arxiv-survey

Survey arXiv papers from a given year to present on a specific theme. Automatically categorizes papers, translates abstracts to Chinese, and generates a structured markdown report.

Generate comprehensive survey reports on arXiv papers for a specific theme and time range.

Usage

bash

/skill arxiv-survey survey <year>:<theme>

Parameters

• year: Starting year for paper collection (e.g., 2026)
• theme: Topic or theme description (can be a short keyword or detailed description)

Examples

bash

/skill arxiv-survey survey 2026:AI 辅助开源贡献
/skill arxiv-survey survey 2024:RAG retrieval augmented generation
/skill arxiv-survey survey 2025:LLM code generation automated program repair
/skill arxiv-survey survey 2026:multimodal learning

Workflow

1. Search: Query arXiv API for papers from <year> to present matching <theme>
2. Filter: Collect 10-50 relevant papers (more if the theme has abundant literature)
3. Categorize: Automatically classify papers into thematic categories based on titles and abstracts
4. Translate: Translate abstracts to fluent Chinese
5. Generate Report: Create a structured markdown file with:
   • Table of Contents: Categories with paper titles listed under each
   • Detailed Content: For each category, list papers with:
     • English title
     • Authors
     • Chinese translated abstract (expanded if full text is accessible)

Output

Generates a markdown file: arxiv-survey-<year>-<theme-slug>.md

Report Structure

markdown

# ArXiv Survey: <Theme> (<Year>-Present)

## Table of Contents

### Category 1
- Paper Title 1
- Paper Title 2

### Category 2
- Paper Title 3

---

## Detailed Papers

### Category 1

#### Paper Title 1
**Authors**: Author List  
**arXiv**: [Link](https://arxiv.org/abs/xxxx.xxxxx)  
**Abstract (中文)**: Translated abstract...

#### Paper Title 2
...

Scripts

• scripts/survey_arxiv.sh <year> <theme> - Main survey script that searches, categorizes, and generates the report

Notes

• Paper count adapts to availability: niche topics may yield ~10 papers, popular topics up to 50
• Abstracts are translated to Chinese for readability
• If full paper content is accessible, abstracts may be expanded with additional context
• All file names are in English (no Chinese characters)