ZeeLin Academic Paper

📋 Privacy and Transparency

Local Processing:

All paper generation processes are completed locally
No user paper content, references, or research data is uploaded to any server
No personal information or research data is collected

Data Security:

✅ Fully localized operation
✅ No external API calls
✅ No risk of data leakage
✅ Complies with academic confidentiality requirements

When to Use

When the user requests:

Write a science and engineering paper / generate an academic paper
Generate research hypotheses and research design
Output a results analysis chapter
Produce text according to a science and engineering paper template
Write a paper based on references

Not suitable for:

Humanities papers (history, philosophy, literature, etc.)
Business reports or market analysis
Non-academic writing

Core Features

🎯 Full-Process Automation

One-stop generation from scratch to a complete paper:

Intelligent Topic Generation
- Automatically extract research gaps based on references
- Ensure innovation and academic value
- Avoid duplication and low-value topics
Structured Outline Design
- Automatically generate a standard five-section structure (introduction, literature review, argumentation, suggestions, conclusion)
- Support customizing the number of arguments and suggestions
- Present the logical framework clearly in JSON format
Chapter-by-Chapter Content Generation
- Introduction: research background, problem statement, innovation points
- Literature review: literature review, research process
- Argumentation: theoretical framework, research hypotheses, research design
- Results analysis: data interpretation, explanation of findings
- Conclusion: summary, limitations, future research directions
- Abstract and keywords: distill the core content

✨ Academic Quality Assurance

Logical consistency check: ensure methods, hypotheses, and results correspond to one another
Academic language: rigorous, concise, avoid colloquialism
Theoretical depth: emphasize innovation and theoretical contribution
Verifiability: research hypotheses, data, methods, and conclusions are closely linked

🛠️ Flexible Customization

Customize the number of argument subsections (default 3)
Customize the number of suggestions (default 3)
Customize the number of keywords (default 5)
Support user-provided topics or automatic generation
Support plain text or structured JSON output

Agent Execution Flow

Step 1: Collect Input

Ask for or infer what the user needs:

Required Input：

Reference list (at least 1-3 papers)
Review materials or research background

Optional Input：

Paper title (automatically generated if none)
Number of arguments (default 3)
Suggested quantity (default 3)
Number of keywords (default 5)

Interaction example：

"Hello! I can help you generate a science and engineering paper 📝

Please provide:

Reference titles or content (at least 1-3)

Research background or review materials

(Optional) paper title, number of arguments, suggested quantity

What direction do you want to research?"

Step 2: Generate the title (if needed)

If the user has not provided a title, call the title generation template:

Analyze the research gap based on the references
Propose an innovative research title
Ensure academic value and theoretical potential

Example output：

"Based on your references, the paper title I suggest is:

"Research on an AI-Assisted Urban Traffic Flow Prediction Model"

Are you satisfied? (Reply 'confirm' or 'change to XXX')"

Step 3: Generate the outline

Call the outline generation template and output a JSON structure:

{
  "引言": "概述研究背景、问题与创新点",
  "综述": [
    {"zs": "城市交通流量预测研究现状"},
    {"zs": "人工智能在交通预测中的应用"}
  ],
  "论证": [
    {"lz": "基于深度学习的流量预测理论框架"},
    {"lz": "研究假设与变量设定"},
    {"lz": "实验设计与数据采集方案"}
  ],
  "建议": [
    {"jy": "优化城市交通管理策略"},
    {"jy": "推广智能交通系统应用"},
    {"jy": "加强跨学科协作研究"}
  ],
  "结论": "总结研究发现、局限性与未来方向"
}

Show it to the user for confirmation：

"The paper outline has been generated, please review: [display JSON outline]

Are you satisfied? Do you need to adjust the number of chapters or the titles?"

Step 4: Generate content chapter by chapter

Generate each chapter in order:

Introduction → show to the user for confirmation
Review concluding paragraph and research process → confirm
Argumentation (generate each subsection one by one) → confirm
Results analysis → confirm
Conclusion → confirm
Abstract and keywords → complete

After each chapter is generated, ask the user:

"The 【Introduction】 chapter has been generated, please review: [content]

Are you satisfied? (Reply 'confirm' to continue to the next chapter, or 'change XXX')"

Step 5: Output the complete paper

Merge all chapters and output the complete paper (plain text format, without Markdown):

【题目】
人工智能辅助的城市交通流量预测模型研究

【摘要】
本研究基于深度学习技术...

【关键词】
人工智能；交通流量预测；深度学习；...

一、引言
...

二、综述
...

三、论证
（一）基于深度学习的流量预测理论框架
...

四、建议
（一）优化城市交通管理策略
...

五、结论
...

Resource usage

Load template files：

Read references/prompts_zh.md to obtain all chapter templates
Directly reuse the placeholders and constraints in the templates
Do not rewrite the template structure

Template priority：

If user constraints conflict with template constraints, prioritize user constraints and explain the adjustment

Data generation：

When the template requires simulated data or table content, generate synthetic examples consistent with the topic
Ensure the data is logically self-consistent and fits the research context

Output rules

Academic standards

Language style: rigorous, concise, academic, avoid colloquialisms
Logical thread: clarify causal, progressive, inductive, or contrastive relationships
Innovativeness: raise new questions, new perspectives, and develop the argument
Practical significance: reflect real-world relevance and provide usable insights
Consistent style: all chapters have a consistent style, do not use Markdown

Format requirements

Plain text output (unless the user specifies another format)
Clear heading hierarchy (1, 2, 3 / (1)(2)(3))
No special symbols (asterisks, quotation marks, book title marks, etc.)
No redundant line breaks

Quality Check

Method consistency: methods mentioned in the introduction must be reflected in the argumentation and results analysis
Hypothesis-result correspondence: research hypotheses are closely linked to the results analysis
Citation standards: all statements are verifiable (based on references)

Quick Command Examples

/paper 基于深度学习的流量预测研究
/paper 根据这 3 篇文献写一篇论文 [文献列表]
/paper 生成研究假设与研究设计

Notes

Academic Integrity

⚠️ Important Reminders：

Content generated by this tool is for reference and inspiration only
Users are responsible for the final paper content
All reference sources must be cited
Do not plagiarize or fabricate data
Manual review and revision of generated content is recommended

Usage Recommendations

Provide high-quality references: The quality of references directly affects the quality of generation
Confirm step by step: Confirm each chapter promptly after generation to avoid major subsequent revisions
Manual polishing: After generation, manual review, supplementing details, and adjusting wording are recommended
Data authenticity: If real data is needed, please replace the generated simulated data

Support and Feedback

Developer: 刘东江 (@lidngjing317853)
Project homepage: https://melodylab.top
Issue feedback: Contact via ClawHub or the project homepage

Example Dialogue Flow

User:"Help me write a paper on the application of artificial intelligence in medical diagnosis"

Agent：

"Hello! I can help you generate a STEM paper 📝

Please provide:

Reference titles or content (at least 1-3)

Research background or review materials

Regarding 'the application of artificial intelligence in medical diagnosis,' what references do you have?"

User:"Reference 1: Research progress on deep learning in medical imaging diagnosis; Reference 2: ..."