About

This tool will daily crawl https://arxiv.org and use LLMs to summarize them.

Try in: https://dw-dengwei.github.io/daily-arXiv-ai-enhanced/

Features

• Using the free features of GitHub Actions and GitHub Pages, no server is required
• Crawling data starts at dawn every day, and using DeepSeek to summarize. This period is during the off-peak discount period of DeepSeek, and it only costs about 0.2 CNY per day.
• Provides a GitHub Pages front-end interface, uses LocalStorage to store personalized preference information (such as keywords and authors of interest), and highlights papers that matches the preferences.
• GitHub Pages takes into account the display effects of both the computer and mobile devices, ensuring that papers can be easily reviewed on mobile devices

Screenshots

• Main page. Highlight the interested keywords and authors.

• Setting page. Set up keywords and authors and store them in your browser.

• Detail page. Show details of the paper you clicked.

• Date select. Enable selecting a single date or a date range for filtering papers (Notice: a large date range will show lots of papers, which may lead your browser to get stuck.).

• Statistics page (in developing). Help you analyze papers. Extract keywords for papers in the day(s) you select. In addition, if you select a range of dates, the keyword trends will be illustrated. (Fortunately, selecting a large range of papers will not stuck your browser to be stuck because this page will not show all papers. It may take a few seconds to process the keywords.)

How to use

This repo will daily crawl arXiv papers about cs.CV, cs.GR, cs.CL and cs.AI, and use DeepSeek to summarize the papers in Chinese. If you wish to crawl other arXiv categories, use other LLMs, or other languages, please follow the instructions. Otherwise, you can directly use this repo in https://dw-dengwei.github.io/daily-arXiv-ai-enhanced/. Please star it if you like :)

Instructions:

1. Fork this repo to your own account
2. Go to: your-own-repo -> Settings -> Secrets and variables -> Actions
3. Go to Secrets. Secrets are encrypted and used for sensitive data
4. Create two repository secrets named OPENAI_API_KEY and OPENAI_BASE_URL, and input corresponding values.
5. Go to Variables. Variables are shown as plain text and are used for non-sensitive data
6. Create the following repository variables:
   1. CATEGORIES: separate the categories with ",", such as "cs.CL, cs.CV"
   2. LANGUAGE: such as "Chinese" or "English"
   3. MODEL_NAME: such as "deepseek-chat"
   4. EMAIL: your email for push to GitHub
   5. NAME: your name for push to GitHub
7. Go to your-own-repo -> Actions -> arXiv-daily-ai-enhanced
8. You can manually click Run workflow to test if it works well (it may take about one hour). By default, this action will automatically run every day. You can modify it in .github/workflows/run.yml
9. Set up GitHub pages: Go to your own repo -> Settings -> Pages. In Build and deployment, set Source="Deploy from a branch", Branch="main", "/(root)". Wait for a few minutes, go to https://<username>.github.io/daily-arXiv-ai-enhanced/. Please see this issue for more precise instructions.

To-do list

• Feature: Replace markdown with GitHub pages front-end.
• Bugfix: In the statistics page, the number of papers for a keyword is not correct.
• Bugfix: In the date picker, the date and week do not correspond.
• Feature: Extract keywords with DeepSeek.
• Update instructions for fork users about how to use GitHub Pages.

Contributors

Thanks to the following special contributors for this project!!!

JianGuanTHU

Chi-hong22

Acknowledgement

We sincerely thank the following individuals and organizations for their promotion and support!!!

Github_Daily

AIGCLINK

阮一峰的网络日志 科技爱好者周刊（第 353 期）

《HelloGitHub》月刊 第 111 期

Star history