ziwenhahaha/daily-paper-reader

What it solves

Daily Paper Reader provides a centralized, automated system for researchers to discover and read AI papers without needing a dedicated server. It solves the problem of information overload by filtering arXiv, OpenReview, and major AI conferences (like NeurIPS, ICLR, ICML, ACL) to create a personalized feed of research based on specific keywords and interests.

How it works

The project uses a serverless architecture leveraging GitHub Actions for automation and GitHub Pages for the frontend. It employs a multi-stage pipeline:

1. Discovery: Automatically fetches new papers from arXiv, OpenReview, and other sources (including bioRxiv, medRxiv, and ChemRxiv).
2. Recommendation: Uses a combination of semantic search (embedding models like BGE), reranking (Qwen-Reranker), and LLM refinement (DeepSeek) to filter and rank papers based on user-defined tags and intent queries.
3. Reading Experience: Generates AI-powered summaries and provides an integrated reading interface where users can read abstracts, full-text PDFs, and long summaries in one place.
4. Interaction: Integrates an AI chat interface for asking questions about the papers while reading.

Who it’s for

• Individual Researchers: Those who need to track specific research directions and stay updated on daily publications.
• Academic Labs: Teams wanting to create a shared public paper board to track collective research interests.
• ** uma Reading Workstations**: Users looking for combine discovery, reading, and AI-assisted summary in a single tool.

Highlights

• Zero-Server Deployment: Runs entirely on GitHub Actions and GitHub Pages.
• Multi-Source Integration: Syncs data from arXiv, OpenReview, and 9+ major AI conferences.
• AI-Enhanced Reading: Features AI-generated summaries, automated figure extraction from PDFs, and an integrated AI chat for paper-specific Q&A.
• Personalized Feed: Highly customizable recommendation flow based on keywords and research directions.