The open-source LLM textbook that puts each chapter in the zoo

What it does This repository hosts Foundations of LLMs, a Chinese-language textbook that walks readers through language model fundamentals, architecture evolution, prompt engineering, parameter-efficient fine-tuning, model editing, and retrieval-augmented generation. The authors from Zhejiang University provide the full PDF, chapter-by-chapter PDFs, and a living directory of related papers for each topic. It is explicitly positioned as an open, community-driven alternative to closed academic courses, with the authors committing to monthly revisions and expanding the catalog to cover inference acceleration and agents.

The interesting bit Each chapter adopts a different animal as its narrative backdrop to make the technical content more approachable—hence the six-animal cover. It is a rare attempt to make a rigorous, citation-heavy textbook read like a guided safari rather than a sleep aid.

Key highlights

• Covers six core topics, from statistical language models to RAG, with architecture deep-dives including non-Transformer designs.
• Maintains a curated paper list per chapter that the authors update to track latest progress.
• Released as both a complete PDF and split chapter files for modular reading.
• Explicitly plans future chapters on inference acceleration and LLM agents.
• Community feedback is solicited via GitHub issues, with contributors acknowledged by name.

Caveats

• The entire text is in Chinese, so it is inaccessible to readers without proficiency in the language.
• The repository contains PDFs and paper lists rather than executable code or interactive notebooks.

Verdict Worth bookmarking if you read Chinese and want a structured, actively maintained survey of LLM fundamentals. If you are looking for hands-on code or an English resource, this is not your habitat.