I am Mingzu Liu (ๅๆ็ฅ), an undergraduate student in Automation Science at Shandong University. I am currently a research assistant supervised by Prof. Runmin Cong in the Multimedia and Vision Processing Group (MVP Lab). I am also a CTF player with SU-Team and HASHTEAM, focusing mainly on web security, vulnerability analysis, and AI security.
โจ Research Interests
- Trustworthy and Robust AI
- Multimodal Large Languages Models
๐ฅ News
- 2026.05: One paper about backdoor defense for MLLMs has been accepted by ICML 2026.
๐ Publications
* indicates co-first author.
ICML 2026
TCAP: Tri-Component Attention Profiling for Unsupervised Backdoor Detection in MLLM Fine-Tuning
Mingzu Liu*, Hao Fang*, Runmin Cong
-
Accepted by the Forty-Third International Conference on Machine Learning (ICML), 2026.
-
Efficient and unsupervised backdoor detection in MLLMs via attention allocation profiling.
๐ Education
- 2023.09 - 2027.06 (Expected): B.Eng. in Automation Science, School of Control Science and Engineering, Shandong University
๐ป Internships
- 2025.09 - now, Multimedia and Vision Processing Group (MVP Lab), Undergraduate Research Assistant, China.
โ๏ธ Contact
- Email: mingzuliu@mail.sdu.edu.cn