vr-peer-review

name	vr-peer-review
description	Critically reviews a written paper from the perspective of a top-venue reviewer, providing strengths/weaknesses analysis and improvement suggestions.
user-invocable	true
argument-hint	["paper-directory"]

Peer Review

Systematically reviews a written paper from the perspective of a top-venue reviewer.

Input

$ARGUMENTS - Paper directory path (e.g., "papers/paper_20260305_llm_reasoning")

Process

Step 1: Read the Entire Paper

• Read main.tex and all section files
• Examine figures and tables
• Review references.bib

Step 2: Summary

Summarize the core of the paper in 3-5 sentences:

• The problem being addressed
• The proposed method
• Key results

Step 3: Strengths Analysis

Describe 3-5 strengths in detail:

• S1: [Strength title] - [Specific description and supporting evidence]
• Support each strength by citing specific parts of the paper
• Evaluation criteria: novelty, technical soundness, significance, clarity, reproducibility

Step 4: Weaknesses Analysis

Describe 3-5 weaknesses in detail:

• W1: [Weakness title] - [Specific description and suggested improvement]
• Provide a concrete actionable suggestion for each weakness
• Weakness types: missing baselines, insufficient analysis, unclear motivation, weak evaluation, overclaiming, etc.

Baseline Completeness Check

Read .claude/skills/_shared/paper-search-protocol.md and execute a Level 2 search focused on: "recent methods addressing [paper's problem] that could serve as missing baselines or competing approaches". Flag any strong recent method not cited or compared against.

Step 5: Questions to Authors

3-5 questions for the authors:

• Intent behind experimental design
• Validity of result interpretation
• Details of the methodology
• Generalizability

Step 6: Minor Comments

Detailed corrections:

• Typos, grammatical errors
• Suggestions for improved phrasing
• Reference errors (\ref, \cite)
• Figure/table caption improvements
• Layout issues

Step 7: Overall Assessment

• Overall Score: 1-10 (1: strong reject, 10: strong accept)
  • 1-3: Reject
  • 4-5: Borderline reject
  • 6-7: Borderline accept
  • 8-10: Accept
• Confidence: 1-5 (1: not an expert in this field, 5: expert in this field)
• Recommended Decision: Accept / Weak Accept / Borderline / Weak Reject / Reject

Output

Save the results in the paper directory.

File Structure

papers/<paper-dir>/
└── review.md               # Review report

review.md Format

# Peer Review: [Paper Title]
Date: [YYYY-MM-DD]
Reviewer: AI Reviewer (Claude)

## Summary
[3-5 sentence summary]

## Strengths
- **S1: [Title]** - [Description]
- **S2: [Title]** - [Description]
- **S3: [Title]** - [Description]

## Weaknesses
- **W1: [Title]** - [Description] → *Suggested improvement: [suggestion]*
- **W2: [Title]** - [Description] → *Suggested improvement: [suggestion]*
- **W3: [Title]** - [Description] → *Suggested improvement: [suggestion]*

## Questions to Authors
1. [Question]
2. [Question]
3. [Question]

## Minor Comments
- [Location]: [Comment]
- [Location]: [Comment]

## Overall Assessment
- **Score**: X/10
- **Confidence**: X/5
- **Decision**: [Accept / Weak Accept / Borderline / Weak Reject / Reject]

### Key Improvements (Priority)
1. [Most important improvement]
2. [Second most important improvement]
3. [Third most important improvement]

## Next Steps
Based on the review results, run `/vr-rebuttal <paper-dir>` to write a rebuttal, or revise the paper directly.

Notes

• Follow review standards of top venues such as NeurIPS, ICML, ICLR, ACL, etc.
• Provide constructive and specific feedback (avoid mere criticism)
• After the review, discuss with the user which feedback to incorporate
• If needed, suggest additional experiments and connect to /vr-ablation-study