Back to Details

paper-review

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

Academic Manuscript Review

学术手稿评审

Provides structured, rigorous peer review of academic manuscripts. Reviews prioritize methodological soundness, statistical validity, logical consistency, and reproducibility.
Note on review calibration: This skill reflects an opinionated review style that prioritizes methodological precision, statistical rigor, and reproducibility. It is direct, evidence-based, and holds manuscripts to high standards. The severity calibration (Critical, Major, Minor) follows a strict hierarchy: Critical issues block publication; Major issues require significant revision; Minor issues improve polish. Reviewers using this skill should adapt the tone and depth to their own standards and the target journal's expectations.
为学术手稿提供结构化、严谨的同行评审服务。评审重点关注方法学合理性、统计有效性、逻辑一致性和可重复性。
评审校准说明：本技能采用带有明确倾向的评审风格，优先关注方法学精度、统计严谨性和可重复性。评审结果直接、基于证据，并对手稿设定高标准。问题严重程度（关键、主要、次要）遵循严格层级：关键问题会阻碍发表；主要问题需要大幅修改；次要问题用于提升文稿打磨度。使用本技能的评审员应根据自身标准和目标期刊的期望调整评审语气和深度。

When to Use

使用场景

Activate when the user wants peer-review feedback on a manuscript (journal article, conference paper, preprint) evaluated for methodological soundness, statistical validity, and clarity of presentation. The output is a structured review with categorized concerns and constructive suggestions.
当用户需要对手稿（期刊文章、会议论文、预印本）进行同行评审反馈，评估其方法学合理性、统计有效性和表述清晰度时，激活本技能。输出结果为包含分类问题和建设性建议的结构化评审报告。

Manuscript Intake

手稿接收

Manuscripts for peer review are typically provided as PDFs from journal submission systems. Convert to both markdown and PNG for a complete review: markdown for efficient text analysis, PNG for exact page/line citations and figure inspection.
PDF (most common): Use a hybrid approach: convert to both markdown and PNG. Markdown gives efficient searchable text for content analysis; PNG preserves exact page layout, line numbers, and figure positions for precise citations.
Step 1: Convert to markdown for text analysis:
bash
uvx opencite convert manuscript.pdf -o manuscript.md
Step 2: Convert to PNG for page/line references and figure inspection:
bash
uv run --with pdf2image --with pillow python -c "
from pdf2image import convert_from_path
pages = convert_from_path('manuscript.pdf', dpi=200)
for i, page in enumerate(pages):
    page.save(f'manuscript_page_{i+1}.png', 'PNG')
"
Note: requires poppler (
brew install poppler
on macOS, 
apt install poppler-utils
on Linux). Alternatively, use 
pdftoppm -png -r 200 manuscript.pdf manuscript_page
.
Workflow: Read the markdown for content review (methods, statistics, logic, literature). When citing a specific issue, refer to the PNG pages to provide exact page and line numbers (e.g., "page 4, line 23" or "p4 l23"). Use the PNGs to inspect figures, tables, and overall layout.
For large PDFs (>10 pages), read PNGs in batches as needed.
Markdown or LaTeX: Read directly; no conversion needed.
Read all sections including supplementary materials, appendices, and figures. Note the target journal if known, as expectations differ across venues (transactions vs. letters vs. conference proceedings).
用于同行评审的手稿通常以期刊投稿系统中的PDF格式提供。需将其转换为markdown和PNG两种格式以完成完整评审：markdown用于高效文本分析，PNG用于精确的页码/行号引用和图表检查。
PDF（最常见格式）：采用混合处理方式：同时转换为markdown和PNG。markdown提供可搜索的文本内容，便于内容分析；PNG保留精确的页面布局、行号和图表位置，用于精准引用。
步骤1：转换为markdown 用于文本分析：
bash
uvx opencite convert manuscript.pdf -o manuscript.md
步骤2：转换为PNG 用于页码/行号引用和图表检查：
bash
uv run --with pdf2image --with pillow python -c "
from pdf2image import convert_from_path
pages = convert_from_path('manuscript.pdf', dpi=200)
for i, page in enumerate(pages):
    page.save(f'manuscript_page_{i+1}.png', 'PNG')
"
注意：需要安装poppler（macOS系统使用
brew install poppler
，Linux系统使用
apt install poppler-utils
）。或者使用
pdftoppm -png -r 200 manuscript.pdf manuscript_page
命令。
工作流程：阅读markdown版本进行内容评审（方法、统计、逻辑、文献）。当引用具体问题时，参考PNG页面提供精确的页码和行号（例如：“第4页，第23行”或“p4 l23”）。使用PNG版本检查图表、表格和整体布局。
对于篇幅较长的PDF（超过10页），可按需分批阅读PNG页面。
Markdown或LaTeX格式：直接阅读，无需转换。
阅读所有内容，包括补充材料、附录和图表。若已知目标期刊，需注意不同期刊的期望存在差异（如汇刊、快报、会议论文集）。

Review Process

评审流程

1. Read the full manuscript

1. 通读完整手稿

Read everything: abstract, introduction, methods, results, discussion, conclusion, figures, tables, supplementary materials. Take note of:
  • The stated hypothesis or research question
  • The methods used to test the hypothesis
  • The statistical approach and sample size
  • The claims made in the discussion and conclusion
  • Whether figures and tables support the narrative
阅读全部内容：摘要、引言、方法、结果、讨论、结论、图表、表格、补充材料。记录以下内容：
  • 明确提出的假设或研究问题
  • 用于验证假设的研究方法
  • 统计方法和样本量
  • 讨论和结论中提出的观点
  • 图表和表格是否支持叙述内容

2. Assess methodological soundness

2. 评估方法学合理性

This is the core of the review. Evaluate using the checklist in 
references/methodology-checklist.md
. Key areas:
Experimental design:
  • Is the design appropriate for the research question?
  • Are controls adequate?
  • Is the sample size justified (power analysis, or at minimum acknowledged)?
  • Are inclusion/exclusion criteria clearly stated and justified?
  • Are there potential confounds that are not addressed?
Signal processing and data analysis (when applicable):
  • Are filtering parameters appropriate? Check Nyquist constraints: the analysis bandwidth must not exceed half the sampling rate (Nyquist frequency) and should not exceed the low-pass filter cutoff.
  • Are artifact rejection/correction methods validated for the specific data type?
  • Are analysis parameters (e.g., window lengths, frequency bands) justified?
  • Is there any "double-dipping" where the same data features used for selection/clustering are also the analysis target?
Statistical methods:
  • Are the chosen tests appropriate for the data distribution and design?
  • Are parametric assumptions tested (normality, homogeneity of variance)?
  • For paired vs. unpaired comparisons, is the correct test variant used?
  • Are main effects tested before post-hoc comparisons?
  • Are multiple comparisons corrected?
  • Are effect sizes reported, not just p-values?
  • Do small sample sizes warrant the statistical conclusions drawn?
  • Are figures appropriate for the data? (Bar plots with error bars for N<5 are misleading; use individual data points instead.)
这是评审的核心环节。使用
references/methodology-checklist.md
中的清单进行评估。关键领域包括：
实验设计：
  • 设计是否适合研究问题？
  • 对照设置是否充分？
  • 样本量是否合理（是否进行功效分析，或至少已说明）？
  • 纳入/排除标准是否明确且合理？
  • 是否存在未解决的潜在混杂因素？
信号处理与数据分析（如适用）：
  • 滤波参数是否合适？检查奈奎斯特约束：分析带宽不得超过采样率的一半（奈奎斯特频率），且不得超过低通滤波器的截止频率。
  • 伪迹剔除/校正方法是否针对特定数据类型进行了验证？
  • 分析参数（如窗口长度、频段）是否合理？
  • 是否存在“双重 dipping”问题：即用于选择/聚类的数据特征同时也是分析目标？
统计方法：
  • 所选检验方法是否适合数据分布和研究设计？
  • 是否检验了参数假设（正态性、方差齐性）？
  • 配对与非配对比较是否使用了正确的检验变体？
  • 是否在事后比较前检验了主效应？
  • 是否对多重比较进行了校正？
  • 是否报告了效应量，而非仅报告p值？
  • 小样本量是否能支撑得出的统计结论？
  • 图表是否适合展示数据？（样本量N<5时使用带误差棒的柱状图具有误导性；应改用单个数据点展示。）

3. Check logical consistency

3. 检查逻辑一致性

Trace the argument from introduction through methods to results and discussion:
  • Do the methods actually test the stated hypothesis?
  • Do the results support the claims made in the discussion?
  • Are conclusions proportional to the evidence? (Do not overreach.)
  • If the introduction frames a problem, do the methods address that exact problem?
  • Are terms and definitions used consistently throughout?
  • If a concept is introduced in the introduction, is it operationalized the same way in the methods?
Watch for contradictions: claims in the introduction that the authors' own methods cannot test, or discussion points that go beyond what the data show.
追踪从引言到方法、结果再到讨论的论证过程：
  • 研究方法是否真正验证了提出的假设？
  • 结果是否支持讨论中提出的观点？
  • 结论是否与证据相符？（不得过度延伸。）
  • 如果引言提出了某个问题，研究方法是否准确解决了该问题？
  • 术语和定义是否在全文中保持一致？
  • 如果引言引入了某个概念，方法部分是否以相同方式对其进行操作化定义？
注意矛盾点：引言中提出的观点作者自身方法无法验证，或讨论内容超出数据所能支撑的范围。

4. Evaluate literature coverage

4. 评估文献覆盖情况

  • Is the literature review current? (Check if key papers from the last 2-3 years are missing.)
  • Are the authors' claims supported by the cited literature, or do the cited papers actually argue otherwise?
  • Is related work from other groups or approaches acknowledged?
  • For the specific techniques used, are validation/limitation papers cited?
  • Are there relevant studies the authors should compare their results against?
Use opencite to verify literature claims and search for potentially missing references:
bash
uvx opencite search "topic keywords" --max 10 --sort citations
uvx opencite canonical "field or method" --max 5
When citing references in the review to support a methodological argument, include the full citation so the authors can verify the claim.
  • 文献综述是否与时俱进？（检查是否遗漏了过去2-3年的关键论文。）
  • 作者的观点是否得到引用文献的支持，或引用文献是否实际持相反观点？
  • 是否认可了其他团队或方法的相关研究？
  • 对于所使用的特定技术，是否引用了验证/局限性相关的论文？
  • 是否存在作者应将其结果与之对比的相关研究？
使用opencite验证文献观点，并搜索可能遗漏的参考文献：
bash
uvx opencite search "topic keywords" --max 10 --sort citations
uvx opencite canonical "field or method" --max 5
在评审中引用参考文献以支持方法学论证时，需提供完整引文，以便作者验证观点。

5. Check reproducibility and transparency

5. 检查可重复性与透明度

  • Are methods described in sufficient detail to reproduce?
  • Are data, code, and materials shared or is sharing addressed?
  • Are custom tools, software versions, and parameters specified?
  • For hardware or device papers: is enough detail provided (schematics, component lists, block diagrams) for independent reproduction?
  • Are conflicts of interest disclosed? Check author affiliations, patents, and commercial products related to the work.
  • 方法描述是否足够详细以支持重复实验？
  • 是否共享了数据、代码和材料，或提及了共享计划？
  • 是否指定了自定义工具、软件版本和参数？
  • 对于硬件或设备类论文：是否提供了足够细节（原理图、组件清单、框图）以支持独立复现？
  • 是否披露了利益冲突？检查作者 affiliations、专利和与研究相关的商业产品。

6. Evaluate figures and tables

6. 评估图表质量

Read each figure and table carefully:
  • Do figures accurately represent the data?
  • Are axes labeled, legends present, and units specified?
  • Are statistical annotations (significance stars, error bars) defined?
  • Are bar plots used appropriately? (For small N, show individual data points.)
  • Is the time/frequency scale appropriate for the data shown?
  • Do the figures match what is described in the text?
  • Are color scales defined with legends?
仔细阅读每个图表：
  • 图表是否准确呈现数据？
  • 是否标注了坐标轴、图例和单位？
  • 是否定义了统计标注（显著性星号、误差棒）？
  • 柱状图是否使用恰当？（样本量较小时，展示单个数据点。）
  • 时间/频率尺度是否适合所展示的数据？
  • 图表是否与文本描述一致？
  • 是否通过图例定义了颜色尺度？

7. Assess writing quality

7. 评估写作质量

  • Are technical terms defined before or immediately after first use?
  • Is terminology consistent throughout? (Do not introduce synonyms mid-paper.)
  • Is the writing concise? Flag unnecessary repetition.
  • Are abbreviations defined on first use and not redefined?
  • Does the abstract accurately reflect the paper's content and findings?
  • Is the methods section complete per the target journal's guidelines?
  • 技术术语是否在首次使用前或使用后立即定义？
  • 术语是否在全文中保持一致？（不得在文稿中途引入同义词。）
  • 写作是否简洁？标记不必要的重复内容。
  • 缩写是否在首次使用时定义，且未重复定义？
  • 摘要是否准确反映了论文的内容和发现？
  • 方法部分是否符合目标期刊的指南要求？

Review Output Format

评审输出格式

Structure the review according to the template in 
references/review-output-template.md
:
  1. Synopsis - 1 paragraph summarizing the paper's goal, methods, findings, strengths, and overall assessment
  2. Critical Issues - Numbered list of issues that would prevent publication (methodological flaws, invalid statistics, unsupported claims)
  3. Major Concerns - Numbered list of significant issues requiring revision (incomplete analysis, missing comparisons, overreached conclusions)
  4. Minor Concerns - Numbered list of issues that improve clarity and polish (writing, figures, references)
  5. Editor Note (optional) - Brief summary for the editor with recommendation
Every concern must:
  • Cite the specific location (page, line, figure, or section)
  • Explain what the problem is and why it matters
  • Provide a constructive suggestion or alternative approach
  • Cite supporting references when arguing a methodological point
按照
references/review-output-template.md
中的模板组织评审内容：
  1. 概要 - 1段文字总结论文的目标、方法、发现、优势和整体评估
  2. 关键问题 - 编号列表，列出会阻碍发表的问题（方法学缺陷、无效统计、无依据观点）
  3. 主要问题 - 编号列表，列出需要大幅修改的重要问题（分析不完整、缺失对比、结论过度延伸）
  4. 次要问题 - 编号列表，列出提升清晰度和打磨度的问题（写作、图表、参考文献）
  5. 编辑说明（可选） - 给编辑的简短总结及建议
每个问题必须：
  • 引用具体位置（页码、行号、图表或章节）
  • 说明问题是什么以及为什么重要
  • 提供建设性建议或替代方案
  • 论证方法学观点时引用支持性参考文献

Review Principles

评审原则

Consult 
references/review-principles.md
for the full rationale. Summary:
  1. Be direct but constructive - Every weakness must include a suggestion for improvement
  2. Be evidence-based - Cite literature when challenging methods or claims; do not rely on authority alone
  3. Be proportional - Calibrate severity to actual impact on the paper's validity
  4. Acknowledge strengths genuinely - Do not manufacture weaknesses; recognize good work
  5. Question logical consistency - If the intro says X, the methods must test X, and the discussion must conclude about X
  6. Demand statistical appropriateness - Wrong tests invalidate conclusions regardless of significance
  7. Insist on reproducibility - Papers that cannot be reproduced have limited scientific value
  8. Check the literature - Missing relevant work suggests incomplete understanding of the field
  9. Scrutinize figures - Figures are often where misleading presentations hide
  10. Hold claims to the data - The discussion must not exceed what the results demonstrate
查阅
references/review-principles.md
获取完整依据。摘要如下：
  1. 直接且具建设性 - 每个不足之处必须包含改进建议
  2. 基于证据 - 质疑方法或观点时引用文献；不得仅依赖权威
  3. 比例适当 - 根据问题对论文有效性的实际影响校准严重程度
  4. 真诚认可优势 - 不得刻意制造问题；认可优秀工作
  5. 质疑逻辑一致性 - 如果引言提出X，方法必须验证X，讨论必须围绕X得出结论
  6. 要求统计恰当性 - 错误的检验方法会使结论无效，无论显著性如何
  7. 坚持可重复性 - 无法复现的论文科学价值有限
  8. 检查文献 - 遗漏相关研究表明对领域理解不完整
  9. 仔细审查图表 - 误导性表述常隐藏在图表中
  10. 观点需符合数据 - 讨论内容不得超出结果所能证明的范围

Additional Resources

额外资源

Reference Files

参考文件

  • references/review-output-template.md
     - Complete review output format with examples
  • references/methodology-checklist.md
     - Detailed methodological assessment checklist
  • references/review-principles.md
     - Review philosophy and calibration guidance
  • references/statistical-review-guide.md
     - Common statistical issues and how to identify them
  • references/figure-review-guide.md
     - Figure quality assessment criteria
  • references/review-output-template.md
     - 完整的评审输出格式及示例
  • references/methodology-checklist.md
     - 详细的方法学评估清单
  • references/review-principles.md
     - 评审理念及校准指南
  • references/statistical-review-guide.md
     - 常见统计问题及识别方法
  • references/figure-review-guide.md
     - 图表质量评估标准