Sy-Tuyen Ho (Tuyen)

I am a PhD student at University of Maryland, College Park working with Prof. Furong Huang. Previously, I eared a Master's degree from Singapore University of Technology and Design advised by Prof. Ngai-Man Cheung; and a Bachelor from the University of Information Technology - Vietnam National University Ho Chi Minh City working with Dr. Tiep Nguyen and Prof. Tam Nguyen.

I’m deeply interested in ML with current interest in Agentic Learning in its application and training data perspectives.

• March, 2026 Up to 61.21%: Training data of VLMs can be reconstructed via our SMI-AW attack. We introduce Paper accepted at CVPR 2026
• March, 2026 Current model inversion evaluation framework is misleading, with up to 90% false positives. Our proposed multimodal evaluation framework mitigates this. Paper accepted at CVPR 2026 Findings
• Sep, 2025 Started my PhD in CS at University of Maryland, College Park
• Sep, 2024 Embedding dimension is the key factor in optimizing OoD generalization of ViT. Our new ViT design improves both efficience and OoD generalization. Paper accepted at NeurIPS 2024
• Jul, 2024 Alert! Skip connection reinforces adversary to inverse training data of ML models. Our new architecture desin mitigates this. Paper accepted at ECCV 2024
• Mar, 2024 Transfer learning prevents data leakage. Our simple TL-DMI defense outperforms SOTA methods. Paper accepted at CVPR 2024
• Jan, 2024 Started my Master of Engineering at SUTD

• University of Maryland, College Park (UMD)
  PhD in Computer Science
  

• Singapore University of Technology and Design (SUTD)
  M.Eng (Research)
  

• University of Information and Technology (UIT), Vietnam Nation University HCMC (VNUHCM)
  B.Sc in Computer Science (Honour)
  

[6] Memory Makes The Poison: Over Memorization Drives Visual Poisoning in LVLMs

Sy-Tuyen Ho, Yaseema Rusiru Ariyarathna Epa, Yaseema_Rusiru_Ariyarathna_Epa, Yasoda Lasiru Ariyarathna Epa, Kecheng Liu, Xudong Jiang, Alex Kot, Furong Huang, Ngai-Man Cheung.
[project page]

[5] Model Inversion Attacks on Vision-Language Models: Do They Leak What They Learn?

Ngoc-Bao Nguyen, Sy-Tuyen Ho, Koh Jun Hao, Ngai-Man Cheung.
CVPR'26 [paper] [project page] [code]

[4] Revisiting Model Inversion Evaluation: From Misleading Standards to Reliable Privacy Assessment.

Sy-Tuyen Ho, Koh Jun Hao, Ngoc-Bao Nguyen, Alexander Binder, Ngai-Man Cheung.
CVPR'26 Findings [paper] [benchmark+code] [project page]

[3] Vision Transformer Neural Architecture Search for Out-of-Distribution Generalization: Benchmark and Insights.

Sy-Tuyen Ho*, Tuan Van Vo*, Somayeh Ebrahimkhani*, Ngai-Man Cheung. (* joint first authors)
NeurIPS'24 (Main Track) [paper] [benchmark+code] [project page]

[2] On the Vulnerability of Skip Connections to Model Inversion Attacks.

Koh Jun Hao*, Sy-Tuyen Ho*, Ngoc-Bao Nguyen, Ngai-man Cheung. (* joint first authors)
ECCV'24 [paper] [code] [project page]

[1] Model Inversion Robustness: Can Transfer Learning Help?

Sy-Tuyen Ho, Koh Jun Hao, Keshigeyan Chandrasegaran, Ngoc-Bao Nguyen, Ngai-man Cheung.

CVPR'24 [paper] [code] [project page]