📚 学术研究 AI Prompt 完整合集 v2.1 | Academic Research AI Prompt Library

简体中文 | English

语言说明 / Language scope: 本 README 提供中英文版本；子目录中的提示词、模板和流程图目前主要为中文。This README is available in Chinese and English; the prompt files, templates, and flowcharts are currently primarily in Chinese.

说明 / Note: 以下规模、耗时和效果数字沿用原项目说明，不代表经过验证的评测结果，请以当前仓库内容和实际使用情况为准。The collection sizes, timelines, and effectiveness figures below are retained from the original project overview, not validated benchmark results. Refer to the current repository contents and your own use case.

---

简体中文

项目文件 | 使用示例 | 快速开始 | English

一套为研究生和学术研究者设计的完整AI Prompt库

89+个精心设计的Prompt，覆盖从论文选题到文献综述的全流程

节省50-70%的研究时间，让你有更多精力做真正重要的工作

---

🎯 这个项目解决什么问题？

如果你是研究生，可能遇到过这些困境：

• 选题困难：花了2-3周还在纠结题目，不知道最后会不会被打回重来
• 找论文费时：手动搜索相关论文要花10-20小时，还是找不全
• 没有思路：不知道怎样系统地组织和分析这些论文
• 文本问题：AI生成的初稿AI率太高、文风不一致、需要优化
• 效率低下：重复做一些低效的工作，时间总是不够用
• 压力大：截止日期逼近，心理压力很大

**这个项目就是为了解决这些问题。**我们提供了一套系统的方法和现成的Prompt，帮你大幅加快科研进度。

---

✨ 能帮你做什么？

1️⃣ 论文选题 - 3-5天而不是2-3周

用系统的方法快速生成和评估论文选题：

• 用AI生成100个选题候选
• 用5维度评估表科学地评分
• 深度分析Top 3选题
• 准备与导师讨论的方案

效果：不再纠结，信心满满地敲定选题，还能节省后期2-3个月的修改时间

2️⃣ 论文查找 - 2小时而不是20小时

快速找到相关论文：

• 8个不同的快速查找方案（从3分钟到15分钟）
• AI直接列举相关论文，省去手动搜索的苦恼
• 在Google Scholar中一键验证

效果：节省80%的论文查找时间

3️⃣ 文献综述 - 清晰框架，效率提升50%

系统地整理和分析论文：

• 用Prompt帮你分类论文
• 生成清晰的综述框架
• 更高效地撰写文献综述

效果：有了好框架，写得快得多

4️⃣ 论文写作 - 2-3周高效完成 ⭐ 新增v2.1

撰写顶刊级别的论文：

• 从创新性评估到结构设计
• 从标题摘要到方法结果
• 全新的9个文本优化Prompt（降低AI率、润色、扩写、缩写、提升表达）
• 综合优化确保高质量成稿

效果：专业的顶刊级论文，综合优化效果显著

---

📁 项目包含什么？

academic-ai-prompt/
├─ README.md                              (中英双语项目说明)
├─ 索引                                   (总索引)
├─ 论文选题系列/
│  ├─ 论文选题AI提示词库.md               (20+个选题Prompt)
│  ├─ 论文选题论证方案模板.md             (3个论证模板)
│  └─ 论文选题5维度评估表.xlsx            (Excel评估工具)
├─ 论文查找系列/
│  ├─ 论文查找和文献综述AI提示词库.md     (30+个查找Prompt)
│  └─ AI直接找论文的Prompt.md             (8个快速方案)
├─ 文献综述系列/
│  └─ 文献综述AI提示词库.md               (20+个综述Prompt)
├─ 论文撰写系列/
│  └─ 顶刊论文写作AI提示词库-2.md         (34+个写作Prompt，新增9个文本优化)
├─ flowchart_topic_selection.png          (论文选题流程图)
├─ flowchart_paper_finding.png             (论文查找流程图)
├─ flowchart_literature_review.png         (文献综述流程图)
└─ LICENSE                                (MIT许可证)

总计：89+个Prompt | 165,000+字 | 3个流程图 | 1个Excel工具 | 3个完整模板

文件直达：选题提示词 | 论证模板 | Excel评估表 | 快速找论文 | 查找与综述 | 文献综述 | 论文写作

---

💡 使用示例：两个详细的使用场景

以下示例中的耗时和日期沿用原项目说明，仅供参考。使用时请更新研究时间范围，并核验所有文献信息。

✨ 案例1：快速生成100个论文选题（30分钟完成！）

场景：计算机系硕士生，机器学习+推荐系统方向，想创新选题

传统做法的痛点：

• ❌ 自己想选题？2-3周还是纠结，经常被打回
• ❌ 咨询导师？导师很忙，可能只给你3-5个方向
• ❌ 查文献？看了50篇论文还是不知道选什么
• ❌ 最后？匆匆选一个，后期才发现选错了...

用我们的Prompt的做法：

步骤1：复制下面的Prompt，填入你的信息

我是一名硕士研究生，计划选择关于机器学习和推荐系统的论文选题。

我的专业背景和兴趣：
- 本专业：计算机科学与技术（机器学习方向）
- 研究兴趣：深度学习、推荐系统、自然语言处理
- 已有知识：深度学习基础、Python编程、论文阅读能力
- 特别关注的问题：如何让推荐系统更准确、更快、更公平

目前学界的热点和方向：
深度学习在推荐系统中的应用、图神经网络在推荐中的应用、
冷启动问题、跨域推荐、个性化推荐的可解释性

我的导师的研究方向：
推荐系统、信息检索、用户建模

请帮我生成100个可能的论文选题，要求：
1. 选题应该在机器学习和推荐系统范围内
2. 选题应该既有创新性（不是简单应用现有方法），又具有可行性（两年内能完成）
3. 选题应该包括不同的研究方向：理论突破、方法改进、应用创新、工程优化等
4. 选题应该考虑当前学界的热点（2023-2024年的研究方向）
5. 选题应该与推荐系统和深度学习有关联

请以列表形式给出这100个选题，每个选题包括：
- 选题名称（简洁、有学术性）
- 简短的描述（一句话，说明这个选题研究什么）
- 为什么这个选题有价值（创新点是什么？实际应用价值是什么？）

步骤2：粘贴到ChatGPT或Claude，点击发送

在5-30分钟内，你会获得一个完整的选题列表，包括：

• ✅ 100个不同角度的选题
• ✅ 每个选题的详细说明
• ✅ 为什么这个选题有价值
• ✅ 从基础理论到前沿应用的全覆盖

步骤3：快速筛选和评估

1. 用「论文选题5维度评估表.xlsx」给这100个选题评分
   • 只需5-10分钟，Excel自动计算加权评分
2. 排出Top 10最好的选题
3. 再用Prompt 3.1深度分析Top 3

步骤4：确定最终选题

• 几天内就能确定最终选题，信心满满
• 导师问你「为什么选这个题目」，你能详细论证
• 而不是「我也不知道」或「因为简单」

时间对比：

• 🐢 传统做法：2-3周纠结，选题质量50%
• 🚀 我们的方法：3-5天完成，选题质量90%+

---

✨ 案例2：5分钟找到30-50篇相关论文（vs 10-20小时！）

场景：已选题「基于深度学习的医学图像分割」，需要快速查找相关论文

传统做法的痛点：

• ❌ 打开Google Scholar搜索「medical image segmentation」
• ❌ 出现2000+论文，怎么选？
• ❌ 一篇一篇看摘要，10小时过去了...
• ❌ 还没看完一半，脑子已经糊涂了
• ❌ 最后抓了30篇论文，质量参差不齐

用我们的Prompt的做法：

复制这个Prompt（来自「AI直接找论文的Prompt.md」）：

请为我推荐30-50篇关于"基于深度学习的医学图像分割"的高质量相关论文。

我的具体需求：
- 研究方向：医学图像处理、计算机视觉
- 核心技术：深度学习、CNN、图像分割（U-Net、FCN等）
- 应用领域：肺部病变分割、肿瘤分割、器官分割、血管分割等
- 我希望包含的类型：
  * 经典基础论文（奠基性工作）
  * 前沿方法论文（最新的CNN/Transformer架构）
  * 应用案例（真实医学图像分割的成功案例）
  * 挑战和改进（现有方法的局限和改进方向）
- 论文时间范围：最近8年（2016-2024），重点关注2022-2024年的论文
- 期刊/会议偏好：IEEE、Nature、MICCAI、CVPR等顶级期刊/会议优先

请给出论文推荐清单，每篇论文包括：
- 论文标题（英文）
- 作者和发表年份
- 发表期刊/会议
- 一句话核心贡献
- 为什么推荐这篇论文
- 推荐指数（1-5星）

请按照推荐指数从高到低排序。

另外，请在最后添加：
- 这个领域的研究脉络总结（过去10年如何发展的）
- 当前还存在的主要问题和机遇
- 建议我重点关注哪几篇论文作为切入点

示例输出（文献信息需核验）：

⭐⭐⭐⭐⭐ 必读论文
1. U-Net: Convolutional Networks for Biomedical Image Segmentation
   作者：Ronneberger et al., 2015
   期刊：MICCAI 2015
   核心贡献：提出U-Net架构，成为医学图像分割的标准方法
   推荐理由：这是医学图像分割的奠基性工作，几乎所有现代方法都基于或参考U-Net
   必读指数：5/5

2. Attention U-Net: Learning Where to Look for the Pancreas
   作者：Ozan Oktay et al., 2018
   期刊：MIDL 2018
   核心贡献：在U-Net基础上加入注意力机制，提升分割准确率
   推荐理由：展示了如何改进U-Net，对你的创新很有参考价值
   必读指数：5/5

... (还有26-47篇)

领域研究脉络：
2010-2015: 传统机器学习方法 → FCN出现 → CNN成为主流
2015-2018: U-Net及其变体统治医学图像分割
2018-2021: 注意力机制、多尺度方法、多任务学习出现
2021-2024: Vision Transformer、自监督学习、少样本学习成为新热点

然后：在Google Scholar验证，下载论文，3分钟内拥有完整的论文库

时间对比：

• 🐢 传统搜索：10-20小时，质量随机
• 🚀 我们的方法：5分钟得到推荐，30-50篇精选论文

---

🎯 为什么这套库这么强大？

不仅仅是Prompt集合，更是思维模式的升级：

✨ 传统方法：「我要选个题目」→ 瞎想 → 错选 ✨ 我们的方法：「系统地生成候选」→ 科学评估 → 最优选择

✨ 传统方法：「找论文太费时」→ 几周后放弃 ✨ 我们的方法：「用AI辅助」→ 5分钟有清单 → 高效阅读

✨ 传统方法：「AI初稿」→ 质量差、AI率高 ✨ 我们的方法：「AI初稿」→ 用9个文本优化Prompt → 高质量成稿

这就是为什么能节省50-70%的时间！

---

📊 使用效果对比

任务	传统方式	使用本项目	节省
论文选题	2-3周纠结	3-5天确定	2-3个月后期修改
论文查找	10-20小时	5分钟	80%时间
文献综述	无框架，低效	有框架，高效	50%写作时间
论文写作	4-6周	2-3周	40-50%时间
文本优化	手工反复修改	用9个Prompt优化	显著质量提升
整体科研	40-60小时	16小时	50-70%

---

🆕 v2.1新增：文本润色与表达专题

9个全新Prompt完全解决论文文本问题：

1. Prompt 7.1 - 论文润色（提升学术质量）
2. Prompt 7.2 - 降低AI率⭐（可降低30-50%）
3. Prompt 7.3 - 扩写段落（丰富内容）
4. Prompt 7.4 - 缩写段落（精简冗余）
5. Prompt 7.5 - 提升学术表达（权威性增强）
6. Prompt 7.6 - AI改进指南（改进策略）
7. Prompt 7.7 - 统一文风（保持一致）
8. Prompt 7.8 - 准确性检查（学术审视）
9. Prompt 7.9 - 综合优化✨（一站式解决）

详见：论文写作库中的「文本润色与表达」。

---

🚀 立即开始（3种场景）

场景1：我需要选论文题目 (3-5天)

1. 打开 论文选题AI提示词库.md
2. 复制 Prompt 1.1（生成100个选题）
3. 粘贴到ChatGPT/Claude，填入信息
4. 用 论文选题5维度评估表.xlsx 评分
5. 用 Prompt 3.1 深度分析Top 3

场景2：选题已定，我要找论文 (5分钟)

1. 打开 AI直接找论文的Prompt.md
2. 复制方案8（3分钟极速版）或方案1（5分钟标准版）
3. 填入你的选题信息
4. 发给ChatGPT/Claude，获得30-50篇论文清单

场景3：我要写高质量论文 (1-2周)

1. 用场景2找到50-100篇论文
2. 打开 文献综述AI提示词库.md（框架→内容→写作）
3. 打开 顶刊论文写作AI提示词库-2.md（各部分写作）
4. 使用新增的9个文本优化Prompt优化论文
   • Prompt 7.2：降低AI率
   • Prompt 7.1：润色表达
   • Prompt 7.3/7.4：字数调整
   • Prompt 7.9：综合优化

---

💡 使用建议

✅ 推荐做法

1. 按顺序来：选题 → 查找 → 综述 → 写作
2. AI是辅助，不是替代：结合你的判断效果最好
3. 多尝试不同Prompt：同一问题的多个版本
4. 根据需要调整：根据反馈调整参数逐步改进
5. 进行人工检查：确保准确性和学术严谨性

⚠️ 注意事项

• AI列举的论文需要在Google Scholar验证
• 某些论文可能已过时，确认时间和引用量
• 不同学科差异较大，可能需要根据学科特点调整
• AI生成的文本需要人工改进和最后检查
• AI检测分数不等同于原创性或学术质量；请遵守所在机构和目标期刊的AI使用与披露要求

---

🎓 适用人群

✅ 硕士研究生、博士研究生、本科毕业生 ✅ 学术工作者、学术内容创作者 ✅ 所有学科（工程、理学、社科、人文、医学等）

---

📚 项目规模

指标	数值
Prompt总数	89+
总字数	165,000+
主要库文件	5个
工具和模板	2个
可视化资源	3张
完整覆盖	论文全流程

---

❓ 常见问题

Q: 这些Prompt适用于所有学科吗？ A: 基本上适用。核心方法论对所有学科都有效。

Q: AI生成的论文清单准确吗？ A: AI提供初步清单 → Google Scholar验证 → 获得最终论文库。人工验证很重要。

Q: 可以直接用这些Prompt吗？ A: 可以直接复制使用。需要填入你的具体信息（[ ]标记的部分）。

Q: 这个项目会继续更新吗？ A: 是的。会持续增加更多学科版本、制作教程、改进Prompt。

---

🎉 现在就开始

3个快速选择：

1. 快速指南（5分钟） → 总索引
2. 项目介绍（10分钟） → 本文档
3. 直接用Prompt → 选择场景 → 复制 → 使用 → 获得结果

---

祝你的科研之旅顺利！ 如果有帮助，请给个Star支持一下 🌟

---

项目信息：v2.1 | 2024年2月 | MIT许可证 | 89+个Prompt | 165,000+字

返回顶部 | Read in English

---

English

Project files | Usage examples | Quick start | 简体中文

An AI prompt library for graduate students and academic researchers.

89+ prompts covering the workflow from research topic selection to literature review, with manuscript-writing and text-refinement resources added in v2.1.

The original overview estimates time savings of 50-70%, leaving more time for substantive research. These figures are not validated benchmarks.

The prompt libraries, templates, spreadsheet, and flowcharts linked below are primarily in Chinese. This English README explains how to use them and includes English versions of the two example prompts.

---

What problems does this project address?

Graduate students may recognize these challenges:

• Choosing a topic: Spending two or three weeks weighing ideas without knowing whether a supervisor will approve the final choice.
• Finding papers: Spending 10-20 hours searching manually and still worrying about missing relevant work.
• Organizing the literature: Not knowing how to classify, compare, and synthesize the papers already collected.
• Improving drafts: Revising AI-generated text with formulaic phrasing, an inconsistent voice, or other writing problems.
• Working efficiently: Repeating low-value tasks while running short of time.
• Managing pressure: Facing an approaching deadline with an unfinished research plan.

This project brings together structured workflows and ready-to-use prompts to help with these tasks. AI assistance complements, rather than replaces, your research judgment.

---

What can it help you do?

1. Research topic selection

Use a structured process to generate and evaluate potential topics:

• Generate 100 candidate topics with AI.
• Score candidates using a five-dimensional evaluation spreadsheet.
• Analyze the top three topics in depth.
• Prepare a proposal for discussion with your supervisor.

Original time estimate: Three to five days instead of two to three weeks. The original overview also suggests that better initial planning could avoid two to three months of later revisions; this is not a guaranteed outcome.

2. Paper discovery

Find an initial set of relevant papers more efficiently:

• Choose from eight search approaches, originally described as taking three to fifteen minutes.
• Ask AI to suggest candidate papers rather than starting every search manually.
• Verify each candidate using Google Scholar and the paper's original source.

Original time estimate: Two hours instead of twenty hours, with an estimated 80% reduction in search time. These figures are illustrative, not measured guarantees.

3. Literature review

Organize and analyze papers systematically:

• Classify papers using structured prompts.
• Develop a review framework and outline.
• Draft and revise the literature review around that framework.

Intended benefit: A clearer structure and a more efficient writing process. The original overview estimates a 50% improvement in writing efficiency.

4. Manuscript writing, expanded in v2.1

Work through the main stages of preparing a journal manuscript:

• Assess novelty and design the manuscript's structure.
• Draft the title, abstract, introduction, methods, and results.
• Use nine text-refinement prompts for polishing, expanding, shortening, and improving expression.
• Review the manuscript as a whole for consistency and clarity.

Original time estimate: Two to three weeks. The library is intended to support high-quality academic writing; it cannot guarantee publication or acceptance by a leading journal.

---

What is included?

File and directory names remain in Chinese so that the paths below match the repository.

academic-ai-prompt/
|-- README.md
|-- 索引                                           # Overall index
|-- 论文选题系列/
|   |-- 论文选题AI提示词库.md                       # Topic-selection prompts
|   |-- 论文选题论证方案模板.md                     # Three proposal templates
|   `-- 论文选题5维度评估表.xlsx                    # Five-dimensional evaluation tool
|-- 论文查找系列/
|   |-- 论文查找和文献综述AI提示词库.md             # Search and review prompts
|   `-- AI直接找论文的Prompt.md                     # Eight paper-discovery approaches
|-- 文献综述系列/
|   `-- 文献综述AI提示词库.md                       # Literature-review prompts
|-- 论文撰写系列/
|   `-- 顶刊论文写作AI提示词库-2.md                 # Writing and text-refinement prompts
|-- flowchart_topic_selection.png
|-- flowchart_paper_finding.png
|-- flowchart_literature_review.png
`-- LICENSE

Original collection summary: 89+ prompts, 165,000+ Chinese characters, three flowcharts, one Excel tool, and three complete proposal templates. The stated collection sizes are retained from the original overview; the current files are the source of truth.

Resource	Open the file
Topic-selection prompts	论文选题AI提示词库.md
Topic proposal templates	论文选题论证方案模板.md
Five-dimensional evaluation spreadsheet	论文选题5维度评估表.xlsx
Direct paper-discovery prompts	AI直接找论文的Prompt.md
Paper search and literature review	论文查找和文献综述AI提示词库.md
Literature-review prompts	文献综述AI提示词库.md
Manuscript-writing prompts	顶刊论文写作AI提示词库-2.md

---

Two detailed usage scenarios

The examples below translate the original README's scenarios. Their time estimates, date ranges, and sample outputs are illustrative. Update the dates for your project and verify all bibliographic details before use.

Example 1: Generate 100 potential research topics

Scenario: A master's student in computer science wants to identify a research topic combining machine learning and recommender systems.

Common difficulties:

• Brainstorming alone can take weeks without producing a clear choice.
• A busy supervisor may only be able to suggest a few broad directions.
• Reading fifty papers does not automatically reveal a suitable research question.
• A rushed decision may reveal feasibility problems later.

Step 1: Copy this prompt and adapt the background information.

I am a master's student planning a research topic in machine learning and
recommender systems.

My background and interests:
- Major: Computer Science and Technology, specializing in machine learning
- Interests: Deep learning, recommender systems, and natural language processing
- Existing skills: Deep learning fundamentals, Python programming, and reading
  academic papers
- Questions I care about: How can recommender systems become more accurate,
  faster, and fairer?

Research directions I am considering:
Deep learning for recommendation, graph neural networks for recommendation,
the cold-start problem, cross-domain recommendation, and explainable
personalized recommendation.

My supervisor's research interests:
Recommender systems, information retrieval, and user modeling.

Please generate 100 possible research topics that meet these requirements:
1. Stay within machine learning and recommender systems.
2. Balance novelty beyond a straightforward application of existing methods
   with feasibility within a two-year master's project.
3. Include theoretical contributions, methodological improvements, new
   applications, and engineering optimization.
4. Consider research directions discussed in 2023-2024. Replace this date
   range if a different period is appropriate for the project.
5. Connect to recommender systems and deep learning.

For each topic, provide:
- A concise academic title
- A one-sentence description of the research question
- An explanation of its potential novelty and practical value

Step 2: Send the prompt to an AI assistant such as ChatGPT or Claude.

The requested output is a candidate list with topic titles, descriptions, and explanations of potential value, spanning foundational questions and applied problems. The original example estimates five to thirty minutes to generate the initial list; reviewing it takes additional time.

Step 3: Screen and evaluate the candidates.

1. Use the five-dimensional evaluation spreadsheet to score the topics and calculate weighted scores.
2. Shortlist the ten most promising candidates.
3. Use Prompt 3.1 in the topic-selection library to analyze the top three in depth.

Step 4: Discuss and choose a topic.

Use the analysis to explain why the topic matters, how it differs from existing work, and whether it is feasible with your time, data, and resources. Bring that rationale to your supervisor rather than relying on an unexplained preference.

Original comparison: Two to three weeks of unstructured selection versus three to five days using the workflow. The original description also mentions topic-quality figures of 50% versus 90%+, but does not define or validate those scores.

---

Example 2: Build an initial list of 30-50 relevant papers

Scenario: You have selected deep-learning-based medical image segmentation as your topic and need a starting bibliography.

Common difficulties:

• A broad query such as "medical image segmentation" returns many results.
• Reading abstracts one by one is time-consuming.
• An unsystematic shortlist may vary widely in relevance and quality.

Copy and adapt this prompt from the direct paper-discovery library:

Please recommend 30-50 relevant research papers on deep-learning-based
medical image segmentation.

My requirements:
- Research area: Medical image processing and computer vision
- Core methods: Deep learning, CNNs, and segmentation architectures such as
  U-Net and FCN
- Applications: Lung lesion, tumor, organ, and vessel segmentation
- Include foundational papers, recent methodological work, application
  studies, and papers discussing limitations or open challenges
- Date range: 2016-2024, with an emphasis on 2022-2024. This range is retained
  from the original example; update it for your own review.
- Venue preferences: Relevant IEEE and Nature journals, MICCAI, CVPR,
  and comparable venues

For each candidate paper, provide:
- Its English title
- Authors and publication year
- Journal or conference
- A one-sentence summary of its main contribution
- Why it is relevant to my topic
- A recommendation score from 1 to 5

Rank the papers by recommendation score.

At the end, provide:
- A summary of how the field has developed over the previous ten years
- Major remaining challenges and research opportunities
- Suggested starting points for closer reading

Illustrative output from the original example:

Suggested priority reading

1. U-Net: Convolutional Networks for Biomedical Image Segmentation
   Authors: Ronneberger et al., 2015
   Venue: MICCAI 2015
   Contribution: Introduced the U-Net architecture for biomedical image
   segmentation.
   Relevance: A foundational architecture frequently used as a starting
   point for later segmentation methods.
   Suggested priority: 5/5

2. Attention U-Net: Learning Where to Look for the Pancreas
   Authors: Ozan Oktay et al., 2018
   Venue as listed in the original example: MIDL 2018
   Contribution: Introduced attention mechanisms into a U-Net-based
   segmentation approach.
   Relevance: An example of extending an existing architecture.
   Suggested priority: 5/5

... additional candidate papers ...

Example high-level timeline:
2010-2015: Traditional machine learning, FCNs, and increasing use of CNNs
2015-2018: U-Net and related architectures
2018-2021: Attention, multi-scale approaches, and multi-task learning
2021-2024: Vision Transformers, self-supervised learning, and few-shot learning

This sample is not a verified bibliography. Check every title, author list, year, venue, and contribution against the original paper or publisher record before citing it. AI recommendations are a starting point, not a substitute for a systematic search.

Original comparison: Ten to twenty hours of manual searching versus approximately five minutes to generate a candidate list. Verification, access to full texts, and reading require additional time.

---

Why use a structured prompt library?

The aim is not just to collect prompts, but to make the workflow more explicit:

• Move from open-ended brainstorming to generating, comparing, and evaluating candidate topics.
• Move from repeated ad hoc searches to AI-assisted candidate lists followed by human verification and reading.
• Move from an unreviewed AI draft to focused revision for clarity, consistency, and accuracy.

The original overview attributes its estimated 50-70% time savings to this more structured process. Actual results depend on the task, discipline, source material, model, and the amount of checking required.

---

Original workflow comparison

These are the original overview's estimates, not results from a controlled evaluation.

Task	Traditional approach described in the original	With this library	Originally stated benefit
Topic selection	Two to three weeks of uncertainty	Three to five days	Avoid two to three months of later revision
Paper discovery	Ten to twenty hours	Five minutes for an initial list	80% less search time
Literature review	No clear framework	Framework-guided writing	50% less writing time
Manuscript writing	Four to six weeks	Two to three weeks	40-50% less time
Text refinement	Repeated manual editing	Nine focused prompts	Improved draft quality
Overall workflow	Forty to sixty hours	Sixteen hours	50-70% less time

---

New in v2.1: Text refinement and academic expression

The writing library includes nine prompts for reviewing and revising manuscript text:

1. Prompt 7.1: Academic polishing. Improve clarity and academic expression.
2. Prompt 7.2: More natural phrasing. The source file calls this "lowering the AI-detection rate" and mentions a 30-50% reduction. This is an unverified claim, not a guarantee or a measure of originality.
3. Prompt 7.3: Paragraph expansion. Develop content and supporting arguments.
4. Prompt 7.4: Paragraph shortening. Remove redundancy and improve concision.
5. Prompt 7.5: Academic expression. Refine precision and scholarly tone.
6. Prompt 7.6: Human revision guidance. Plan improvements to AI-assisted drafts.
7. Prompt 7.7: Consistent style. Align wording and voice across the manuscript.
8. Prompt 7.8: Accuracy review. Examine the precision of academic claims and expression.
9. Prompt 7.9: Integrated revision. Review the text across several dimensions together.

See the text-refinement section of the writing library.

---

Quick start: Three scenarios

Scenario 1: I need a research topic

1. Open the topic-selection prompt library.
2. Copy Prompt 1.1 to generate candidate topics.
3. Replace the background information with your own and send it to your chosen AI assistant.
4. Score the candidates using the evaluation spreadsheet.
5. Use Prompt 3.1 to analyze the top three and discuss the shortlist with your supervisor.

Original planning estimate: Three to five days.

Scenario 2: I have a topic and need relevant papers

1. Open the direct paper-discovery prompts.
2. Choose Approach 8 for a brief request or Approach 1 for a more detailed request.
3. Fill in your topic, scope, and date range.
4. Ask for an initial list of thirty to fifty candidate papers, then verify their bibliographic details and relevance.

Original planning estimate: Around five minutes for an initial AI response, excluding verification and reading.

Scenario 3: I want to write a high-quality manuscript

1. Use Scenario 2 to begin building a relevant reading list; the original workflow suggests fifty to one hundred papers.
2. Open the literature-review library to work through the framework, synthesis, and writing stages.
3. Open the manuscript-writing library for section-by-section drafting.
4. Revise using the text-refinement prompts: 7.1 for polishing, 7.3/7.4 for length, 7.7 for consistency, and 7.9 for an integrated review. Prompt 7.2 addresses formulaic phrasing; follow applicable AI-use and disclosure requirements.

Original planning estimate: One to two weeks for this scenario. The original overview gives other writing estimates elsewhere; treat all of them as illustrative rather than fixed deadlines.

---

Usage recommendations

Recommended practices

1. Follow the workflow: Topic selection, paper discovery, literature review, then manuscript writing.
2. Keep AI in a supporting role: Use your own judgment to assess its output.
3. Try different prompts: Compare approaches when the first response is not useful.
4. Adapt the details: Refine the scope, constraints, and requested format as you learn more.
5. Review the result yourself: Check accuracy, sources, reasoning, and academic standards.

Important limitations

• Verify AI-suggested papers using Google Scholar and original sources.
• Check whether papers are current and relevant rather than relying only on citation counts.
• Adapt the prompts to the conventions of your discipline.
• Revise and check AI-generated text before using it in academic work.
• AI-detection scores are not measures of originality or academic quality. Follow your institution's and target publication's policies on AI use and disclosure.

---

Who is this for?

• Master's and doctoral students.
• Undergraduate students preparing a dissertation or final-year project.
• Academic researchers and creators of academic content.
• Researchers in engineering, natural sciences, social sciences, humanities, medicine, and related fields, with discipline-specific adaptation.

---

Project scale

The figures below reproduce the original project summary rather than an independently audited inventory.

Metric	Original stated value
Total prompts	89+
Total text length	165,000+ Chinese characters
Main prompt-library files	Five
Tools and template resources	Two
Visual resources	Three flowcharts
Coverage	The research-writing workflow

---

Frequently asked questions

Do the prompts work across disciplines?

The library is designed around general research workflows. Adapt its terminology, methods, evaluation criteria, and writing conventions to your own field.

Are the AI-generated paper lists accurate?

Treat them as candidate lists. Verify each entry before adding it to your bibliography or relying on its claims.

Can I copy and use the prompts directly?

Yes. Replace the placeholders marked with square brackets and supply the details of your own project. The linked libraries are primarily in Chinese; this README includes two English examples to help you get started.

Will the project continue to be updated?

The original roadmap describes additional discipline-specific versions, tutorials, and prompt improvements. Check the repository's latest commits for available updates.

---

Get started

1. Read the quick guide: Open the overall index, currently in Chinese.
2. Explore the project: Use the resource table to choose a library.
3. Try a prompt: Select a scenario, adapt the example, review the output, and refine it.

If the library is useful, consider starring the repository.

Project information from the original overview: v2.1 | February 2024 | MIT License | 89+ prompts | 165,000+ Chinese characters.

Back to English | 返回中文