paper-rebuttal
Compare original and translation side by side
🇺🇸
Original
English🇨🇳
Translation
ChinesePaper Rebuttal
论文反驳信写作
A systematic approach to writing rebuttals after receiving peer review feedback. The goal is not to defend every point — it's to move scores by addressing the concerns that actually drive them.
本指南系统介绍了收到同行评审反馈后撰写反驳信的方法。核心目标并非为每个观点辩护——而是通过解决真正影响评分的问题来提升分数。
When to Use This Skill
本技能适用场景
- User received reviewer comments and needs to write a rebuttal
- User asks how to respond to specific reviewer criticism
- User wants to analyze reviews strategically before responding
- User mentions "rebuttal", "reviewer comments", "review feedback", "respond to reviewers"
For pre-submission self-review and catching weaknesses before they become reviewer complaints, use theskill.paper-review
- 用户收到评审意见，需要撰写反驳信
- 用户询问如何回应对特定评审批评
- 用户希望在回应前对评审意见进行策略性分析
- 用户提及“反驳信”“评审意见”“评审反馈”“回复评审人”
若需在投稿前进行自我评审，提前发现可能被评审人指出的问题，请使用技能。paper-review
Step 1: Diagnose Reviews
步骤1：诊断评审意见
Before writing a single word, answer: "Why did this reviewer give this exact score?" Not what they wrote — what drove the score. Most researchers skip this and address every comment equally. That is a mistake.
在动笔前，请先回答：“这位评审人为何给出这个具体分数？” 不是看他们写了什么，而是看驱动分数的核心原因。大多数研究者会跳过这一步，平等对待每条意见——这是错误的做法。
Score Diagnosis
评分诊断
For each reviewer, ask: "What would move this reviewer from their current score to acceptance?"
| Score Range | Typical Situation | Your Strategy |
|---|---|---|
| 7+ | Already your champion | Arm them with ammunition for the discussion phase |
| 5-6 | On the fence, 1-2 concerns holding them back | Identify and resolve those specific concerns |
| 3-4 | Fundamental objection | Determine if the objection is addressable; if not, focus elsewhere |
针对每位评审人，思考：“如何让这位评审人从当前分数改为同意录用？”
| 评分区间 | 典型情况 | 应对策略 |
|---|---|---|
| 7分及以上 | 已经是你的支持者 | 为他们提供讨论阶段可用的论据 |
| 5-6分 | 持观望态度，1-2个问题阻碍其给出高分 | 识别并解决这些特定问题 |
| 3-4分 | 存在根本性异议 | 判断该异议是否可解决；若不可行，将精力集中在其他方面 |
Color-Code Every Comment
为每条意见标注颜色
Read through each review and mark every comment:
| Color | Meaning | Action | Budget |
|---|---|---|---|
| Red | Score-driving concern — this is why the score is low | Address first, maximum effort and evidence | 60% |
| Orange | Addressable concern — can be resolved | Respond with concrete data or revision | 30% |
| Gray | Minor or cosmetic | Acknowledge briefly, confirm fix | 10% |
| Green | Positive comment or praise | Note as ammunition for your champion | — |
通读所有评审意见，为每条意见标记颜色：
| 颜色 | 含义 | 行动 | 精力分配 |
|---|---|---|---|
| 红色 | 影响评分的核心问题——这是分数偏低的原因 | 优先处理，投入最多精力和证据 | 60% |
| 橙色 | 可解决的问题——能够进行修正 | 提供具体数据或修订方案回应 | 30% |
| 灰色 | 次要或格式类小问题 | 简要致谢并确认将修复 | 10% |
| 绿色 | 正面评价或赞扬 | 记录下来作为支持者的论据 | — |
Identify the Invisible Question
挖掘隐藏的问题
Behind every reviewer comment is an unspoken question. A comment like "The baselines are outdated" really asks: "Is this method actually competitive with current approaches?" Address the invisible question, not just the surface request.
每条评审意见背后都有未明说的问题。比如“基线已过时”实际上是在问：“该方法是否能与当前主流方法竞争？”要回应隐藏的问题，而非仅停留在表面要求。
Step 2: Plan Response Strategy
步骤2：制定回应策略
Categorize Every Concern
对所有问题分类
| Category | Response Strategy |
|---|---|
| Misunderstanding | Clarify with specific references to the paper; restate the key point |
| Missing experiment | Provide the experiment inline if feasible; otherwise explain constraints honestly |
| Missing baseline | Add comparison or explain precisely why the baseline is not applicable |
| Writing clarity | Acknowledge and provide revised text in the rebuttal |
| Fundamental concern | Address directly with technical arguments AND additional evidence |
| Minor issue | Thank the reviewer and confirm the fix |
| 类别 | 回应策略 |
|---|---|
| 误解类 | 结合论文中的具体内容进行澄清，重申核心观点 |
| 缺少实验 | 若可行，直接在反驳信中提供实验结果；否则如实说明限制条件 |
| 缺少基线对比 | 补充对比，或详细说明为何该基线不适用 |
| 写作清晰度问题 | 表示认可，并在反驳信中提供修订后的文本 |
| 根本性异议 | 用技术论证和额外证据直接回应 |
| 次要问题 | 感谢评审人并确认将修复 |
Identify Common Themes
识别共性问题
If multiple reviewers raise the same concern, it's almost certainly a real weakness. Consolidate these into a "Common Response" section — this saves word count and demonstrates that you understand the pattern.
如果多位评审人提出相同问题，那几乎可以肯定这是论文的真实缺陷。将这些问题整合到“共性回应”部分——这既能节省字数，也能体现你已发现问题的共性。
Distinguish Actionable vs. Subjective
区分可执行问题与主观问题
- Actionable: "Missing comparison with Method X" — you can do this
- Subjective: "The novelty is limited" — harder to address, but can be reframed with evidence
- 可执行问题：“缺少与方法X的对比”——你可以补充该内容
- 主观问题：“创新性不足”——较难回应，但可以用证据重新阐释
The Champion Strategy
支持者策略
Your rebuttal's real audience is not the negative reviewer — it's the positive one.
Your champion argues on your behalf in the AC discussion, often using your exact words. Write your rebuttal to arm them:
- Make key arguments copy-pasteable — your champion will quote you directly
- Highlight where reviewers agree with each other — consensus strengthens the champion's position
- Flag contradictions between reviewers — if R1 says "limited novelty" but R2 says "interesting approach," your champion can use this
- Lead with strengths before weaknesses — remind the AC what your paper does well
See references/rebuttal-tactics.md for the full 18 tactical rules.
你的反驳信真正的受众并非给出负面评价的评审人——而是支持你的评审人。
你的支持者会在领域主席（AC）的讨论中为你辩护，且常常会直接引用你的话。撰写反驳信时要为他们提供有力支持：
- 让核心论点便于复制粘贴——你的支持者会直接引用你的话
- 突出评审人之间的共识——共识能强化支持者的立场
- 指出评审人之间的矛盾——如果评审人1说“创新性不足”但评审人2说“方法有趣”，你的支持者可以利用这一点
- 先讲优势再讲不足——提醒领域主席（AC）你的论文的亮点
完整的18条写作技巧请参考references/rebuttal-tactics.md。
Step 3: Write the Rebuttal
步骤3：撰写反驳信
Structure
结构
- Opening: One line thanking reviewers (keep it short)
- Common concerns: Address issues raised by multiple reviewers first — these are highest priority
- Per-reviewer responses: Address remaining concerns in priority order (red → orange → gray), NOT in the order the reviewer wrote them
- 开头：用一句话感谢评审人（保持简短）
- 共性问题：优先回应多位评审人提出的问题——这些是最高优先级
- 分评审人回应：按优先级（红色→橙色→灰色）而非评审人原有的顺序，回应剩余问题
Per-Concern Format
单条问题回应格式
For each concern, follow this three-part structure:
- Acknowledge: Show you understand the concern (one sentence)
- Respond: Provide your answer — evidence, clarification, new experiment results
- Action: State what you changed in the revision (specific section/table/figure)
Use a fillable template at assets/rebuttal-template.md.
对每个问题，遵循以下三部分结构：
- 认可：表明你理解该问题（一句话）
- 回应：给出你的答案——证据、澄清、新实验结果
- 行动：说明你在修订版中修改了哪些内容（具体章节/表格/图）
可使用assets/rebuttal-template.md中的可填写模板。
The Neutral Third-Party Test
第三方中立测试
Before submitting, have someone who hasn't read your paper read only the reviews and your rebuttal. Ask: "Can you tell whether the concerns were addressed?" If not, rewrite.
提交前，找一位未读过你论文的人，只让他看评审意见和你的反驳信，然后问：“你能看出问题是否已得到回应吗？”如果不能，重新撰写。
Counterintuitive Rebuttal Principles
反直觉的反驳原则
-
Submit a rebuttal even with extreme scores. A paper with scores of 3/8/8 has better odds than you think. The negative reviewer may realize they are an outlier during discussion. But only if you submit a rebuttal — without one, the AC has nothing to work with.
-
Concede something small, win something big. Acknowledging a minor weakness ("We agree that Table 2 could include dataset X for completeness") makes your defense of major points more credible. Pure defense with zero concession reads as unobjective.
-
One new experiment beats three paragraphs of explanation. Reviewers are trained to be skeptical of arguments. They are not trained to be skeptical of data. A small new experiment that directly addresses a concern is worth more than any amount of reasoning.
-
The best rebuttal is written before submission. Draft responses to likely attacks while writing the paper ("prebuttal"). Two benefits: you often realize the attack is valid and fix the paper, and if the attack comes, you have a polished response ready.
-
Don't defend every point equally. Equal effort signals you don't know which points matter. Allocate your word budget according to the color-coding: 60% red, 30% orange, 10% gray. Reviewers notice when you nail the big issues.
- 即便分数极端也要提交反驳信。一篇分数为3/8/8的论文，录用概率比你想象的高。负面评价的评审人可能会在讨论阶段意识到自己是少数派。但前提是你要提交反驳信——如果不提交，领域主席（AC）将没有依据进行判断。
- 做出小让步，换取大成果。承认一个小缺陷（如“我们认可表2可以补充数据集X以保证完整性”）会让你对核心问题的辩护更可信。完全不做任何让步的辩护会显得不够客观。
- 一个新实验胜过三段解释。评审人受过训练，对论点持怀疑态度，但不会对数据持怀疑态度。一个直接回应问题的小型新实验，价值远超任何长篇大论的推理。
- 最佳反驳信在投稿前就已撰写。撰写论文时就针对可能的质疑草拟回应（“预反驳”）。这样做有两个好处：你通常会意识到质疑是合理的，从而修改论文；如果真的收到该质疑，你已有打磨好的回应。
- 不要对所有问题投入同等精力。同等投入意味着你不知道哪些问题才是关键。按照颜色编码分配精力：60%给红色问题，30%给橙色问题，10%给灰色问题。评审人会注意到你是否解决了核心问题。
Common Reviewer Concerns
常见评审意见
Prepare responses for these frequent concerns. Having a prepared response doesn't mean copying it verbatim — adapt to your specific paper and the reviewer's specific framing.
| Common Concern | Response Strategy |
|---|---|
| "Limited novelty" | Articulate the specific insight; show what prior work cannot do; narrow and sharpen the claim |
| "Marginal improvement" | Emphasize other advantages (speed, generalizability, simplicity); add challenging test cases |
| "Missing ablations" | Provide the ablation table inline in the rebuttal |
| "Missing baselines" | Add the comparison or explain precisely why it's not applicable |
| "Not reproducible" | Add implementation details; commit to code release with a specific timeline |
| "Limited evaluation" | Add diverse datasets or metrics; if infeasible, explain resource constraints honestly |
| "No limitation discussed" | Add a limitation section in the revision; acknowledge this was an oversight |
| "Overclaimed results" | Weaken specific claims to match evidence; show the revised wording |
| "Unfair comparison" | Use standard evaluation protocols; add commonly reported baselines |
| "Method is engineering, not research" | Identify the scientific insight behind the design; explain why the choice is non-obvious |
| "Metrics don't match claims" | Align each claim with a specific metric; add the missing metric if feasible |
| "Related work incomplete" | Add the missing references; explain the relationship to your work |
Need to run new experiments for the rebuttal? Use theskill for targeted debugging, orexperiment-craftfor a full new experiment stage.experiment-pipeline
提前准备针对这些常见问题的回应。准备好的回应并非让你直接复制粘贴——要根据你的论文和评审人的具体表述进行调整。
| 常见评审意见 | 回应策略 |
|---|---|
| “创新性不足” | 明确阐述具体的创新点；展示现有工作无法实现的内容；缩小并强化你的核心主张 |
| “改进幅度有限” | 强调其他优势（速度、通用性、简洁性）；补充更具挑战性的测试用例 |
| “缺少消融实验” | 在反驳信中直接提供消融实验表 |
| “缺少基线对比” | 补充对比，或详细说明为何该基线不适用 |
| “无法复现” | 补充实现细节；承诺在特定时间范围内发布代码 |
| “评估范围有限” | 补充多样化的数据集或指标；若不可行，如实说明资源限制 |
| “未讨论局限性” | 在修订版中添加局限性章节；承认这是之前的疏漏 |
| “结果夸大” | 弱化特定主张以匹配证据；展示修订后的表述 |
| “对比不公平” | 使用标准评估协议；补充常用的基线对比 |
| “方法属于工程实现，而非研究” | 识别设计背后的科学见解；解释为何该选择并非显而易见 |
| “指标与主张不匹配” | 让每个主张对应具体的指标；若可行，补充缺失的指标 |
| “相关工作不完整” | 补充缺失的参考文献；说明这些工作与你的研究的关系 |
需要为反驳信开展新实验？ 若需针对性调试，请使用技能；若需完成完整的新实验阶段，请使用experiment-craft技能。experiment-pipeline
Handoff from Paper Review
与Paper Review技能的衔接
This skill picks up where leaves off. If you used before submission, these artifacts are especially useful for rebuttal:
paper-reviewpaper-review| Artifact from paper-review | How It Helps Rebuttal |
|---|---|
| Reject-first simulation | You've already anticipated likely attacks |
| Claim-evidence audit table | Quickly verify whether a reviewer's concern about unsupported claims is valid |
| Prebuttal drafts (Phase 6) | Ready-made response templates for common criticisms |
| Trust scorecard | Identifies weaknesses you can proactively concede |
本技能承接技能的工作。如果你在投稿前使用过技能，以下产出物对撰写反驳信尤为有用：
paper-reviewpaper-review| Paper Review技能产出物 | 对反驳信的帮助 |
|---|---|
| 拒稿模拟 | 你已提前预判了可能的质疑 |
| 主张-证据审核表 | 快速验证评审人关于“主张无证据支持”的质疑是否成立 |
| 预反驳草稿（第6阶段） | 针对常见批评的现成回应模板 |
| 可信度评分卡 | 识别可主动承认的缺陷 |
Reference Navigation
参考资源导航
| Topic | Reference File | When to Use |
|---|---|---|
| 18 tactical rules | rebuttal-tactics.md | Detailed writing guidance for structure, content, tone |
| Rebuttal template | rebuttal-template.md | Starting a new rebuttal document |
| 主题 | 参考文件 | 适用场景 |
|---|---|---|
| 18条写作技巧 | rebuttal-tactics.md | 结构、内容、语气方面的详细写作指导 |
| 反驳信模板 | rebuttal-template.md | 开始撰写新的反驳信文档时 |