Automatically fetch, filter, and summarize research papers from arXiv & IACR — with AI-generated bilingual (Chinese/English) summaries. Deployed for free on GitHub Pages, updated daily via GitHub Actions.
- Tired of manually checking arXiv every day? Paper Pulse fetches papers automatically.
- Only care about specific topics? Flexible keyword filtering (AND/OR logic) keeps only what matters.
- Need Chinese + English summaries? AI generates both — toggle per card with one click.
- Don't want to pay for hosting? Runs entirely on GitHub Actions + Pages. Zero cost.
| Feature | Description |
|---|---|
| Multi-source | Fetches from arXiv (configurable categories) and IACR ePrint |
| Keyword filtering | OR between lines, AND within a line — fine-grained control |
| Bilingual AI summaries | Chinese + English via Qwen (DashScope API), toggle per card |
| Daily automation | GitHub Actions cron job, auto-commits results |
| RSS feed | Subscribe in any reader — feed.xml auto-generated |
| BibTeX export | Single paper or bulk export |
| Email digest | Daily report with stats and token usage |
| Static site | No server needed — GitHub Pages serves everything |
| Markdown summaries | Rich formatting in AI-generated summaries |
Click "Use this template" (or fork) to create your own copy.
Sign up at DashScope Console — free tier available.
Go to Settings → Secrets and variables → Actions → New repository secret:
- Name:
MODELSCOPE_API_KEY(orDASHSCOPE_API_KEY) - Value: your API key
Go to Settings → Pages:
- Source: Deploy from a branch
- Branch:
master, Folder:/ (root)
Go to Actions → Fetch Papers → Run workflow. After it completes, visit https://<username>.github.io/<repo-name>/.
From now on, papers are fetched automatically every day at 00:00 UTC.
Paper Pulse can send a daily email report with statistics (new papers, failed summaries, token usage). To enable it, add three more secrets in Settings → Secrets and variables → Actions:
| Secret | Description |
|---|---|
EMAIL_USERNAME | Gmail address used to send the report (e.g. you@gmail.com) |
EMAIL_PASSWORD | Gmail App Password (not your login password) |
EMAIL_TO | Recipient address (can be the same as EMAIL_USERNAME) |
Note: Gmail requires an App Password — you must enable 2-Step Verification on your Google account first, then generate an App Password under Security → App passwords. Regular Gmail passwords will not work.
If these secrets are not set, the workflow still runs normally — the email step is simply skipped.
# Each line = OR condition. Words on same line = AND condition. transformer # Papers containing "transformer" neural backdoor # Papers containing BOTH "neural" AND "backdoor" federated learning # Papers containing "federated learning" Keyword filtering can be independently toggled per source (apply_to_arxiv / apply_to_iacr in config.toml).
| Setting | Where | Default |
|---|---|---|
| Paper retention | general.days_back | 30 days |
| arXiv categories | fetchers.arxiv.categories | cs.CR, cs.AI, cs.LG, cs.CL |
| AI model | summarizer.model | qwen-plus |
| RSS items | rss.max_items | 50 |
| Site URL | general.site_url | (your GitHub Pages URL) |
See CONFIG_GUIDE.md for all options.
paper-pulse/ ├── .github/workflows/ │ └── fetch-papers.yml # Daily automation ├── scripts/ │ ├── fetchers/ │ │ ├── arxiv.py # arXiv API fetcher │ │ └── iacr.py # IACR RSS fetcher │ ├── filter.py # Keyword filtering engine │ ├── summarizer.py # Bilingual AI summarization │ ├── rss.py # RSS feed generator │ └── main.py # Pipeline orchestrator ├── data/ │ ├── papers.json # Paper database │ └── failed.json # Failed summarization queue ├── config.toml # All configuration ├── keywords.txt # Keyword filter rules ├── index.html / app.js / styles.css # Frontend └── feed.xml # RSS feed (auto-generated) Fetch (arXiv + IACR) → Filter (keyword matching) → Summarize (Qwen AI, bilingual) → Merge & Deduplicate → Save (papers.json) → Generate RSS (feed.xml) → Commit & Push (GitHub Actions) Failed summaries are retried automatically on the next run.