Instructor: Edgar Dobriban
This course explores Large Language Models (LLMs), from the basics to cutting-edge research.
- Course Syllabus.
- Lecture Notes; Work in progress.
| Link | Topic |
|---|---|
| 01 | Motivation and Context |
| 02 | AI: Goals and definitions. The role of LLMs. |
| 03 | LLM architectures: attention and transformers. |
| 04 | Insight into transformer architectures. |
| 05 | Position encoding. |
| 06 | Specific LLM families: GPT, Llama, DeepSeek, LLM360. |
| 07 | Training LLMs: pre- and post-training, supervised fine-tuning, learning from preferences (PPO, DPO, GRPO). |
| 08 | Test-time computation: sampling, prompting, reasoning. |
| 09 | Empirical Behaviors: scaling laws, emergence, memorization, super-phenomena. |
- Calibrated Language Models Must Hallucinate by Georgy Noarov.
- Representations in Deep Neural Networks by Joseph H. Rudoler.
- First-Person Fairness in Chatbots by Varun Gupta.
- LLMs Know More Than They Show: On the Intrinsic Representation of LLM Hallucinations by Ryan Chan.
- Representational Abilities of Transformers by Soham Mallick and Manit Paul.
- AI Control: Protocols and methods for deploying untrusted AI models by Davis Brown.
- Diffusion LLMs by Zhihan Huang and Kevin Jiang.
- Various forms of preference optimization by Tao Wang and Sunay Joshi.
- Adversarial Reasoning in LLMs by Mahdi Sabbaghi.
- Transformer Circuits: Mathematical Framework and In-context Learning by Hwai-Liang Tung & Yu Huang.
- Model Collapse by Xuyang Chen & Xianglong Hou.
- Foundations of Large Language Models, U of Michigan, 2024
- Language Modeling from Scratch, Stanford, Spring 2024
- Recent Advances on Foundation Models, U of Waterloo, Winter 2024
- Large Models, U of Toronto, Winter 2025
- Advanced NLP, CMU, Spring 2025
- Andrej Karpathy's Neural Networks: Zero to Hero video lectures. 100% coding-based, hands-on tutorial on implementing basic autodiff, neural nets, language models, and GPT-2 mini (124M params).
- The Llama 3 Herd of Models describes the Llama "open-weights LLM" developed by Meta. Possibly the highest information content anywhere about LLMs.
- DeepSeek-V3 Technical Report and DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning describe the open-weights DeepSeek V3 and R1 models, which bring together several innovations in training LLMs to make them achieve comparable performance to some top closed models.
- The corresponding sections in the Understanding Deep Learning book. See also the associated tutorial posts: LLMs; Transformers 1, 2, 3; Training and fine-tuning; Inference
- Foundations of Large Language Models book