This project provides an automated system for daily crawling and summarizing arXiv papers using Large Language Models (LLMs). It targets researchers and academics seeking concise, AI-generated summaries of new papers, with a focus on specific AI categories. The primary benefit is efficient discovery and understanding of relevant research without needing to read full papers.

How It Works

The system leverages GitHub Actions for daily execution and GitHub Pages for hosting a front-end interface. It crawls specified arXiv categories (defaulting to cs.CV, cs.GR, cs.CL, cs.AI) and uses DeepSeek for summarization, aiming for cost-efficiency by running during off-peak hours. The front-end utilizes browser LocalStorage for user preferences like keywords and authors, highlighting matching papers and offering filtering by date range.

Quick Start & Requirements

• Install/Run: Fork the repository, configure GitHub Secrets (OPENAI_API_KEY, OPENAI_BASE_URL) and Variables (CATEGORIES, LANGUAGE, MODEL_NAME, EMAIL, NAME), and manually trigger the workflow via GitHub Actions.
• Prerequisites: GitHub account, API access to an LLM (e.g., DeepSeek), and basic understanding of GitHub Actions and repository settings.
• Setup: Configuration involves setting secrets and variables. Workflow execution can take approximately one hour. GitHub Pages deployment requires branch configuration.
• Links: Demo

Highlighted Details

• Serverless architecture using GitHub Actions and Pages.
• Personalized paper highlighting based on user-defined keywords and authors.
• Front-end interface designed for both desktop and mobile viewing.
• Includes a "Statistics" page for keyword trend analysis (under development).

Maintenance & Community

Notable contributors include JianGuanTHU and Chi-hong22. The project acknowledges support from Github_Daily and AIGCLINK.

Licensing & Compatibility

The repository's license is not explicitly stated in the README. Compatibility for commercial use or closed-source linking is not specified.

Limitations & Caveats

The statistics page has known bugs regarding keyword counts and date/week correspondence. Summarization quality is dependent on the chosen LLM and prompt engineering. Large date range selections in the date picker may cause browser performance issues.