| name | openbs-rebuttal |
| description | Write a conference paper rebuttal following Prof. Bingsheng He's (NUS) proven rebuttal strategies. Analyzes reviews, categorizes concerns, drafts structured responses with evidence, and prepares confidential comments to SPC/AC. Use after receiving paper reviews. |
| disable-model-invocation | true |
| argument-hint | [reviews-file-path or paste reviews] |
| allowed-tools | Read, Grep, Glob, Bash |
Prof. Bingsheng He's Rebuttal Writing Skill
You are helping write a rebuttal following the proven strategies of Prof. Bingsheng He's research group at NUS, whose work appears at top database, systems, and ML venues (SIGMOD, VLDB, ICDE, and venues such as CVPR/TKDE for their federated-learning line).
Sourcing note. raintreebook has no standalone rebuttal chapter; the rebuttal logistics come from Ch. 2.3 ("During Review and Rebuttal"), and the mindset is consistent with his reviewer experience (SIGMOD Distinguished PC 2022). For the fuller collection of stories and mottos, read ${CLAUDE_SKILL_DIR}/../../references/stories-and-philosophy.md (Glob to locate if needed).
Read the reviews provided in $ARGUMENTS (file path or pasted text). Then follow the strategy below.
BEFORE YOU WRITE: the co-author protocol (raintreebook Ch. 2.3)
A rebuttal is a team document. Prof. He's group process:
- Forward ALL reviews to every co-author immediately. Submission systems often notify only the contact author -- do not let co-authors learn the outcome late.
- Collect input from all co-authors BEFORE drafting. The senior authors often see the winning framing you missed.
- Recirculate the finalized rebuttal to everyone before submitting. No surprises, ever.
- Remember the long game: "Do not upset the reviewers -- you are very likely to meet them at a conference; they can be your future employer or interviewer." Even a rejection should leave a stronger paper and no burned bridges.
CORE PHILOSOPHY
Prof. He's central rebuttal principle:
"We need to convince Reviewers to upgrade their rating."
Key mindset rules:
- Reviewers believe they are smart and that they understand the paper. Top conferences accept only 20-30%.
- It is psychologically difficult for a reviewer to change their rating ("losing face"). You must make it easy for them by providing irrefutable evidence.
- You cannot change the reviewer; you must change yourself (and your paper). Do not argue the reviewer is wrong in the abstract. Provide evidence.
- Even if the reviewer doesn't upgrade, other reviewers and the AC will see the rebuttal during discussion. A strong rebuttal helps regardless.
- Worst case: you now have a stronger paper for the next venue.
STEP 1: Analyze and Categorize Each Review
For each reviewer, extract and categorize every concern into:
Category A: Reviewer is factually wrong
Strategy: Provide evidence (references to other top publications) or proof (mathematical proof or experimental evidence). Be polite but firm. Use specific page/line numbers to show where the misunderstanding occurred.
Category B: Missing experiments or results
Strategy: Provide the results IN the rebuttal itself -- do NOT merely promise to add them in revision. The implicit message: "I can address your concern in one day, so it is not a major issue."
Category C: Scope or relevance questions
Strategy: Acknowledge the broader context, explain why your scope is appropriate, and frame expansions as "interesting future work."
Category D: Presentation or clarity issues
Strategy: Acknowledge graciously and describe specific changes you will make. Quote the revised text if space allows.
Category E: Fundamental technical concerns
Strategy: Address head-on with detailed explanations, examples, and/or additional analysis.
STEP 2: Draft the Rebuttal
Opening
Start with a polite but purposeful opening:
"We thank the reviewers for their valuable feedback. We sincerely hope you can carefully read our responses and adjust the score if appropriate."
Structure for Each Reviewer
Address each reviewer separately with clearly labeled Q/A pairs:
## Response to Reviewer #X
**Q1 (Category): [Summarize the concern in one sentence]**
[Your response with evidence, data, or references]
**Q2 (Category): [Next concern]**
[Response]
Response Patterns from Prof. He's Group
When correcting a factual error:
"We respectfully note that [specific concept] works differently from what is described. Specifically, [explanation with references]. As stated in [Author, Venue Year], [supporting quote]. We have clarified this in Section X, Page Y of the revised paper."
When providing new experimental results:
Include actual tables/numbers directly in the rebuttal:
"We have conducted the requested experiments. The results are shown below:
[Table with actual numbers]
These results confirm that [conclusion]."
When a concern is out of scope:
"We agree that [broader topic] is an important research direction. However, as noted in [reference], a comprehensive study of [topic] would itself constitute a separate paper. Our work focuses specifically on [your scope], which is the primary challenge in [your domain]."
When acknowledging a limitation:
"We acknowledge that [limitation]. However, our empirical results on [datasets] show that [limitation has minimal impact in practice]. Achieving guarantees for [ideal solution] warrants further research, and we have added this discussion to Section X."
When notation or presentation caused confusion:
"We apologize that the usage of [notation] may have caused some confusion. We have revised the presentation in Section X to clarify [specific change]."
STEP 3: Draft Confidential Comments to SPC/AC
If any reviewer has a clear factual misunderstanding, draft a confidential comment following Prof. He's template:
"We would like to bring to the attention of the SPC that Reviewer #X appears to have [specific misunderstanding]. Specifically, [evidence with page numbers]. [Explain the error factually, not emotionally].
We are NOT complaining about the reviews and misunderstanding. We appreciate that reviewers have spent a lot of effort. However, we do hope that our paper can receive a fair chance."
Use this ONLY when there is clear, provable factual error by a reviewer. Keep it factual, not emotional.
STEP 4: Post-Rebuttal Strategy
Suggest:
- If the venue has a discussion period, recommend sending a polite reminder 2 days before discussion closes:
"As the discussion period is coming to a close in just TWO days, we would appreciate your prompt response regarding our rebuttal."
- Identify which reviewers are most likely to change their rating and focus energy accordingly.
OUTPUT FORMAT
## Rebuttal Analysis Summary
### Review Score Summary
| Reviewer | Score | Key Concerns | Likely to Upgrade? |
|----------|-------|-------------|-------------------|
### Concern Classification
[Table mapping each concern to Category A-E]
---
## Draft Rebuttal
[Full rebuttal text ready for submission]
---
## Confidential Comments to SPC/AC (if applicable)
[Draft confidential comment]
---
## Post-Rebuttal Strategy
[Recommendations for discussion period]
IMPORTANT REMINDERS
- Word/page limits: Check the venue's rebuttal length limit and respect it strictly.
- Tone: Always respectful, never defensive or hostile. Prof. He's group uses phrases like "We respectfully note..." and "We appreciate the reviewer's insight..."
- Evidence over promises: Always show results, never just promise them.
- Cascading benefit: Even concerns you cannot fully address should be acknowledged gracefully -- the AC reads everything.
- AI-era use (augment, don't automate): AI is useful to categorize concerns, draft candidate response wording, and surface the strongest counter-evidence to hunt for. But YOU decide the strategy, and every experimental number in the rebuttal must be real and yours. A polished-but-empty rebuttal (promises dressed as results) is transparent to an experienced AC. See
/openbs-ai-era.
