LLM、RAG 与知识图谱

Large Language Models, Retrieval-Augmented Generation, and Knowledge Graphs

This course introduces the foundations of large language models, retrieval-augmented generation, knowledge graph construction, GraphRAG, evaluation, and research applications. It is designed for students who want to build knowledge-grounded AI systems for literature analysis, policy intelligence, patent mining, healthcare information, and enterprise knowledge services.

LLM-RAG-KG Graduate students / applied AI learners Chinese with English terms

Course Introduction

LLM、RAG 与知识图谱关注的是如何把大语言模型从“通用生成器”转化为“知识约束下的智能分析系统”。在真实研究和应用场景中，模型不能只依赖参数中的隐性知识，而需要连接外部文献、政策、专利、企业文档、医学知识库和结构化知识资源。

This course emphasizes three abilities: designing retrieval-augmented workflows, constructing knowledge graphs from textual evidence, and evaluating whether an LLM system produces grounded, traceable, and useful answers.

Why This Course Matters

RAG 解决的是“从哪里找证据”的问题，知识图谱解决的是“知识之间如何组织和推理”的问题，而 LLM 则负责把证据、结构和任务目标转化为可读、可解释的输出。三者结合，可以支撑更加可靠的知识发现、问答系统、研究综述、专利情报分析和医学信息服务。

Outline

Course Directory

Click a section to open the Markdown-based teaching document.

Module 1

LLM Foundations and Knowledge Grounding

Understand why LLMs need external knowledge, and how retrieval, tools, and structured knowledge reduce hallucination.

LLM-01

LLM 与知识 grounding

Understand the limits of parametric knowledge and why LLM systems need external evidence, retrieval, and structured constraints.

→

Module 2

RAG Pipeline and Vector Retrieval

Learn document preprocessing, chunking, embeddings, vector databases, reranking, and prompt assembly.

LLM-02

RAG 流程：切分、向量化与检索

Learn how to build a retrieval-augmented generation pipeline from raw documents to grounded answers.

→

Module 3

Knowledge Graph Construction and Semantic Retrieval

Build entities, relations, triples, ontologies, and graph-based retrieval structures from documents and domain knowledge.

LLM-03

知识图谱构建：实体、关系与本体

Learn how to construct knowledge graphs from domain concepts, textual evidence, and structured relations.

→

Module 4

GraphRAG, Evaluation, and Research Applications

Combine vector retrieval and knowledge graphs, evaluate factuality and usefulness, and apply the system to research tasks.

LLM-04

GraphRAG、评估与研究应用

Combine vector retrieval and knowledge graphs, evaluate grounded generation, and apply GraphRAG to research scenarios.

→