Hi! (Xin Chào!), I'm Tuyen. I earned my Bachelor of Computer Science (Honours) from the University of Information Technology (UIT), Vietnam National University Ho Chi Minh City (VNU-HCM). Presently, I work as a Research Assistant while pursuing my master's degree at the Singapore University of Technology and Design (SUTD) being advised by Prof. Ngai-Man Cheung. My research interest is to develop Responsible AI systems.

• Sep, 2024, One paper accepted by NeurIPS-24
• Jul, 2024, One paper accepted by ECCV-24
• Mar, 2024, One paper accepted by CVPR-24
• Jan, 2024, Started my Master of Engineering at SUTD

• Singapore University of Technology and Design (SUTD)
  M.Eng (Research)
  January 2024 - August 2025

• University of Information and Technology (UIT), Vietnam Nation University (VNU) HCMC
  B.Sc in Computer Science (Honour)
  August 2017 - August 2021

• Vision Transformer Neural Architecture Search for Out-of-Distribution Generalization

  Sy-Tuyen Ho, Tuan Van Vo, Somayeh Ebrahimkhani, Ngai-Man Cheung Conference on Neural Information Processing Systems (NeurIPS) 2024

  [paper] [code]
  TL;DR: This work presents OoD-ViT-NAS, the first benchmark for out-of-distribution generalization, revealing that the embedding dimension is the key ViT architectural factor influencing OoD performance.

• On the Vulnerability of Skip Connections to Model Inversion Attacks

  Koh Jun Hao*, Sy-Tuyen Ho*, Ngoc-Bao Nguyen, Ngai-man Cheung European Conference on Computer Vision (ECCV) 2024

  [paper] [code]

  * indicates co-first authors
  TL;DR: We systematically reveal, for the first time, that skip connections reinforce model inversion attacks and propose MI-resilient architectures to improve model inversion robustness based on this discovery.

• Model Inversion Robustness: Can Transfer Learning Help?

  Sy-Tuyen Ho, Koh Jun Hao, Keshigeyan Chandrasegaran, Ngoc-Bao Nguyen, Ngai-man Cheung Conference on Computer Vision and Pattern Recognition (CVPR) 2024

  [paper] [code]
  TL;DR: We offer a fresh approach to achieve SOTA model inversion robustness by leveraging transfer learning without bells and whistles.