kinhluan Internal Critique
Simulate a tough-but-fair reviewer before anyone else sees your work.
Reviewer Mindset
Top conference reviewers ask:
- Is this problem important and unsolved?
- Is the proposed method novel?
- Are experiments convincing and fair?
- Are claims supported by evidence?
- Is the writing clear enough to follow?
Critique Protocol
Severity Grading System
Every issue must be assigned a severity level. This determines priority and revision strategy.
| Severity | Symbol | Definition | Revision Strategy |
|---|---|---|---|
| Critical | 🔴 | Paper would be rejected. Fundamental flaw in contribution, methodology, or ethics. | Must fix before any submission |
| Major | 🟡 | Significant weakness requiring substantial revision. Weakens core claims. | Fix before submission; may need new experiments |
| Minor | 🟢 | Polish issue. Doesn't affect scientific validity but hurts readability. | Fix if time permits; acceptable for initial submission |
| Suggestion | 🔵 | Optional improvement. Would strengthen but not required. | Address if reviewer raises it |
Level 1 — Critical Issues (🔴 Reject if any present)
Check each. If yes → fix before any sharing.
- Contribution unclear: Can't state in 2 sentences what is new
- Baselines unfair: Missing key baselines, or baselines disadvantaged
- No statistical significance: Single-run results with no std dev
- Claims unsupported: Results table doesn't support abstract claims
- Reproducibility impossible: Hyperparameters missing, no code planned
- Overclaiming: "state-of-the-art" without comprehensive comparison
- Ethical concern: Missing IRB approval, biased dataset, harmful application
- Plagiarism risk: Uncited verbatim text, unattributed figures
Level 2 — Major Weaknesses (🟡 Major revision territory)
- Ablation missing: Don't know which component causes improvement
- Dataset too small/narrow: Only 1 dataset, too easy, not standard
- Hyperparameter sensitivity unknown: Did you tune on test set?
- Limitations not discussed: Honest papers discuss failure cases
- Related work gaps: Missing the 2-3 most relevant papers
- Theoretical grounding weak: Claims lack theoretical justification
- Generalization untested: Only tested on one domain/setting
Level 3 — Minor Issues (🟢 Polish)
- Notation inconsistency
- Figures unclear (missing axis labels, legend, caption)
- Paragraph without clear main point
- Abstract doesn't match paper content
- Future work too vague
- Grammar/spelling errors
- Citation format inconsistent
Level 4 — Suggestions (🔵 Optional)
- Additional analysis could strengthen claims
- Broader impact statement could be expanded
- Visualization could be more intuitive
- Code could be better documented
Section-Specific Review
Introduction
- Hook motivates real problem?
- Gap is specific, not "performance is important"?
- Contributions are concrete (numbers, not adjectives)?
Related Work
- Papers organized by theme, not dump of citations?
- Each cluster clearly differentiated from yours?
- Most recent papers included (within 2 years)?
Methodology
- Problem formally defined (notation, objective)?
- Algorithm reproducible from text alone?
- Assumptions stated explicitly?
Experiments
- Baselines are current SOTA?
- Same compute budget for all methods?
- Results table complete (all methods × all metrics)?
- Ablation covers all key design choices?
- At least 1 analysis beyond main table?
Writing
- Abstract ≤ 250 words?
- No paragraph > 8 sentences?
- No hedging words (very, clearly, obviously)?
- Every claim has citation or experimental support?
Output Format
Produce a structured critique report with severity grading:
## Internal Critique: [Paper Title]
### 🔴 Critical Issues (Must Fix)
| # | Issue | Location | Suggested Fix | Effort |
|---|-------|----------|---------------|--------|
| 1 | [description] | [section] | [action] | [hours] |
### 🟡 Major Weaknesses (Should Fix)
| # | Issue | Location | Suggested Fix | Effort |
|---|-------|----------|---------------|--------|
| 1 | [description] | [section] | [action] | [hours] |
### 🟢 Minor Issues (Fix if Time)
| # | Issue | Location | Suggested Fix | Effort |
|---|-------|----------|---------------|--------|
| 1 | [description] | [section] | [action] | [minutes] |
### 🔵 Suggestions (Optional)
| # | Suggestion | Benefit if Implemented |
|---|------------|------------------------|
| 1 | [description] | [impact] |
### Strongest Points
- [what works well — helps identify what to protect in revisions]
### Overall Assessment
| Criterion | Score (1-5) | Notes |
|-----------|-------------|-------|
| Contribution Clarity | [score] | |
| Methodological Rigor | [score] | |
| Experimental Validation | [score] | |
| Writing Quality | [score] | |
| Reproducibility | [score] | |
**Ready for:** advisor / workshop / top venue / needs more work
**Estimated revision time:** [X hours/days]
**Priority fixes:** [top 3 actions]
Simulated Reviewer Comments
For each weakness, write in reviewer voice:
"The authors claim X but only demonstrate Y. The baseline Z is missing, which is the strongest published method on this task. Without this comparison, the improvement claim is not convincing."
Then write your rebuttal response:
"We thank the reviewer. We will add Z as baseline. Preliminary results show our method still outperforms (Table attached). We will include in final version."
This prepares you for actual rebuttal if the paper gets reviewed.
Links to Other Skills
- Requires →
paper-writing(draft to review) - Feeds into →
publication-strategy(after critique, choose venue) - Can iterate back to →
experiment-tracking(if more experiments needed)