flonat/flonat-research/review-response

Review Response

Systematic workflow for responding to reviewer comments on academic papers. Covers the full cycle from parsing comments through to a polished rebuttal document.

When to Use

• "Help me write a rebuttal"
• "Respond to reviewer comments"
• "Handle this R&R"
• "Develop a review response strategy"
• Paper has received referee reports and needs a structured response

Workflow

1. Receive reviewer comments
2. Parse and classify each comment (Major / Minor / Typo / Misunderstanding)
2b. Re-audit the paper against its artifact (empirical papers)
3. Develop response strategy per comment (Accept / Defend / Clarify / Experiment)
4. Write structured responses
5. Tone check — every response must pass the tone checklist
6. Assemble final rebuttal document

Step 1: Parse and Classify

Read all reviewer comments and classify each one:

Keyword signals for classification:

• Major: "major concern", "fundamental issue", "missing experiments", "insufficient evidence", "not convincing"
• Minor: "minor concern", "could be improved", "please clarify", "suggestion"
• Typo: "typo", "grammar", "formatting", "inconsistent"
• Misunderstanding: "The authors did not..." (but they did), "It is unclear..." (but it is stated)

Priority order: Major > Misunderstanding > Minor > Typo

Present the full classification table to the user before proceeding to strategy.

Step 1b: Re-audit the paper against its artifact (empirical papers)

Before assigning any strategy, check the submitted paper against the code, the run outputs and the replication package. Reviewers see the artifact too. A rebuttal that defends a claim the artifact contradicts does more damage than the original error.

• Every reviewer claim about the artifact. Check each one directly against the files (model, framework, environment, "the code does not do X"). Classify it as reviewer-correct, reviewer-misread or ambiguous in the paper, with file:line evidence.
• Qualitative methods claims. Check the paper's descriptions of tools, models, what was logged, procedures performed and pre-registration status against the code and artifacts. Treat protocol or plan documents as the source of those claims, never as evidence for them.
• Headline numbers. Recompute them from the saved outputs, and spot-read the raw records behind any extraction or matching step (parsers, path or ID normalisers). Treat any anomaly a reviewer points to, or any count that looks too convenient, as a lead to follow.

If the re-audit finds a genuine error, stop and present it to the user before drafting, with the corrected figures and the claims that no longer hold. Whether and how to disclose it is his decision. When the audit was prompted by a reviewer's scrutiny, credit that scrutiny without implying the reviewer found the error. Trigger incident: 2026-09-23 ICSE #1402. A re-audit prompted by one reviewer's code inspection found a result-changing bug and four protocol-vs-implementation drifts that 25 logged pre-submission review passes (12 distinct checks) had missed.

Step 2: Develop Response Strategy

For each classified comment, assign a strategy:

Decision flow:

Comment → Is the reviewer correct?
  Yes → Is the fix feasible?
    Yes → Accept
    No  → Accept principle + explain constraint + offer alternative
  Partially → Accept valid part + Defend the rest with evidence
  No (misunderstanding) → Clarify with specific location references
  Requests new work → Is it feasible?
    Yes → Experiment
    No  → Explain limitation + offer alternative analysis

Strategy combinations (common in practice):

• Accept + Clarify: "We agree and have expanded Section 3. We also note this was partially addressed in Table 2."
• Defend + Experiment: "Our choice of X is justified because [reasons], but we have also added the comparison with Y as requested."

Full strategy library with templates: references/response-strategies.md

Step 3: Write Responses

Each response follows this structure:

**Comment N.M**: [Reviewer's original comment, quoted verbatim]
 
**Response**: [Our substantive reply]
 
**Changes**: [Specific modifications with section/page/table references]

Non-negotiable rules:

1. Every comment gets a response — even typos
2. Every response starts with thanks
3. Every claim in the response has evidence (data, citation, or specific manuscript reference)
4. Every change states its location (Section X, page Y, Table Z)
5. Never say "The reviewer is wrong" — use "We would like to respectfully clarify"
6. Every "Changes:" entry is verified against the manuscript before the letter is assembled — read/grep the actual .tex at the cited location; never trust the comment tracker, an earlier letter draft, or memory of the edit. A promised change with no matching manuscript edit is either downgraded to an honest acknowledgment ("we discuss rather than change...") or flagged to the user as unfulfilled — a letter must never describe the manuscript as more revised than it is. Record the outcome per commitment in the tracker's Fulfillment Ledger (fulfilled / partial / not_fulfilled / acknowledgment_only; see templates/referee-comments/comment-tracker.md).

Step 4: Tone Check

Before finalising, run every response through this checklist:

• [ ] Opens with genuine thanks (not perfunctory)
• [ ] Uses "We" not "I"
• [ ] No defensive or aggressive language
• [ ] No "obviously", "clearly", or "it is well-known"
• [ ] Specific location references for all changes
• [ ] No vague promises ("We will..." without specifics)
• [ ] Misunderstandings addressed with "We apologise for the confusion" not "The reviewer failed to notice"

Full tone guide with good/bad examples: references/tone-guidelines.md

Step 5: Assemble Rebuttal Document

Standard structure

# Response to Reviewers
 
We sincerely thank all reviewers for their valuable feedback. We have carefully
addressed all comments and made substantial revisions. Below we provide detailed
responses to each reviewer's comments.
 
---
 
## Response to Reviewer 1
 
### Major Comments
 
**Comment 1.1**: [verbatim]
**Response**: [reply]
**Changes**: [locations]
 
### Minor Comments
 
**Comment 1.2**: [verbatim]
**Response**: [reply]
**Changes**: [locations]
 
---
 
## Response to Reviewer 2
[same structure]
 
---
 
## Summary of Major Changes
 
1.[Change 1] (addressing Reviewer X, Comment Y)
2.[Change 2] (addressing Reviewers X and Z)
3....
 
We believe these revisions have significantly strengthened the manuscript.

Where multiple reviewers raise the same issue

Consolidate: "We thank Reviewers 1 and 3 for raising this important point. We have [action] which addresses both concerns."

Where reviewers contradict each other

Acknowledge both perspectives: "After careful consideration, we have [chosen approach] because [reasons], which we believe addresses both reviewers' core concerns."

Full template library: references/rebuttal-templates.md

Output

The skill produces a complete rebuttal document saved to the project's correspondence/ directory (or paper directory if no correspondence folder exists).

Naming convention: rebuttal-{venue}-{round}.md (e.g., rebuttal-jmp-r1.md)

Templates

Three templates live under templates/referee-comments/. Use them together across a revision cycle:

When to use the LaTeX scaffold vs. the Markdown rebuttal

• Markdown rebuttal (the default rebuttal-{venue}-{round}.md above) — for venues that accept Markdown/plain-text uploads, internal review by co-authors, or quick R&Rs.
• LaTeX `response-letter-ansrev.tex` — when the venue requires a typeset PDF response, when the main paper has many labels/citations the response needs to reference, or when reviewers will be assigned numbers and cross-referenced ("see our reply to Reviewer 1 Comment 3"). Strongly preferred for OR / Management Science / journals with structured R&R.

Using response-letter-ansrev.tex

1. Copy templates/referee-comments/response-letter-ansrev.tex into the project's correspondence/ (or paper-{venue}/paper/) directory.
2. Copy templates/referee-comments/ansrev/{ansrev.sty,revquote.sty} next to it — these are vendored from GitHub (not on CTAN). See templates/referee-comments/ansrev/README.md for provenance.
3. Keep the canonical project .latexmkrc unchanged. Copy templates/referee-comments/ansrev/.latexmkrc.local beside it so latexmk auto-recompiles the main paper for xr cross-refs. If .latexmkrc.local already exists, merge the custom dependency into that supplement rather than overwriting it.
4. Set main={<main-file-basename>} in the scaffold's \usepackage{ansrev} options.
5. Compile the main paper first, then the response file. Refer to comment-tracker IDs (R1-C1, AE-C2) as \label{}s inside each \QA{}{} — \ref{R1-C1} elsewhere expands to "Reviewer 1 Comment #1".

Integration

• Before writing: Read the reviewer comments and the paper to understand context
• During writing: Cross-reference the paper for specific section/page numbers
• After writing: Run proofread on the rebuttal document itself for tone and clarity
• If paper changes are needed: Track them separately — the rebuttal documents the response, not the revision

Reference Documents

• references/review-classification.md — Classification criteria with keyword signals
• references/response-strategies.md — Strategy library with templates for each type
• references/rebuttal-templates.md — Full rebuttal document templates
• references/tone-guidelines.md — Tone guide with good/bad expression pairs
• references/figure-interpretation.md — Figure interpretation guide (useful when discussing figures in responses)

Adapted from

galaxy-dawn/claude-scholar review-response skill, adapted for general academic research (not ML-specific).