Back to Details

academic-writing-cs

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

Academic Writing for Computer Science

计算机科学学术写作

Overview

概述

This skill provides end-to-end support for writing high-quality computer science research papers. It focuses on constructing clear, compelling technical narratives while adhering to field-specific conventions.
Core Philosophy:
  • Academic papers are narrative arcs (Problem → Solution → Evidence → Implications), not template fill-ins
  • Clarity comes from structure: place familiar information first, new information last
  • Every design choice must be justified; every claim must be supported
Scope:
  • Conference papers (6-12 pages, competitive venues)
  • Journal articles (15-30 pages, comprehensive)
  • Thesis chapters (flexible length, deep coverage)
  • All CS subfields: AI/ML, Systems, Theory, HCI, Security, etc.
本技能为撰写高质量计算机科学科研论文提供端到端支持，专注于在遵循领域专属规范的同时，构建清晰、有说服力的技术叙事。
核心理念：
  • 学术论文是叙事弧（问题 → 解决方案 → 证据 → 启示），而非模板填空
  • 清晰度源于结构：把熟悉的信息放在前面，新信息放在后面
  • 每个设计选择都必须有依据，每个主张都必须有支撑
适用范围：
  • 会议论文（6-12页，高竞争力学术会场）
  • 期刊文章（15-30页，内容全面）
  • 学位论文章节（长度灵活，内容深度高）
  • 所有CS子领域：AI/ML、Systems、Theory、HCI、Security等

When to Use This Skill

何时使用本技能

Invoke this skill when:
  • Planning paper structure and narrative flow
  • Drafting any section (Abstract, Introduction, Methods, Results, Discussion, Conclusion)
  • Revising for clarity, coherence, or compliance with venue requirements
  • Reviewing sentence-level writing for clarity issues
  • Seeking CS-specific conventions (notation, figures, citations)
  • Checking completeness with section-by-section quality checklists
  • Responding to reviewer comments
当出现以下场景时可调用本技能：
  • 规划论文结构和叙事逻辑
  • 起草任何章节（摘要、引言、方法、结果、讨论、结论）
  • 修订内容以提升清晰度、连贯性，或符合投稿会场要求
  • 检查句子级写作的清晰度问题
  • 查找CS专属写作规范（符号标注、图表、引用）
  • 通过分章节质量检查表检查内容完整性
  • 回复审稿人意见

Workflow Decision Tree

工作流决策树

Stage 1: Planning and Structure

阶段1：规划与结构搭建

When starting a new paper or major revision:
  1. Define the Narrative Arc
    • What problem does this solve, and why does it matter? (1-2 sentences)
    • What is the single main contribution? (1 sentence)
    • What are the 3 key results that support the contribution?
    • What are the main limitations?
    Reference: 
    references/narrative_framework.md
    — Read the "Core Principle" and "Section-Level Narrative Structure" sections to understand how to structure the paper's story.
  2. Identify Target Venue and Constraints
    • Conference or journal?
    • Page limits, formatting requirements, anonymization rules?
    • Subfield conventions (ML vs. Systems vs. Theory)?
    Reference: 
    references/cs_conventions.md
    (Section 8: Venue-Specific Guidelines, Section 5: Subfield-Specific Conventions)
  3. Outline Section-by-Section
    • For each major section, define:
      • What is the purpose of this section?
      • What are the 2-3 key points to convey?
      • What figures/tables will support this?
    Tool: Use 
    assets/section_checklists.md
    (Quick Pre-Draft Planning Checklist) to ensure all key questions are answered before writing begins.
当开始撰写新论文或进行重大修订时：
  1. 定义叙事弧
    • 本研究解决了什么问题，为什么这个问题很重要？（1-2句话）
    • 研究唯一的核心贡献是什么？（1句话）
    • 支撑该贡献的3个关键结果是什么？
    • 研究的主要局限性是什么？
    参考资料： 
    references/narrative_framework.md
    — 阅读“核心原则”和“章节级叙事结构”部分，了解如何搭建论文的故事线。
  2. 确定目标投稿会场及约束
    • 投会议还是期刊？
    • 页数限制、格式要求、匿名化规则？
    • 子领域规范（ML/Systems/Theory各有差异）？
    参考资料： 
    references/cs_conventions.md
    （第8节：会场专属指南、第5节：子领域专属规范）
  3. 分章节搭建大纲
    • 为每个主要章节明确：
      • 本章节的目的是什么？
      • 要传达的2-3个核心要点是什么？
      • 哪些图表/表格可以支撑这些要点？
    工具： 使用
    assets/section_checklists.md
    （草稿前快速规划检查表），确保动笔前所有关键问题都已明确。

Stage 2: Drafting

阶段2：草稿撰写

For each section, follow this process:
撰写每个章节时，遵循以下流程：

Abstract

摘要

  1. Use the 4-sentence structure: Context → Gap → Contribution → Impact
  2. Check against 
    assets/section_checklists.md
    (Abstract Checklist)
  3. Ensure it's self-contained and within word limit (150-250 words)
Common mistakes:
  • Vague contribution: "We improve X" → Be specific: "We achieve 15% higher accuracy"
  • No concrete results: Always include numbers/metrics
  1. 使用4句话结构：背景 → 研究空白 → 贡献 → 影响
  2. 对照
    assets/section_checklists.md
    （摘要检查表）检查
  3. 确保摘要独立成义，且符合字数限制（150-250词）
常见错误：
  • 贡献表述模糊：“我们改进了X” → 要具体：“我们将准确率提升了15%”
  • 没有具体结果：始终要包含数值/指标

Introduction

引言

  1. Follow the funnel structure: Broad → Narrow → Specific
    • Para 1: Problem domain and importance
    • Para 2-3: Specific problem, motivation, why existing work falls short
    • Para 4: Gap statement ("However, existing approaches lack...")
    • Para 5: Contribution overview (what this paper provides)
    • Para 6: Results summary (2-3 concrete findings)
    • Para 7: Paper organization (optional)
  2. Key requirement: By the end of paragraph 4-5, the reader must clearly understand the contribution.
  3. Include at least one figure (architecture or key result) for ML/systems papers.
  4. Check against 
    assets/section_checklists.md
    (Introduction Checklist)
Reference: 
references/narrative_framework.md
(Introduction section) for detailed guidance and examples.
  1. 遵循漏斗结构：宽泛 → 收窄 → 具体
    • 第1段：问题领域及重要性
    • 第2-3段：具体问题、研究动机、现有工作的不足
    • 第4段：研究空白说明（“然而，现有方法缺少…”）
    • 第5段：贡献概述（本论文提出的内容）
    • 第6段：结果总结（2-3个具体结论）
    • 第7段：论文章节安排（可选）
  2. 核心要求： 读者在读完第4-5段时，必须清晰理解论文的贡献。
  3. ML/系统类论文至少要包含一张图（架构图或核心结果图）。
  4. 对照
    assets/section_checklists.md
    （引言检查表）检查。
参考资料： 
references/narrative_framework.md
（引言部分）获取详细指导和示例。

Related Work

相关工作

  1. Organize thematically (not chronologically): Group into 3-5 categories
  2. For each category:
    • Describe the general approach
    • Cite 3-5 representative works with 1-sentence descriptions
    • Point out limitations relevant to your contribution
  3. End with positioning paragraph: "In contrast to [X], our approach..."
    • Clearly articulate differences and advantages
  4. Check against 
    assets/section_checklists.md
    (Related Work Checklist)
Common mistakes:
  • Laundry list of citations without synthesis
  • Failing to position your work relative to prior work
  • Being dismissive (respect prior work while differentiating)
  1. 按主题组织（而非按时间排序）：分为3-5个类别
  2. 对每个类别：
    • 描述通用方法
    • 引用3-5个代表性工作并附上1句话说明
    • 指出与你的研究贡献相关的局限性
  3. 以定位段落收尾： “与[X]相比，我们的方法…”
    • 清晰阐述差异和优势
  4. 对照
    assets/section_checklists.md
    （相关工作检查表）检查。
常见错误：
  • 仅罗列引用文献没有整合分析
  • 没有将你的工作与前人工作做定位区分
  • 态度轻蔑（区分工作的同时要尊重前人研究）

Methodology

方法论

  1. Dual objectives:
    • Reproducibility: Enough detail for reimplementation
    • Intuition: Explain why the approach works
  2. Structure varies by paper type:
    • ML/AI papers: Problem Formulation → Overview + Figure → Detailed Design → Implementation → Complexity
    • Systems papers: Architecture Overview → Component Design → Key Mechanisms → Implementation
    • Theory papers: Formal Definitions → Main Results (theorems) → Proof Sketch
  3. Always include:
    • Clear notation (define all symbols on first use)
    • High-level overview before diving into details
    • Justification for design choices (or defer to Ablations)
  4. Check against 
    assets/section_checklists.md
    (Methodology Checklist)
Reference: 
references/narrative_framework.md
(Methodology section) and 
references/cs_conventions.md
(Section 1: Notation and Mathematical Writing)
  1. 双重目标：
    • 可复现性：提供足够细节支持他人复现实现
    • 可理解性：解释方法为什么有效
  2. 结构因论文类型不同有差异：
    • ML/AI论文：问题定义 → 总览+图示 → 详细设计 → 实现 → 复杂度分析
    • 系统类论文：架构总览 → 组件设计 → 核心机制 → 实现
    • 理论类论文：形式化定义 → 核心结果（定理） → 证明梗概
  3. 必须包含：
    • 清晰的符号标注（首次使用时定义所有符号）
    • 进入细节前先给出高层级总览
    • 设计选择的依据（或放在消融实验部分说明）
  4. 对照
    assets/section_checklists.md
    （方法论检查表）检查。
参考资料： 
references/narrative_framework.md
（方法论部分）和
references/cs_conventions.md
（第1节：符号标注与数学写作）

Experiments/Results

实验/结果

  1. Experimental Setup (subsection):
    • Datasets: Size, splits, preprocessing
    • Baselines: What you compare against (with citations)
    • Metrics: What you measure and why
    • Hardware/Software: Infrastructure and versions
    • Hyperparameters: How selected
  2. Main Results (subsection):
    • Table/figure showing primary comparison
    • Text: "Table 1 shows that our method outperforms..."
    • Highlight key findings with concrete numbers
    • Report statistical significance (confidence intervals, p-values, or std dev)
  3. Ablation Studies (subsection, critical):
    • Demonstrate necessity of each component
    • Table: effect of removing/modifying components
  4. Analysis (subsection):
    • Where does the method excel? Where does it fail?
    • Qualitative analysis, error analysis, failure cases
  5. Computational Cost (if relevant):
    • Training time, inference time, memory usage
    • Comparison with baselines
  6. Check against 
    assets/section_checklists.md
    (Experiments/Results Checklist)
Reference: 
references/narrative_framework.md
(Experiments/Results section)
  1. 实验设置（子章节）：
    • 数据集：规模、划分方式、预处理逻辑
    • 基线方法：对比的基准方法（附上引用）
    • 评估指标：测量的指标及选择理由
    • 硬件/软件：基础设施及版本
    • 超参数：选择方式
  2. 核心结果（子章节）：
    • 展示核心对比结果的表格/图
    • 正文描述：“表1显示我们的方法优于…”
    • 用具体数值突出核心发现
    • 报告统计显著性（置信区间、p值或标准差）
  3. 消融实验（子章节，非常重要）：
    • 证明每个组件的必要性
    • 表格展示移除/修改组件的效果影响
  4. 分析（子章节）：
    • 方法在什么场景下表现好？什么场景下表现差？
    • 定性分析、错误分析、失败案例
  5. 计算成本（如相关）：
    • 训练时长、推理时长、内存占用
    • 与基线方法的对比
  6. 对照
    assets/section_checklists.md
    （实验/结果检查表）检查。
参考资料： 
references/narrative_framework.md
（实验/结果部分）

Discussion

讨论

  1. Summarize findings (1 para): Restate key results
  2. Interpret results (1-2 paras): Why does the method work? What insights?
  3. Acknowledge limitations (0.5-1 para): Be honest about scope and failure cases
  4. Broader implications (0.5-1 para): Impact on the field, applications, future directions
  5. Check against 
    assets/section_checklists.md
    (Discussion Checklist)
Tone: Balanced—confident but not overselling. Limitations increase credibility.
  1. 总结发现（1段）：重述核心结果
  2. 解读结果（1-2段）：方法为什么有效？有哪些洞见？
  3. 承认局限性（0.5-1段）：坦诚说明适用范围和失败案例
  4. 更广泛的启示（0.5-1段）：对领域的影响、应用场景、未来方向
  5. 对照
    assets/section_checklists.md
    （讨论检查表）检查。
语气： 平衡——自信但不过度吹嘘。提及局限性会提升可信度。

Conclusion

结论

  1. Restate contribution (1 para): Recap problem, solution, key findings
  2. Broader impact (0.5 para): Significance and applications
  3. Future work (0.5 para): Open questions and extensions
    • Phrase as opportunities: "An interesting direction is..." (not "In future work, we will...")
  4. Check against 
    assets/section_checklists.md
    (Conclusion Checklist)
Do NOT: Introduce new ideas, copy-paste Abstract, or be vague.
  1. 重述贡献（1段）：概括问题、解决方案、核心发现
  2. 更广泛的影响（0.5段）：研究意义和应用场景
  3. 未来工作（0.5段）：待解决的问题和扩展方向
    • 以机会的方式表述：“一个有趣的方向是…”（不要用“未来我们将…”）
  4. 对照
    assets/section_checklists.md
    （结论检查表）检查。
禁止： 引入新想法、直接复制摘要、表述模糊。

Stage 3: Revision for Clarity

阶段3：清晰度修订

After drafting, apply sentence-level clarity principles:
草稿完成后，应用句子级清晰度原则：

The Three Golden Rules (Gopen & Swan)

三大黄金规则（Gopen & Swan）

  1. Old Before New: Start sentences with familiar information; end with new information
    • This creates coherent flow where each sentence builds on what came before
  2. Subject-Verb Proximity: Keep the verb close to the subject
    • Long gaps between subject and verb strain comprehension
  3. Stress Position Power: Place the most important information at sentence end
    • Readers remember and emphasize what comes at the end
Apply these rules systematically:
  • For each paragraph, check that sentences flow (old-to-new)
  • For each sentence, check that:
    • Topic position (start) contains familiar info
    • Stress position (end) contains important new info
    • Verb appears soon after subject
Reference: 
references/sentence_clarity.md
— Read this in full for detailed principles, examples, and common anti-patterns.
Practical Checklist:
  • Familiar information at sentence start (topic position)
  • Important new information at sentence end (stress position)
  • Verb close to subject
  • Active voice (unless passive is intentionally better)
  • Parallel structures for parallel ideas
Common anti-patterns to fix:
  • "Buried Verb" Syndrome: Converting verbs to nouns (nominalization)
    • ❌ "The comparison of the methods is shown..."
    • ✅ "Table 1 compares the methods..."
  • "Throat-Clearing": Weak starts like "It is important to note that..."
    • ❌ "It is important to note that our method improves accuracy."
    • ✅ "Our method improves accuracy."
  • "Dangling Emphasis": Ending sentences with weak elements
    • ❌ "This approach significantly improves performance, as shown in [23]."
    • ✅ "As shown in [23], this approach significantly improves performance."
  1. 旧信息在前，新信息在后：句子以熟悉的信息开头，新信息收尾
    • 这样可以形成连贯的逻辑流，每个句子都建立在前文内容的基础上
  2. 主语和动词靠近：让动词紧跟主语
    • 主语和动词之间间隔太长会增加理解负担
  3. 强调位置效应：把最重要的信息放在句子末尾
    • 读者会记住并重视句子末尾的内容
系统应用这些规则：
  • 对每个段落，检查句子是否符合旧到新的逻辑流
  • 对每个句子，检查：
    • 主题位置（开头）包含熟悉的信息
    • 强调位置（末尾）包含重要的新信息
    • 动词紧跟在主语之后
参考资料： 
references/sentence_clarity.md
— 完整阅读本文获取详细原则、示例和常见反模式。
实用检查表：
  • 句子开头（主题位置）是熟悉的信息
  • 重要的新信息放在句子末尾（强调位置）
  • 动词靠近主语
  • 使用主动语态（除非被动语态确实更合适）
  • 并列的观点使用平行结构
需要修正的常见反模式：
  • “动词掩埋”综合征：把动词转化为名词（名词化）
    • ❌ “方法的对比展示在…”
    • ✅ “表1对比了不同方法…”
  • “清嗓式开头”：比如“需要注意的是…”这类弱开头
    • ❌ “需要注意的是我们的方法提升了准确率。”
    • ✅ “我们的方法提升了准确率。”
  • “强调错位”：句子以次要元素收尾
    • ❌ “该方法显著提升了性能，如[23]所示。”
    • ✅ “如[23]所示，该方法显著提升了性能。”

Stage 4: Polishing and Compliance

阶段4：打磨与合规性检查

Language and Phrasing

语言与措辞

When writing or revising specific academic functions, consult 
references/phrasebank.md
:
  • Introducing work: Establishing territory, identifying gaps, stating contributions
  • Referring to sources: Integral vs. non-integral citations
  • Describing methods: Sequential actions, conditional logic, implementation details
  • Reporting results: Presenting findings, comparing baselines, interpreting
  • Discussing findings: Explaining success, acknowledging limitations, stating implications
  • Writing conclusions: Summarizing, broader impact, future work
General language functions:
  • Being cautious (hedging): "may", "appears to", "likely"
  • Being critical: Identifying weaknesses, questioning validity
  • Compare and contrast: Similarity, difference
  • Describing trends: Increasing, decreasing, stability
  • Explaining causality: Causes, effects, conditions
Usage: Adapt templates to your context; don't copy verbatim. Vary expressions to maintain natural flow.
撰写或修订特定学术表达时，可参考
references/phrasebank.md
：
  • 介绍研究工作：确立研究领域、识别研究空白、说明研究贡献
  • 引用来源：整合式vs非整合式引用
  • 描述方法：顺序动作、条件逻辑、实现细节
  • 报告结果：展示发现、对比基线、解读结果
  • 讨论发现：解释优势、承认局限性、说明启示
  • 撰写结论：总结内容、更广泛影响、未来工作
通用语言功能：
  • 谨慎表述（模糊限制）：“可能”、“似乎”、“大概率”
  • 批判性表述：识别不足、质疑有效性
  • 对比表述：相似性、差异性
  • 描述趋势：上升、下降、稳定
  • 解释因果关系：原因、效果、条件
使用方法： 结合你的上下文调整模板，不要逐字复制。变换表达方式保持自然流畅。

CS-Specific Conventions

CS专属规范

Ensure compliance with field norms:
  1. Notation:
    • Define all symbols on first use
    • Use consistent conventions (bold for vectors, italic for scalars, etc.)
    • Integrate equations into sentences with punctuation
  2. Figures and Tables:
    • Reference all figures/tables in text before they appear
    • Self-contained captions
    • High-resolution, readable fonts (≥8pt)
    • Colorblind-friendly palettes
  3. Citations:
    • Follow venue citation style (author-year or numbered)
    • Cite all prior work you build on or compare against
    • Accurate and complete bibliography
  4. Code and Reproducibility:
    • State code availability
    • Provide sufficient implementation details
    • Report hyperparameters, random seeds, number of runs
  5. Subfield-Specific Variations:
    • ML/AI: Emphasis on ablations, statistical significance, computational cost
    • Systems: Architecture diagrams, throughput/latency, scalability
    • Theory: Formal definitions, theorems, proofs, complexity bounds
    • HCI: User studies, qualitative feedback, interface screenshots
    • Security: Threat models, attack scenarios, defense mechanisms
Reference: 
references/cs_conventions.md
— Comprehensive guide covering notation, figures, citations, code, subfield norms, and venue requirements.
确保符合领域规范：
  1. 符号标注：
    • 首次使用时定义所有符号
    • 使用统一的规范（向量用粗体、标量用斜体等）
    • 把公式融入句子，添加合适的标点
  2. 图表：
    • 正文在图表出现前就引用所有图表
    • 图表标题可独立表意
    • 高分辨率、字体清晰（≥8pt）
    • 使用色盲友好的配色方案
  3. 引用：
    • 遵循投稿会场的引用格式（作者年份或编号式）
    • 引用所有你基于或对比的前人工作
    • 参考文献准确完整
  4. 代码与可复现性：
    • 说明代码是否公开
    • 提供足够的实现细节
    • 报告超参数、随机种子、运行次数
  5. 子领域专属差异：
    • ML/AI：重点关注消融实验、统计显著性、计算成本
    • 系统类：架构图、吞吐量/延迟、可扩展性
    • 理论类：形式化定义、定理、证明、复杂度边界
    • HCI：用户研究、定性反馈、界面截图
    • Security：威胁模型、攻击场景、防御机制
参考资料： 
references/cs_conventions.md
— 涵盖符号标注、图表、引用、代码、子领域规范、会场要求的综合指南。

Quality Assurance

质量保障

Before submission, use 
assets/section_checklists.md
:
  1. Section-by-Section Review:
    • Run through each section's checklist
    • Ensure all required elements are present
    • Check for common pitfalls
  2. Pre-Submission Checklist:
    • Content completeness (all sections, figures, citations)
    • Formatting (venue template, page limits, margins)
    • Anonymization (if double-blind)
    • Reproducibility (sufficient detail, code availability)
    • Final quality checks (spell-check, grammar, co-author review)
  3. Emergency Checklist (if deadline is imminent):
    • Prioritize: Abstract, Introduction contribution statement, Main results table, At least one ablation, Readable figures, Correct bibliography
提交前，使用
assets/section_checklists.md
：
  1. 分章节审核：
    • 对照每个章节的检查表检查
    • 确保所有必要元素都已包含
    • 检查常见陷阱
  2. 提交前检查表：
    • 内容完整性（所有章节、图表、引用）
    • 格式（会场模板、页数限制、页边距）
    • 匿名化（如果是双盲评审）
    • 可复现性（足够的细节、代码可用性）
    • 最终质量检查（拼写检查、语法、共同作者审核）
  3. 紧急检查表（如果临近截止日期）：
    • 优先处理：摘要、引言贡献说明、核心结果表、至少一个消融实验、清晰的图表、正确的参考文献

Stage 5: Responding to Reviews

阶段5：回复审稿意见

After receiving reviewer feedback:
  1. Analyze comments systematically:
    • Categorize: Major issues (experiments, clarity, claims) vs. Minor issues (typos, formatting)
    • Prioritize: Address major issues first
  2. Plan revisions:
    • List all changes to be made
    • If experiments are requested, plan them carefully
    • If clarifications are needed, identify which sections to revise
  3. Revise and respond:
    • Address every comment (in rebuttal or revision)
    • Use respectful, professional tone
    • Clearly mark changes (if required by venue)
  4. Check revised version:
    • Ensure all changes are integrated
    • Re-run relevant checklists from 
      assets/section_checklists.md
      (Revision Checklist)
    • Verify still within page limits
Reference: 
assets/section_checklists.md
(Revision Checklist)
收到审稿人反馈后：
  1. 系统分析意见：
    • 分类：重大问题（实验、清晰度、主张）vs 次要问题（拼写错误、格式）
    • 优先级：优先解决重大问题
  2. 规划修订：
    • 列出所有要做的修改
    • 如果需要补充实验，仔细规划
    • 如果需要澄清内容，确定要修订的章节
  3. 修订与回复：
    • 回应每一条意见（在反驳信或修订稿中）
    • 使用尊重、专业的语气
    • 明确标注修改内容（如果会场要求）
  4. 检查修订版本：
    • 确保所有修改都已整合
    • 对照
      assets/section_checklists.md
      （修订检查表）重新运行相关检查项
    • 确认仍符合页数限制
参考资料： 
assets/section_checklists.md
（修订检查表）

Key Resources Summary

核心资源汇总

Narrative and Structure

叙事与结构

  • references/narrative_framework.md
    : Core paper structure (Abstract, Introduction, Related Work, Methods, Results, Discussion, Conclusion). Use for understanding the narrative arc and section-specific guidance.
  • references/narrative_framework.md
    ：核心论文结构（摘要、引言、相关工作、方法、结果、讨论、结论）。用于理解叙事弧和章节专属指导。

Sentence-Level Clarity

句子级清晰度

  • references/sentence_clarity.md
    : Gopen & Swan principles (topic position, stress position, old-to-new flow). Use for revising individual sentences and paragraphs for maximum clarity.
  • references/sentence_clarity.md
    ：Gopen & Swan原则（主题位置、强调位置、旧到新逻辑流）。用于修订单个句子和段落，最大化清晰度。

Academic Phrases

学术短语

  • references/phrasebank.md
    : Templates for common academic writing functions (introducing work, citing sources, reporting results, discussing findings). Use when drafting or seeking variation in phrasing.
  • references/phrasebank.md
    ：常见学术写作场景的模板（介绍研究、引用来源、报告结果、讨论发现）。用于起草或变换表达方式。

CS Conventions

CS规范

  • references/cs_conventions.md
    : Field-specific norms (notation, figures, citations, code, subfield variations, venue requirements). Use for ensuring compliance with CS writing standards.
  • references/cs_conventions.md
    ：领域专属规范（符号标注、图表、引用、代码、子领域差异、会场要求）。用于确保符合CS写作标准。

Quality Checklists

质量检查表

  • assets/section_checklists.md
    : Comprehensive checklists for every section, plus pre-submission, revision, and emergency checklists. Use for planning, reviewing, and final quality assurance.
  • assets/section_checklists.md
    ：覆盖所有章节的综合检查表，以及提交前、修订、紧急检查表。用于规划、审核和最终质量保障。

Example Workflows

示例工作流

Workflow 1: Starting from Scratch

工作流1：从零开始写作

User: "I need to write a conference paper on my new semi-supervised learning method."
Process:
  1. Planning (Stage 1):
    • Define narrative arc: Problem (labeled data is expensive) → Solution (our semi-supervised method) → Evidence (experiments on 3 datasets) → Implications (reduces labeling cost)
    • Read 
      references/narrative_framework.md
      (Core Principle)
    • Use 
      assets/section_checklists.md
      (Quick Pre-Draft Planning Checklist)
  2. Drafting (Stage 2):
    • Abstract: 4-sentence structure (Context: deep learning needs data; Gap: labeling is expensive; Contribution: our method STCR; Impact: 82% accuracy with 10% labels)
    • Introduction: Funnel (broad: DL success → narrow: labeling cost → gap: existing semi-supervised methods lack X → contribution: STCR leverages consistency → results: 7% improvement)
    • Check each section against 
      assets/section_checklists.md
  3. Revision (Stage 3):
    • Apply 
      references/sentence_clarity.md
      principles to every paragraph
    • Ensure old-to-new flow, stress position usage
  4. Polishing (Stage 4):
    • Use 
      references/phrasebank.md
      for varied phrasing
    • Ensure compliance with 
      references/cs_conventions.md
      (ML/AI conventions)
    • Run Pre-Submission Checklist from 
      assets/section_checklists.md
用户： “我需要写一篇关于我的新半监督学习方法的会议论文。”
流程：
  1. 规划（阶段1）：
    • 定义叙事弧：问题（标注数据成本高）→ 解决方案（我们的半监督方法）→ 证据（3个数据集上的实验）→ 启示（降低标注成本）
    • 阅读
      references/narrative_framework.md
      （核心原则）
    • 使用
      assets/section_checklists.md
      （草稿前快速规划检查表）
  2. 草稿撰写（阶段2）：
    • 摘要：4句话结构（背景：深度学习需要大量数据；空白：标注成本高；贡献：我们的STCR方法；影响：仅用10%标注数据达到82%准确率）
    • 引言：漏斗结构（宽泛：DL的成功 → 收窄：标注成本高 → 空白：现有半监督方法缺少X → 贡献：STCR利用一致性约束 → 结果：提升7%）
    • 对照
      assets/section_checklists.md
      检查每个章节
  3. 修订（阶段3）：
    • 对每个段落应用
      references/sentence_clarity.md
      的原则
    • 确保旧到新的逻辑流、合理使用强调位置
  4. 打磨（阶段4）：
    • 使用
      references/phrasebank.md
      变换表达方式
    • 确保符合
      references/cs_conventions.md
      （ML/AI规范）
    • 运行
      assets/section_checklists.md
      中的提交前检查表

Workflow 2: Revising for Clarity

工作流2：清晰度修订

User: "My introduction is confusing. Reviewers said they couldn't understand the contribution."
Process:
  1. Diagnose issue:
    • Check against 
      assets/section_checklists.md
      (Introduction Checklist)
    • Is the contribution stated clearly by paragraph 4-5?
    • Is the funnel structure followed (broad → narrow)?
  2. Restructure if needed:
    • Read 
      references/narrative_framework.md
      (Introduction section)
    • Ensure: Opening → Background → Gap → Contribution → Results → Organization
    • Explicitly state: "In this paper, we present [X], which addresses [Y] by [Z]."
  3. Revise at sentence level:
    • Apply 
      references/sentence_clarity.md
      principles
    • Check that each sentence flows from the previous one (old-to-new)
    • End key sentences with the important information (stress position)
用户： “我的引言很混乱，审稿人说他们看不懂贡献。”
流程：
  1. 诊断问题：
    • 对照
      assets/section_checklists.md
      （引言检查表）检查
    • 第4-5段是否清晰说明了贡献？
    • 是否遵循了漏斗结构（宽泛 → 收窄）？
  2. 必要时重构结构：
    • 阅读
      references/narrative_framework.md
      （引言部分）
    • 确保结构：开篇 → 背景 → 研究空白 → 贡献 → 结果 → 章节安排
    • 明确表述：“在本论文中，我们提出了[X]，通过[Z]解决了[Y]问题。”
  3. 句子级修订：
    • 应用
      references/sentence_clarity.md
      的原则
    • 检查每个句子是否承接前文（旧到新逻辑）
    • 核心句子把重要信息放在末尾（强调位置）

Workflow 3: Drafting the Results Section

工作流3：撰写结果章节

User: "How should I present my experimental results?"
Process:
  1. Structure:
    • Read 
      references/narrative_framework.md
      (Experiments/Results section)
    • Follow: Setup → Main Results → Ablations → Analysis → Cost
  2. Create tables/figures:
    • Main results table: Methods (rows) vs. Metrics (columns)
    • Bold best results; include standard deviations
    • Check 
      references/cs_conventions.md
      (Figures and Tables section)
  3. Write accompanying text:
    • "Table 1 shows that our method achieves X, outperforming the strongest baseline by Y%."
    • Use 
      references/phrasebank.md
      (Section 4: Reporting Results) for phrasing
  4. Quality check:
    • Run through 
      assets/section_checklists.md
      (Experiments/Results Checklist)
    • Ensure: Statistical significance, Ablations present, Analysis included
用户： “我应该怎么展示我的实验结果？”
流程：
  1. 结构：
    • 阅读
      references/narrative_framework.md
      （实验/结果部分）
    • 遵循结构：设置 → 核心结果 → 消融实验 → 分析 → 成本
  2. 创建表格/图：
    • 核心结果表：方法（行）vs 指标（列）
    • 最优结果加粗；包含标准差
    • 对照
      references/cs_conventions.md
      （图表部分）检查
  3. 撰写配套正文：
    • “表1显示我们的方法达到了X，比最优基线高Y%。”
    • 使用
      references/phrasebank.md
      （第4节：报告结果）参考措辞
  4. 质量检查：
    • 对照
      assets/section_checklists.md
      （实验/结果检查表）检查
    • 确保：统计显著性、包含消融实验、包含分析

Workflow 4: Ensuring CS Compliance

工作流4：确保CS规范合规

User: "Is my notation and citation style correct for ICML?"
Process:
  1. Check venue requirements:
    • Read 
      references/cs_conventions.md
      (Section 8: Venue-Specific Guidelines)
    • ICML uses numbered citations [1], double-blind review, LaTeX template
  2. Notation:
    • Read 
      references/cs_conventions.md
      (Section 1: Notation and Mathematical Writing)
    • Ensure: Vectors are bold, scalars are italic, all symbols defined
  3. Citations:
    • Read 
      references/cs_conventions.md
      (Section 3: Citations and References)
    • Use numbered format: "Method X [1] achieves..."
    • Anonymize self-citations for double-blind
  4. Final check:
    • assets/section_checklists.md
      (Pre-Submission Checklist → Compliance section)
用户： “我的符号标注和引用格式符合ICML要求吗？”
流程：
  1. 检查会场要求：
    • 阅读
      references/cs_conventions.md
      （第8节：会场专属指南）
    • ICML使用编号式引用[1]、双盲评审、LaTeX模板
  2. 符号标注：
    • 阅读
      references/cs_conventions.md
      （第1节：符号标注与数学写作）
    • 确保：向量用粗体、标量用斜体、所有符号都已定义
  3. 引用：
    • 阅读
      references/cs_conventions.md
      （第3节：引用与参考文献）
    • 使用编号格式：“方法X[1]达到了…”
    • 双盲评审时匿名化自引
  4. 最终检查：
    • 对照
      assets/section_checklists.md
      （提交前检查表 → 合规部分）检查

Common Pitfalls and How to Avoid Them

常见陷阱及规避方法

Pitfall 1: Vague Contributions

陷阱1：贡献表述模糊

Problem: "We improve performance on X." Solution: Be specific. "We achieve 15% higher accuracy than the strongest baseline on ImageNet."
问题： “我们提升了X上的性能。” 解决方案： 表述具体。“在ImageNet数据集上，我们的准确率比最优基线高15%。”

Pitfall 2: Missing Ablations

陷阱2：缺少消融实验

Problem: Claiming design choices are important without evidence. Solution: Include ablation studies. Remove each component and measure the performance drop.
问题： 声称设计选择很重要但没有证据支撑。 解决方案： 包含消融实验。移除每个组件并测量性能下降幅度。

Pitfall 3: Poor Information Flow

陷阱3：信息流混乱

Problem: Sentences feel disjointed; readers get lost. Solution: Apply old-to-new flow. Each sentence should start with information from the previous sentence. Reference: 
references/sentence_clarity.md
问题： 句子之间脱节，读者无法理解逻辑。 解决方案： 应用旧到新的逻辑流。每个句子的开头都要承接上一句的信息。 参考资料： 
references/sentence_clarity.md

Pitfall 4: Weak Stress Position

陷阱4：强调位置使用不当

Problem: Sentences end with citations or minor details. Example: ❌ "This approach significantly improves performance, as shown in [23]." Solution: ✅ "As shown in [23], this approach significantly improves performance."
问题： 句子以引用或次要细节收尾。 示例： ❌ “该方法显著提升了性能，如[23]所示。” 解决方案： ✅ “如[23]所示，该方法显著提升了性能。”

Pitfall 5: Ignoring Limitations

陷阱5：忽略局限性

Problem: Overselling without acknowledging scope or failure cases. Solution: Dedicate a paragraph in Discussion to honest limitations. This increases credibility.
问题： 过度吹嘘，不承认适用范围或失败案例。 解决方案： 在讨论部分专门用一段坦诚说明局限性。这会提升可信度。

Pitfall 6: Inconsistent Notation

陷阱6：符号标注不一致

Problem: Using 
x
for input in one section, 
X
in another. Solution: Define all notation upfront. Create a notation table (appendix) if needed. Reference: 
references/cs_conventions.md
(Section 1)
问题： 一个章节里用
x
表示输入，另一个章节用
X
表示输入。 解决方案： 提前定义所有符号。如果需要可以在附录放符号对照表。 参考资料： 
references/cs_conventions.md
（第1节）

Tips for Efficient Writing

高效写作技巧

  1. Draft quickly, revise thoroughly:
    • Don't aim for perfection in the first draft
    • Get ideas down, then refine structure and clarity
  2. Write sections out of order:
    • Start with Methods and Results (most concrete)
    • Then Introduction and Related Work
    • Finally Abstract and Conclusion
  3. Use figures early:
    • Create key figures (architecture, main results) before writing
    • Figures clarify your thinking and guide the narrative
  4. Get feedback early:
    • Share drafts with co-authors and colleagues
    • Mock reviews identify issues before submission
  5. Iterate on structure:
    • If a section feels wrong, revisit the narrative arc
    • Ensure every section advances Problem → Solution → Evidence → Implications
  6. Use the checklists proactively:
    • Before drafting a section, read the checklist to know what to include
    • After drafting, use the checklist to verify completeness
  1. 快速写草稿， thorough修订：
    • 第一版草稿不要追求完美
    • 先把想法写下来，再优化结构和清晰度
  2. 可以不按顺序写章节：
    • 从方法和结果开始写（最具体的部分）
    • 然后写引言和相关工作
    • 最后写摘要和结论
  3. 尽早确定图表：
    • 动笔前先做核心图表（架构图、核心结果图）
    • 图表能理清你的思路，引导叙事逻辑
  4. 尽早获取反馈：
    • 把草稿分享给共同作者和同事
    • 模拟评审能在提交前发现问题
  5. 迭代优化结构：
    • 如果某个章节感觉不对，重新梳理叙事弧
    • 确保每个章节都推进“问题 → 解决方案 → 证据 → 启示”的逻辑
  6. 主动使用检查表：
    • 起草某个章节前，先读对应的检查表，明确要包含的内容
    • 起草完成后，用检查表验证内容完整性

Advanced: Handling Special Cases

进阶：特殊场景处理

Writing for Top-Tier Venues

顶会顶刊写作

  • Higher bar for novelty and rigor: Ensure the contribution is significant, not incremental
  • Strong baselines: Compare against state-of-the-art, not just simple methods
  • Comprehensive evaluation: Multiple datasets, extensive ablations, sensitivity analyses
  • Polished presentation: High-quality figures, clear writing, consistent notation
  • 创新性和严谨性要求更高：确保贡献是显著的，而非增量改进
  • 强基线：和SOTA方法对比，而非仅和简单方法对比
  • 全面的评估：多个数据集、大量消融实验、敏感性分析
  • ** polished的呈现**：高质量图表、清晰的写作、统一的符号标注

Writing Rebuttals

反驳信写作

  • Address all concerns: Even if you disagree, engage respectfully
  • Provide evidence: If reviewers doubt a claim, provide additional results or citations
  • Be concise: Rebuttals have strict length limits; prioritize major issues
  • Highlight changes: "We added an experiment (Table 3) showing..."
  • 回应所有疑问：即使你不同意，也要礼貌沟通
  • 提供证据：如果审稿人质疑某个主张，提供额外的结果或引用
  • 简洁：反驳信有严格的字数限制；优先处理重大问题
  • 突出修改内容：“我们新增了实验（表3）证明…”

Writing Thesis Chapters

学位论文章节写作

  • More comprehensive: Deeper background, extended related work, lessons learned
  • Narrative continuity: Ensure chapters connect (e.g., Chapter 3 builds on Chapter 2)
  • Broader scope: Can include negative results and explorations that didn't pan out
  • Use 
    assets/section_checklists.md
    (Long-Form Paper Checklist)
  • 更全面：更深入的背景、扩展的相关工作、经验教训
  • 叙事连续性：确保章节之间有关联（比如第3章基于第2章的内容）
  • 更宽泛的范围：可以包含负面结果和没有成功的探索
  • 使用
    assets/section_checklists.md
    （长文论文检查表）

Summary: The Golden Workflow

总结：黄金工作流

  1. Plan the narrative: Problem → Solution → Evidence → Implications
  2. Draft section-by-section: Use structure guidelines from 
    references/narrative_framework.md
  3. Revise for clarity: Apply principles from 
    references/sentence_clarity.md
  4. Polish and comply: Use 
    references/phrasebank.md
    and 
    references/cs_conventions.md
  5. Quality check: Run through 
    assets/section_checklists.md
Remember:
  • Papers are stories, not templates
  • Clarity comes from structure (old-to-new, topic/stress positions)
  • Every claim needs evidence; every design choice needs justification
  • Honest limitations increase credibility
When in doubt, ask:
  • "Does this advance the narrative arc?"
  • "Can a reader reproduce this?"
  • "Is this claim supported?"
  • "Is this the simplest, clearest way to express this?"
  1. 规划叙事： 问题 → 解决方案 → 证据 → 启示
  2. 分章节起草： 使用
    references/narrative_framework.md
    的结构指南
  3. 修订提升清晰度： 应用
    references/sentence_clarity.md
    的原则
  4. 打磨与合规： 使用
    references/phrasebank.md
    和
    references/cs_conventions.md
  5. 质量检查： 对照
    assets/section_checklists.md
    检查
记住：
  • 论文是故事，不是模板
  • 清晰度源于结构（旧到新逻辑、主题/强调位置）
  • 每个主张都需要证据；每个设计选择都需要依据
  • 坦诚的局限性会提升可信度
有疑问时，自问：
  • “这部分内容是否推进了叙事弧？”
  • “读者能复现这个结果吗？”
  • “这个主张有支撑吗？”
  • “这是表达这个内容最简单、最清晰的方式吗？”

Getting Started

开始使用

For a new paper:
  1. Read 
    references/narrative_framework.md
    (Core Principle)
  2. Use 
    assets/section_checklists.md
    (Quick Pre-Draft Planning Checklist)
  3. Outline your paper's narrative arc in 4 sentences (Problem, Solution, Evidence, Implications)
  4. Draft section-by-section, checking checklists as you go
For revising an existing draft:
  1. Identify the issue (structure, clarity, compliance)
  2. Consult the relevant reference file
  3. Apply fixes systematically
  4. Re-check with the appropriate checklist
For sentence-level issues:
  1. Read 
    references/sentence_clarity.md
    (Three Golden Rules)
  2. Apply to each problematic paragraph
  3. Check: Old-to-new flow, stress position usage, subject-verb proximity
Ready to write? Let's build a clear, compelling paper together.
新论文写作：
  1. 阅读
    references/narrative_framework.md
    （核心原则）
  2. 使用
    assets/section_checklists.md
    （草稿前快速规划检查表）
  3. 用4句话概括你论文的叙事弧（问题、解决方案、证据、启示）
  4. 分章节起草，边写边对照检查表检查
修订现有草稿：
  1. 明确问题（结构、清晰度、合规性）
  2. 参考对应的参考文件
  3. 系统地修改
  4. 用对应的检查表重新检查
句子级问题：
  1. 阅读
    references/sentence_clarity.md
    （三大黄金规则）
  2. 应用到每个有问题的段落
  3. 检查：旧到新逻辑流、强调位置使用、主语动词靠近
准备好写作了吗？我们一起打造一篇清晰、有说服力的论文吧。