IEEE TPAMI Response to Reviewers: How to Write a Rebuttal That Survives R2 (2026)

TPAMI is an IEEE Computer Society journal run through the ScholarOne Manuscript Central portal and managed by Associate Editors who recruit topic-matched reviewers, with the IEEE Computer Society advocating three reviews per paper. Two features of this process shape how you should write the rebuttal, and both work against the author who treats the response as a formality.

First, the same reviewers re-review a major revision. The IEEE Transactions process re-invites the reviewers from the prior round, including those who recommended rejection. The Associate Editor reads your response against the comments the same people already wrote. This is the opposite of a fresh-eyes second submission: every reviewer can compare your claimed change to what they originally asked for, line by line. A reply that paraphrases a comment to make it easier to answer is caught instantly, because the reviewer wrote the original wording.

Second, this is a journal of methodology, not applications, and the second round is where the additional-experiments and state-of-the-art-comparison norms bite hardest. TPAMI reviewers routinely demand a deeper ablation, comparison to recent baselines on standardized benchmarks such as ImageNet, COCO, and ADE20K, and a reproducibility package. On revision, "we will add this in a future work" is not an acceptable answer to a request for a missing experiment.

If the experiment is feasible, run it; if it is genuinely out of scope, the burden is on you to defend the boundary with evidence, not to defer.

Every response must indicate the exact page and line where the manuscript changed. This is the most common reason a TPAMI reviewer reopens an otherwise-resolved comment. Submit a change-tracked PDF alongside the clean version, typeset reviewer comments in boxes, and color your changed text so the reviewer can confirm the edit without searching. When a change spans a new table or figure, reference it by its new number ("New Table 4, page 9"), because reviewers reading three papers a week will not remember your old numbering.

Rejection after major revision is common at TPAMI. The journal accepts roughly 15 to 20 percent of submissions, and the one-resubmission norm means most major-revised papers face a binary accept-or-reject on the second round. A rejection on revision is not a freak outcome; the majority of post-review papers do not end in acceptance.

When it happens, the instinct is to appeal. Appeals succeed only when you can show a reviewer made a verifiable factual error about your method or your results, not when you disagree with their judgment about novelty or significance. Disagreement is not grounds for an appeal, and the same Associate Editor who issued the decision usually handles it.

The more useful question is whether the rejection was about depth or about fit. If reviewers found the methodology sound but read the contribution as a thin extension of a conference paper, the work was conference-shaped to begin with, and converting it to a CVPR, ICCV, or ECCV submission is often faster and cleaner than a journal appeal.

If reviewers found a genuine methodological gap, a second journal will surface the same gap, so fix the science before resubmitting anywhere. Do not resubmit the identical manuscript to a sister IEEE journal hoping for different reviewers; the IEEE Computer Society portfolio coordinates, and reviewer pools overlap.

• Every reviewer comment has its own numbered reply that ends with an exact page and line reference, and each new result has a labeled table or figure number.
• The comparison table adds at least one state-of-the-art baseline published after your conference version, on the standardized benchmark the reviewer named, with ties and losses reported honestly in the text.
• The requested ablation appears as a dedicated table that varies one factor per row and isolates the proposed component.
• The reproducibility statement points to a real repository, and the released configuration matches the numbers in the paper.
• Your replies stay substantive on methodology and gracious on minor points, because the same reviewers who wrote the comments will read every line.

In our pre-submission review work with IEEE TPAMI submissions, we read response letters and revised manuscripts before authors resubmit, and the same named failure patterns recur across computer-vision and pattern-analysis manuscripts. In our analysis of TPAMI rebuttals, the editorial triage pattern is consistent: because the same reviewers re-review a major revision at TPAMI, these patterns are not cosmetic; each one is something the original reviewer will notice immediately.

The phantom experiment. Across our TPAMI pre-submission reviews, the most common named rejection pattern in a rebuttal is a response that claims an experiment was run without adding the corresponding table or figure to the manuscript. A reviewer who asked for an ablation expects a new ablation table, not a sentence asserting that the module helps.

We flag any point where the response language ("we conducted additional experiments," "we tested extensively") has no matching new result in the revised manuscript. At TPAMI specifically, where additional experiments are the dominant revision request, this gap is the single fastest way to turn a clearable comment into a second major revision or a reject.

SOTA comparison that stops at the conference version. TPAMI reviewers demand comparison to recent state-of-the-art baselines, and they read the comparison table against the calendar. Manuscripts coming through our pre-submission review pipeline for TPAMI often carry a benchmark table that was current when the conference paper was written but is now a year stale.

When the first-round comment was "compare to recent baselines," a revision that adds nothing newer than the original submission reads as non-responsive. We check that the revised comparison table includes baselines published after the conference version, on the standardized benchmarks TPAMI reviewers expect (ImageNet, COCO, ADE20K), and that ties or losses are reported honestly rather than buried.

The ablation that does not isolate the mechanism. A second recurring TPAMI pattern: the response adds an ablation, but the ablation varies several things at once and cannot attribute the gain to the proposed component. Reviewers in pattern analysis and machine intelligence are trained to ask whether the improvement comes from the new module or from the training schedule, the data augmentation, or the longer schedule.

Across our TPAMI-targeted reviews, we flag ablations that change more than one factor per row, because the same reviewer who requested the ablation will reject the conclusion if the experiment cannot support it.

The missing reproducibility package. TPAMI increasingly expects released code and configurations, and a revision that promises code "upon acceptance" without a repository link weakens the rebuttal. In our pre-submission reviews of TPAMI manuscripts, we check whether the reproducibility statement points to an actual artifact and whether the configuration in the paper matches the released code, because a reviewer who cannot reproduce the headline number will not clear the comment that asked for reproducibility.

These four patterns are testable against your own response letter before you upload it. Read each reviewer comment, then ask whether your reply names a specific manuscript change, points to a real new result, and survives being read by the person who wrote the comment.

Check whether your TPAMI rebuttal answers every reviewer with a real change

Source: Manusights internal review of TPAMI-targeted resubmissions, 2025 to 2026 cohort.

Total submission-to-acceptance for a successful TPAMI paper commonly runs 6 to 14 months across one or more revision rounds. The Associate Editor takes 1 to 2 weeks to analyze returned reports and issue the next decision once the re-invited reviewers respond.

• Manusights internal pre-submission review corpus (2025 to 2026 TPAMI-targeted cohort)