What is fun-rec?

Fun-Rec is a Chinese-language textbook and learning resource about recommendation systems, published by Datawhale, a Chinese AI learning community. A recommendation system is the kind of software that decides which videos, products, or articles to show you based on your past behavior. This project covers both the foundational techniques and more recent approaches driven by large AI models. The content is split into two main sections. The first covers the traditional multi-stage pipeline that most large platforms use: an initial candidate retrieval step that quickly narrows millions of items to a few thousand, followed by ranking and re-ranking steps that order those candidates more carefully. Topics include collaborative filtering, vector-based retrieval, feature crossing, sequential modeling, and multi-objective ranking. The second section focuses on generative recommendation, a newer direction where large language models and diffusion models take a more direct role in producing recommendations. Chapters cover scaling laws for recommendation models, end-to-end generative modeling, and reasoning-based approaches where the model thinks through item selection step by step. The final chapter walks through building a production-grade recommendation system from scratch, including offline pipelines, online serving, and deployment. The project is still actively being updated. It is primarily written for readers who already have a machine learning background and want to understand how recommendation algorithms work in real products. The README and all content are in Chinese, with a link to an English version of the README.