Daily Papers: Personalized ArXiv Research Digest
Overview
Daily Papers is an intelligent research paper aggregation and summarization system that helps researchers stay up-to-date with the latest developments in their field. By leveraging Large Language Models (LLMs), it automatically curates personalized daily digests of relevant papers from arXiv and HuggingFace.
Key Features
Multi-Source Paper Aggregation
- ArXiv Integration: Fetches papers from specified arXiv categories (cs.AI, cs.LG, cs.CV, cs.CL, cs.RO, etc.)
- HuggingFace Trending: Tracks trending and daily papers from the HuggingFace platform
- Social Metrics: Displays upvotes, GitHub stars, and trending indicators
AI-Powered Curation
- Smart Filtering: LLM-based relevance assessment based on your research interests and custom prompts
- Intelligent Ranking: Automatically selects the most valuable papers when count exceeds limit
- Deep Summarization: Generates comprehensive summaries highlighting key contributions, methodology, and results
Automated Delivery
- Email Reports: Markdown and HTML-formatted reports delivered directly to your inbox
- Multi-User Support: Configure multiple users with independent settings and preferences
- Cost Tracking: Monitors token usage and provides cost estimates
Deployment Options
- GitHub Actions: Free, zero-maintenance deployment using GitHub’s infrastructure
- Local Scheduler: Option to run on local machines with customizable scheduling
- Minimal Cost: Approximately $0.50/day using DeepSeek V3 (~$15/month)
Technical Implementation
Architecture
- Built in Python with a modular design for easy customization
- RESTful API integration with multiple LLM providers (DeepSeek, OpenAI-compatible endpoints)
- Configurable filtering and summarization prompts for different research domains
- Robust error handling and retry mechanisms
Workflow
- Fetch: Retrieves latest papers from configured sources
- Filter: LLM evaluates relevance based on user preferences
- Rank: Selects top papers when volume exceeds threshold
- Summarize: Generates detailed analysis of each paper
- Deliver: Formats and sends email reports with original abstracts and AI summaries
Configuration
Users can customize:
- arXiv categories and lookback window
- Maximum papers per digest
- Custom filtering, ranking, and summarization prompts
- Multiple user profiles with independent settings
Use Cases
- Researchers: Stay current with developments in specific subfields
- PhD Students: Track relevant papers for literature reviews
- Industry Professionals: Monitor practical applications and sota techniques
- Study Groups: Share curated digests with team members
Cost Efficiency
With DeepSeek V3 pricing (~$0.27/M input tokens, ~$1.10/M output tokens), the system costs approximately:
- $0.50/day for 10 papers (~$15/month)
- Alternative models: DeepSeek V3.2-Exp, Qwen-Turbo, GLM-4-Long (~$0.30-0.40/day)
- Free option: Gemini 2.0 Flash with generous free tier
Acknowledgments
Inspired by:
- arxiv-sanity by Andrej Karpathy
- daily-arXiv-ai-enhanced
- customize-arxiv-daily
- arxiv-pusher