Large Language Models

COMP 4901B, Fall 2025
The Hong Kong University of Science and Technology
Junxian He

Course description: this is an undergraduate-level course to cover the basics and latest advances about large language models. This course will start with the basics of language models, and then cover the latest advances in large language models. It will include core techniques on developing LLMs such as DeepSeek-R1, Qwen, LLama, etc. Applications and potential risks of LLMs will also be covered. Throughout this course, you are expected to (1) know the basic principles of language models and how it works; (2) be able to use LLMs and know how to tame it in practice; (3) understand the core development techniques and are able to train language models in relatively simple scenarios.

Course Information

Course Logistics (Grading, Policy, etc.)

Time and Location

Lectures: Wed, Fri 4:30 PM - 5:50 PM at Multi-function Room, LG4, LIB

Course Staff and Office Hours

Instructor: Junxian He. Office Hour: Wed 10 AM - 11 AM at CYT3004
TA1: Yuzhen Huang. Office Hour: Wed 10AM - 11AM at Zoom
TA2: Wei Liu. Office Hour: Tue 2PM - 3PM at Zoom

Contact and Communication

Canvas is the main platform for communication about technical questions of lecture contents and homeworks. Please do not directly email the teaching staff on technical questions which may easily get lost. The teaching staff will try to answer questions on Canvas in a prompt manner. Some rules:

  • All announcements and communications will happen over Canvas
  • Please utilize the Canvas Discussion page to ask questions. There are pre-existed discussion topics initiated by the staff such as lecture questions and each homework. Please post your questions in the respective thread, you are also encouraged to answer others’ questions if you know the answer.
  • While we allow students to optionally post anonymous questions and answers on Canvas, please do not post anything inappropriate as anonymous users – in such cases, we will turn off the anonymous option.
  • If you have non-technical, private matters, please email the teaching staff directly.
  • Please consult the Course Logistics page before asking logistical questions.
  • For longer discussions, please come to office hours.

Prerequisites and Materials

Students are required to have taken machine learning courses and know basics about machine learning and deep learning.

This course does not have a textbook as large language models is a fast-evolving field, but I will recommend blogs or papers to read for the respective lecture.

Lecture Schedule

The lecture schedule below is tentative and subject to change.

SlidesDateTopicReadingsAssignments
Lecture 003/09 WedIntroduction to Large Language Models  
Lecture 105/09 FriMachine Learning Basics  
Lecture 210/09 WedLanguage Model  
Lecture 312/09 FriNeural Networks and Transformers  
Lecture 417/09 WedTransformers  
Lecture 519/09 FriPretraining: Objectives and Data Curation  
Lecture 624/09 WedIn-Context Learning, Prompt Engineering  
Lecture 726/09 FriEvaluation of Large Language Models  
 01/10 WedThe National Day Holiday  
Lecture 803/10 FriInstruction Tuning and Alignment  
Lecture 908/10 WedInstruction Tuning and Alignment  
Lecture 1010/10 FriReinforcement Learning Basics  
Lecture 1115/10 WedReinforcement Learning from Human Feedback (RLHF)  
Lecture 1217/10 FriScaling Up LLMs: Scaling Laws  
Lecture 1322/10 WedChain-of-Thought Reasoning  
Lecture 1424/10 FriDeep Reasoning Models  
 29/10 WedChung Yeung Festival  
Lecture 1531/10 FriSynthetic Data Training: Data Synthesis and Distillation  
Lecture 1605/11 WedSynthetic Data Training: Self-Improving through RLVR  
Lecture 1707/11 FriLLM Agents  
Lecture 1812/11 WedLLM Agents  
Lecture 1914/11 FriLLM Safety: Attacks and Defense  
Lecture 2019/11 WedHallucination and RAG  
Lecture 2121/11 FriBias and Ethics  
Lecture 2226/11 WedSparse Models  
Lecture 2328/11 FriReview Representative LLMs