Seminar in Computer Science - Robust and Verifiable AI
Seminar time: MW 06:00 PM - 07:50 PM
Location: Spieth Hall Room 1307
Instructor: Zhouxing Shi
Office hours: TBD, at Winston Chung Hall 356 (within the 351 suite)
Prerequisites: A prior course or experience in introduction to machine learning or similar
Grading
- Paper presentation: 20%
- Paper review and discussions: 20%
- Project: 60%
- Proposal: 10%
- Midterm report: 10%
- Final presentation: 10%
- Final submission: 30%
[Tentatitive] Schedule
Paper presentation schedule: TBD
| Date | Topic | |
|---|---|---|
| 9/29 | Introduction | |
| 10/1 | Adversarial ML basics | |
| 10/6 | Jailbreaking in LLM | |
| 10/8 | Jailbreaking in LLM | |
| 10/13 | Prompt injection in LLM | |
| 10/15 | Defenses for LLM safety | Project proposal due on 10/29 |
| 10/20 | Defenses for LLM safety | |
| 10/22 | Vulnerabilities in text-to-image generation | |
| 10/27 | Safety and security of AI agents | |
| 10/29 | Safety and security of AI agents | |
| 11/3 | NN verification | |
| 11/5 | NN verification | Midterm report due on 11/9 |
| 11/10 | Training verifiable NN | |
| 11/12 | Training verifiable NN | |
| 11/17 | Applications of verified NN | |
| 11/19 | Verification for generative AI | |
| 11/24 | Verification for generative AI | |
| 11/26 | ||
| 12/1 | Final project presentation | |
| 12/3 | No class (NeurIPS 2025) | Project due on 12/12 |