@Jolyne_AI: GitHub 上有个开源小工具:daily-arXiv-ai-enhanced,帮你把“追论文”这件事变成每天自动完成的日常。 它会每日抓取 arXiv 最新论文,并用 DeepSeek 等大模型生成中文摘要,让你用更少时间,快速跟上 A…
摘要
daily-arXiv-ai-enhanced 是一个开源工具,通过GitHub Actions每日自动抓取arXiv最新论文,并使用DeepSeek等大模型生成中文摘要,帮助快速跟进AI研究进展。
查看缓存全文
缓存时间: 2026/07/06 12:12
GitHub 上有个开源小工具:daily-arXiv-ai-enhanced,帮你把“追论文”这件事变成每天自动完成的日常。
它会每日抓取 arXiv 最新论文,并用 DeepSeek 等大模型生成中文摘要,让你用更少时间,快速跟上 AI 领域最新研究进展。
GitHub: http://github.com/dw-dengwei/daily-arXiv-ai-enhanced…
官网: https://dw-dengwei.github.io/daily-arXiv-ai-enhanced…
主要亮点:
- 每日自动更新:覆盖计算机视觉 / 图形学 / 自然语言处理三大方向
- 中文摘要自动生成:用 LLM 提炼要点,显著降低阅读成本
- 高度可定制:可自选论文类别、摘要语言与使用的 AI 模型
- 零服务器部署:基于 GitHub Actions 全自动运行,省心省力
- 历史按日期归档:随时回查过去的论文更新
- 个性化配置:支持邮箱与用户信息等定制选项
上手也很简单:Fork 项目,填好 API Key 即可跑起来。适合想每天稳定“刷到”最新论文的人。
dw-dengwei/daily-arXiv-ai-enhanced
Source: https://github.com/dw-dengwei/daily-arXiv-ai-enhanced
🚀 daily-arXiv-ai-enhanced
若您所在法域对学术数据有审查要求,谨慎运行本代码;任何二次分发版本必须履行合规审查(包括但不限于原始论文合规性、AI合规性)义务,否则一切法律后果由下游自行承担。
If your jurisdiction has censorship requirements for academic data, run this code with caution; any secondary distribution version must remove the entrance accessible to China and fulfill the content review obligations, otherwise all legal consequences will be borne by the downstream.
This innovative tool transforms how you stay updated with arXiv papers by combining automated crawling with AI-powered summarization.
✨ Key Features
🎯 Zero Infrastructure Required
- Leverages GitHub Actions and Pages - no server needed
- Completely free to deploy and use
🤖 Smart AI Summarization
- Daily paper crawling with DeepSeek-powered summaries
- Cost-effective: Only ~0.2 CNY per day
💫 Smart Reading Experience
- Personalized paper highlighting based on your interests
- Cross-device compatibility (desktop & mobile)
- Local preference storage for privacy
- Flexible date range filtering
🧩 SKILL System
- Plug-and-play skill modules for customizing paper filtering
⚙️ Easy Preference Export & Integration
- One-click copy in Settings to export your keywords and authors configuration
- Seamlessly combine exported preferences with SKILL for reproducible and shareable setups
👉 Try it now! - No installation required
https://github.com/user-attachments/assets/b25712a4-fb8d-484f-863d-e8da6922f9d7
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 watch the video above first and directly use this repo in https://dw-dengwei.github.io/daily-arXiv-ai-enhanced/. Please star it if you like :)
If you want to customize categories, LLMs, or languages, click here.
Instructions
- Fork this repo to your own account and delete my own information in buy-me-a-coffee.
- Go to: your-own-repo -> Settings -> Secrets and variables -> Actions
- Go to Secrets. Secrets are encrypted and used for sensitive data
- Create two repository secrets named
OPENAI_API_KEYandOPENAI_BASE_URL, and input corresponding values. - [Optional] Set a password in
secrets.ACCESS_PASSWORDif you do not wish others to access your page. (see https://github.com/dw-dengwei/daily-arXiv-ai-enhanced/pull/64) - Go to Variables. Variables are shown as plain text and are used for non-sensitive data
- Create the following repository variables:
CATEGORIES: separate the categories with “,”, such as “cs.CL, cs.CV”LANGUAGE: such as “Chinese” or “English”MODEL_NAME: such as “deepseek-chat”EMAIL: your email for push to GitHubNAME: your name for push to GitHub
- Go to your-own-repo -> Actions -> arXiv-daily-ai-enhanced
- 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 - Set up GitHub pages: Go to your own repo -> Settings -> Pages. In
Build and deployment, setSource="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.
Contributors
Thanks to the following special contributors for contributing code, discovering bugs, and sharing useful ideas for this project!!!
|
JianGuanTHU |
Chi-hong22 |
chaozg |
quantum-ctrl |
Zhao2z |
eclipse0922 |
|
xuemian168 |
Lrrrr549 |
AinzRimuru |
fengxueguiren |
zerocpp |
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
Buy me a coffee
相似文章
@VincentLogic: 每天被 Arxiv 的新论文淹没?头都大了。 刚发现一个宝藏网站,专门聚合最新的 AI 论文和模型基准测试(Benchmarks)。 界面很干净,直接看 Trending 或者按周/月筛选。最爽的是每篇论文都直接关联了它用到的 bench…
推荐一个免费网站 sophon.at/papers,聚合最新 AI 论文和模型基准测试,界面干净,支持按 Trending 或周/月筛选,每篇论文直接关联所用 benchmark 和 model。
@GitHub_Daily: 用 AI 写的文章,读起来总有一股 AI 味,句式工整、用词重复,发布或提交后总担心被检测到。 最近偶然找到了 AI Humanize Text 这个开源工具,专门把 AI 生成的文本改写成更自然的人类写作风格。 提供了四种不同的改写思路…
介绍了一个名为 AI Humanize Text 的开源工具,通过多语言翻译链、大模型多轮改写等方法,将 AI 生成的文本改写得更自然,避免被检测。
@GitHub_Daily: AI-research-feedback,一个给 Claude Code 用的学术审稿技能。 核心能同时跑六个审稿智能体,分别盯语法、前后一致性、公式、图表和论证漏洞。 还能指定 QJE、AER 这类期刊,模拟对应审稿人的挑剔程度,最后合…
AI-research-feedback 是一个为 Claude Code 设计的学术审稿技能,通过六个并行智能体检查语法、一致性、公式、图表和论证漏洞,支持指定期刊模拟审稿人,最后生成结构化审稿报告。
@gaoren7716: 写论文这件事,可能要被流程化 AI 系统重写了 不是帮你润色,不是帮你改一句话,而是从选题开始就有 13 个 Agent 在协作 功能清单: Deep Research(13-agent 调研团队) Systematic Review(P…
介绍了一套名为Academic Research Skills的开源AI工具,通过13个Agent协作实现从选题到写作、审稿的全流程学术研究自动化,可作为Claude Code插件使用,将学术研究变为标准化生产线。
@_ar9av: 连续第6天每天阅读一篇关于AI的arXiv论文并分享真正印象深刻的内容:AutoSci(北京大学)概要:…
一条推文介绍北京大学开发的AutoSci系统,该系统可自动化从文献调研到回复审稿意见的整个研究流程,并在项目间进行自我改进。

