Instructor: Jindong Wang
Office hours: 11am-12pm, Thursday
Email group: TBA
When: Tuesday & Thursday, 9:30–10:50 a.m. (ET)
Where: TBA
Each class: 55min lecture + 5min buffer + 20min paper presentation. Schedule subject to change.
| Week & Date | Topic/Content | Material/Suggested Reading | Note |
|---|---|---|---|
| I. Foundations of Generative AI | |||
| 1 | |||
| January 28, 2027 | L01: Introduction | Opportunities and Risks of Foundation Models | |
| 2 | |||
| February 2, 2027 | |||
| February 4, 2027 | Tue — L02: VAE | ||
| Thu — L03: GAN | Auto-Encoding Variational Bayes | ||
| Generative Adversarial Nets | HW1 starts | ||
| 3 | |||
| February 9, 2027 | |||
| February 11, 2027 | L04: Diffusion & Flow Matching | DDPM | |
| Flow Matching | |||
| 4 | |||
| February 16, 2027 | |||
| February 18, 2027 | L05: Autoregressive Models & Transformers | Attention Is All You Need | |
| Slides | HW1 due February 21, 2027 | ||
| II. Large Language Models | |||
| 5 | |||
| February 23, 2027 | |||
| February 25, 2027 | L06: LLM Fundamentals | ||
| Prompting, inference, evaluation | PromptBench | ||
| PromptRobust | HW2 starts | ||
| 6 | |||
| March 2, 2027 | |||
| March 4, 2027 | L07: Pre-training | ||
| Data, scaling, privacy | Scaling Laws | ||
| Compute-Optimal LLMs | |||
| 7 | |||
| March 9, 2027 | |||
| March 11, 2027 | L08: Post-training & Fine-tuning | ||
| SFT, DPO, RL/RLHF, LoRA/PEFT | InstructGPT | ||
| Direct Preference Optimization | |||
| LoRA | |||
| March 13, 2027 → March 21, 2027 | Spring Break | No classes | |
| 8 | |||
| March 23, 2027 | |||
| March 25, 2027 | L09: Security and Alignment | ||
| Prompt injection, jailbreaks, safety alignment | Indirect Prompt Injection | ||
| Constitutional AI | |||
| III. AI Agents and Reasoning | |||
| 9 | |||
| March 30, 2027 | |||
| April 1, 2027 | L10: AI Agents | ||
| Planning, multi-agent systems, human oversight | ReAct | ||
| CompeteAI | HW3 starts Tue | ||
| HW2 due Thu | |||
| 10 | |||
| April 6, 2027 | |||
| April 8, 2027 | L11: Reasoning | ||
| Chain-of-thought, test-time compute, evaluation | Chain-of-Thought Prompting | ||
| DeepSeek-R1 | |||
| 11 | |||
| April 13, 2027 | |||
| April 15, 2027 | L12: Tool Calling & Memory | ||
| Tools, RAG, persistent memory, safety | Toolformer | ||
| MemGPT | Project proposal due Thu | ||
| IV. Multimodal and World Models | |||
| 12 | |||
| April 20, 2027 | |||
| April 22, 2027 | L13: General Multimodal AI | ||
| Vision, language, audio/video, evaluation | CLIP | ||
| LLaVA | HW4 starts Tue | ||
| HW3 due Thu | |||
| 13 | |||
| April 27, 2027 | |||
| April 29, 2027 | L14: Unified Multimodal Models | ||
| Joint understanding and generation | BAGEL | ||
| Janus-Pro | |||
| 14 | |||
| May 4, 2027 | |||
| May 6, 2027 | L15: World Models | ||
| Dynamics, prediction, planning; course review | World Models | ||
| DreamerV3 | HW4 due Thu | ||
| May 13, 2027 | Final Project submission | Code + report | Due 12:00 p.m. (noon) |