results that is demonstrably incorrect — are worth addressing directly and confidently, with a precise pointer to the relevant section, figure, or line. Opinion differences — where the reviewer has a different view of what constitutes an important contribution, a sufficient baseline comparison, or an interesting problem — require a more delicate approach. You can present additional reasoning and perspective, but arguing that a reviewer is simply wrong about a matter of scientific taste rarely succeeds and often backfires. Acknowledge the legitimate dimension of the concern and explain your design choices without dismissing the reviewer's perspective.
Tone and Length
Tone in a rebuttal should be precisely calibrated: confident but not arrogant, clear but not curt, responsive but not defensive. Avoid phrases that signal frustration with the review process. Even if a review is careless, unfair, or clearly written by someone outside your subfield, your rebuttal must read as if written by a researcher who is grateful for the engagement and committed to improving the work. On length: most rebuttal windows impose word or character limits, but even when they do not, brevity is a virtue. A rebuttal that a reviewer can read in five minutes is more likely to be read carefully than one that requires fifteen. Prioritise depth on the most consequential concerns over breadth of coverage across every minor point.
What Actually Moves Scores
Research on rebuttal effectiveness and the experience of veteran area chairs points to consistent patterns. Score changes are most likely when: a rebuttal corrects a factual misreading that drove a low score; a rebuttal demonstrates that an experiment the reviewer requested was already in the paper (perhaps in the appendix); a rebuttal addresses a concern raised by multiple reviewers simultaneously; or a rebuttal provides a clear, concrete plan for a specific revision that would resolve the primary weakness. Score changes are least likely when the rebuttal merely reasserts the contribution without engaging with the specific concern, when it introduces results not in the original paper, or when it is written in a tone that puts reviewers on the defensive.
Example Rebuttal Structure
A reliable structure for most rebuttals runs as follows:
- Opening sentence: Thank the reviewers briefly and signal that you will address all major concerns.
- Block 1 — Shared concerns: Address issues raised by two or more reviewers together, demonstrating that you recognise their importance.
- Block 2 — Reviewer 1 specific concerns: Numbered responses mirroring the reviewer's numbered points.
- Block 3 — Reviewer 2 specific concerns: Same structure.
- Block 4 — Reviewer 3 specific concerns: Same structure.
- Closing sentence: Commit to specific, named revisions in the camera-ready that address the remaining valid concerns.
If word count is tight, prioritise the shared concerns block and the highest-confidence reviewer's concerns over minor points from borderline reviewers.
Common Rebuttal Mistakes to Avoid
Several mistakes appear consistently in unsuccessful rebuttals:
- Spending the majority of your word budget on a reviewer who is clearly positive and already recommending acceptance.
- Failing to address the primary concern of a reviewer who is borderline — area chairs will notice the omission.
- Writing "We will add experiments" without specifying what those experiments are and what result you expect — vague commitments carry no weight.
- Using the rebuttal to question the reviewer's expertise rather than their arguments.
- Submitting a rebuttal on the last hour of the deadline without having had a colleague review it — fresh eyes catch miscommunications you have become blind to.
When the Rebuttal Does Not Help
Sometimes the reviews reflect a genuine mismatch between your paper and the venue's priorities, and no rebuttal will bridge that gap. When three reviewers independently agree that the contribution is insufficient for this venue, the most efficient response is to accept the decision, implement the substantive improvements that the reviews have identified, and identify a more appropriate target venue for the next cycle. A rejection from a high-quality peer review process — even a painful one — is among the most information-dense feedback you will receive as a researcher. Extract every insight it contains before you move on.