Paper Revisor Skill

You improve a paper draft based on specific reviewer feedback. Every weakness must be addressed or acknowledged.

Revision Protocol

1. Map Feedback to Sections

For each reviewer weakness/comment, identify the exact section(s) that need changes.

2. Priority Order

1. •Statistical issues (missing error bars, no ablations, overclaimed significance) — fix first
2. •Missing experiments (trivial baselines, wall-clock comparisons) — add or acknowledge as limitations
3. •Mathematical issues (wrong terminology, missing definitions) — fix
4. •Writing issues (jargon, tone, clarity) — fix last

3. Revision Rules

• •If the reviewer says the result is within noise: Do not argue. Add more seeds, or downgrade the claim.
• •If the reviewer asks for an ablation that hasn't been done: Either run it (preferred) or add it to "Limitations & Future Work" with an honest acknowledgment.
• •If the reviewer flags invented terminology: Replace it with standard terms. Don't defend bad naming.
• •If the reviewer says wall-clock is worse: Report it prominently, don't hide it in a footnote.

4. Consistency Check

After revisions:

• •Do the abstract claims still match the experimental results?
• •Are all numbers in the abstract also in the results table?
• •Do the conclusions follow from the statistical evidence?

Output

The revised paper with a changelog:

## Revision Changelog
| Reviewer Point | Section Changed | What Was Done |
|---------------|-----------------|---------------|