Reviewer Comment Triage

Use review comments to improve what the paper teaches its reader. A reviewer's question identifies a concern; it does not by itself determine the diagnosis or require a particular experiment.

Read independently

Establish the requested scope and mode: triage, proposed repair, source edit, or verification of a revision. Read the current relevant text and supporting evidence before adopting earlier assistant judgments. A prior PASS may help locate a passage but cannot prove that the present concern is resolved. Retain author-approved choices and well-functioning prose unless the current evidence reveals a concrete defect.

Read each comment for its substantive question. Separate the reviewer's interpretation, the source's actual claim, the evidence available, and the action already taken. If the supplied packet is incomplete, limit the conclusion to it rather than inferring a missing experiment or inventing a revision history.

Establish the right expectation

Choose the disciplinary and methodological lens required by the comment and the paper's study design. Separate the scientific reviewer's question from an editor's clarity or style preference. For each substantive concern, record what a satisfactory argument would establish, why that expectation applies, what the current source establishes, and the remaining difference. Distinguish the actual comment from an inferred reader need. Do not impose a broader scientific claim or another discipline's reporting convention.

For a changed passage, compare the earlier and current versions when supplied. A definition being longer or a reply saying it was clarified does not establish resolution. Check the actual operation and retained claim, while preserving valid motivation, conditions, and interpretation. Accept a supported narrowing or correction, and do not restore incorrect old wording merely for completeness.

Choose a proportionate repair

For each consequential concern, identify its primary cause:

Check comparators and ablations by their actual operations, not their names. Removing examples, masking a field, and removing supervision test different things. A temporal split does not establish transfer to another population, and higher performance alone does not identify a mechanism. If a genuinely missing study is essential to the retained claim, mark that dependency explicitly. New experiments are not the automatic consequence of a reviewer asking a broader question; honor the user's chosen scope.

Convert defensive prose into informative content. “We emphasize that our comparison is fair” usually needs a concrete comparison definition, not a stronger assurance. Keep qualifications that change the claim's truth or scope. Remove reviewer-directed staging from the manuscript when its reader-facing substance can stand directly. Draft a response letter only when requested, and do not describe a proposed repair there as completed.

Prose when drafting or revising

When drafting or revising prose, including audit explanations, read prose editing rules. Preserve scientific meaning, necessary reasoning and valid technical terms. Keep sound passages, do not add unreported properties, and limit edits to the requested scope. The local reference covers concept accuracy, clear sentences, repetition and punctuation, including the required exceptions.

Deliver a useful triage

Provide a concise overall judgment and, where several comments are involved, a compact mapping of concern → current evidence/location → decision → smallest repair → unresolved dependency. Prioritize scientific correctness and reader comprehension over stylistic preferences. Include explicit preserve/no-change decisions. Give replacement wording only for the authorized passages, with reasons tied to evidence.

Distinguish proposed wording, changes actually made, source consistency checked, and rendered results inspected. For a repair spanning Methods, captions, tables, or conclusions, verify the connected locations rather than fixing only the quoted sentence. If evidence is unavailable, state exactly what cannot yet be checked.

Stop when every in-scope concern has a supported disposition and the requested repairs are consistent. An unresolved evidence gap can remain honestly documented. Do not expand the task into a full rewrite, unsupported promises, automatic statistical analyses, or a second review of already sound passages.