Paper Revision

Paper Revision

Systematically revise papers based on reviewer feedback.

Input

• $0 — Reviewer comments/feedback
• $1 — Current paper draft (main.tex or paper directory)

References

• Revision workflow and prompts: ~/.claude/skills/paper-revision/references/revision-prompts.md

Workflow

Step 1: Parse and Prioritize Concerns

For each reviewer comment: 1. Extract the specific concern 2. Classify: major revision, minor revision, question, suggestion 3. Map to affected paper section(s) 4. Prioritize: address major concerns first

Step 2: Plan Revisions

Create a revision plan:

Concern → Affected Section → Required Action → New Content/Experiment

Categories of actions:

• Clarification: Rewrite text for clarity
• Additional experiment: Run new experiment, add results
• New analysis: Add ablation, statistical test, or comparison
• Structural change: Move, merge, or split sections
• Citation: Add missing references

Step 3: Execute Revisions

For each planned revision: 1. Read the current section 2. Apply targeted edits (preserve surrounding structure) 3. If new experiments needed: use experiment-code skill 4. If new figures/tables needed: use figure-generation / table-generation skills 5. Mark changes (use \textcolor{blue}{...} for revised text)

Step 4: Verify Improvements

• Re-run self-review skill to check if scores improved
• Verify all reviewer concerns are addressed
• Check that revisions don't introduce new issues
• Ensure page count still fits venue requirements

Step 5: Write Revision Summary

Generate a diff summary:

• List all changes made with section references
• Note any new experiments, figures, or tables added
• Cross-reference each change to the reviewer concern it addresses

Rules

• Address EVERY reviewer concern — do not skip any
• Preserve paper structure unless structural change is explicitly needed
• New results must come from actual experiments, not hallucinated
• Mark all revised text clearly for the reviewers
• Keep a copy of the previous version before revising
• Compare new scores vs previous scores after revision

Related Skills

• Upstream: self-review
• Downstream: paper-compilation
• See also: rebuttal-writing, paper-writing-section

Paper Revision Prompts

Extracted from AI-Scientist (perform_improvement) and AgentLaboratory (report_refinement).

Concern Classification

For each reviewer comment, classify as:

| Type | Description | Priority |
|------|-------------|----------|
| Major Revision | Fundamental flaw, missing experiment, wrong methodology | P0 |
| Minor Revision | Clarity issues, missing details, presentation | P1 |
| Question | Reviewer seeks clarification | P2 |
| Suggestion | Nice-to-have improvement | P3 |

Revision Planning Template

For each concern:

Concern: [Extracted reviewer concern]
Type: [Major/Minor/Question/Suggestion]
Affected Section(s): [Section name(s)]
Required Action: [Clarification/Experiment/Analysis/Structural/Citation]
Plan:
  - Step 1: ...
  - Step 2: ...
New Content Needed: [Yes/No - describe if yes]

Revision Execution (AgentLaboratory)

Full Pipeline Re-run

When a major revision requires new experiments:

1. Save current state:
   prev_code = current_code
   prev_results = current_results
   prev_paper = current_paper

2. Inject reviewer feedback as notes:
   notes += "Reviewer requested: {concern}"

3. Re-run affected phases:
   - If new experiment needed: re-run experiment phase
   - If new analysis needed: re-run analysis phase
   - Always re-run writing phase for affected sections

4. Compare new vs previous:
   - Score improvement: new_score - prev_score
   - Verify concern addressed

Targeted Edit Prompts

For Clarification Concerns

The reviewer found this passage unclear:
"{original_text}"

Their comment: "{reviewer_comment}"

Rewrite the passage to address the reviewer's concern.
Keep the same meaning but improve clarity.
Mark revised text with \textcolor{blue}{...}.

For Missing Experiment Concerns

The reviewer requests: "{reviewer_comment}"

Current experimental setup: {current_setup}
Available resources: {resources}

Design a new experiment to address this concern:
1. What to test
2. Expected outcome
3. How to present results (table/figure)

For Citation Concerns

The reviewer suggests citing: "{suggested_area}"

Search for relevant papers and add citations.
Integrate citations naturally into the text.
Do not just add a parenthetical — explain the connection.

Change Marking

% Use blue text for all revisions
\usepackage{xcolor}
\newcommand{\revision}[1]{\textcolor{blue}{#1}}

% Example usage:
\revision{We additionally evaluate on the CIFAR-100 dataset
to address the reviewer's concern about generalization.}

Revision Verification Checklist

After all revisions:
[ ] Every reviewer concern has been addressed
[ ] All revised text is marked in blue
[ ] New results are from actual experiments (not fabricated)
[ ] Page count still fits venue requirements
[ ] No new LaTeX errors introduced
[ ] All new citations are in .bib file
[ ] Self-review score improved or maintained
[ ] Revision summary cross-references every concern

Revision Summary Template

# Revision Summary

## Changes Overview
- N sections modified
- M new experiments added
- K new figures/tables added

## Detailed Changes

### Reviewer 1, Concern 1 (Major)
**Concern**: [what they said]
**Action**: [what we did]
**Location**: Section X, paragraph Y
**New content**: [brief description]

### Reviewer 1, Concern 2 (Minor)
...

## Score Comparison
| Metric | Before | After |
|--------|--------|-------|
| Overall | 5 | 7 |
| Clarity | 6 | 8 |
| ...    | ... | ... |