Quang H Nguyen

Quang_2025.jpg

Hi! I am Quang Nguyen (my full name is Nguyen Hung Quang). I am a research assistant at MAIL Research lab at VinUniversity under the supervision of Prof. Khoa Doan. I took my CS bachelor in VNU Hanoi. My current research focuses on understanding the behavior of generative models and developing robust and trustworthy AI systems.

Recently, I am interested in:

  • the mechanisms of reasoning language models and their influence on the alignment and reliability.
  • the reasoning capability of vision-language models.
  • safety concerns in diffusion models.

Besides research, I enjoy reading, playing tennis, and pingpong.

news

2025-10 I am selected as top reviewer at NeurIPS 2025.
2025-09 Our paper studying knowledge attribution in diffusion models is accepted at NeurIPS 2025.
2025-01 Our paper on improving clean-label backdoor attacks in constrained threat models is accepted at ICLR 2025.
2024-05 Our paper applying randomized defense to NLP models is accepted at ACL Findings 2024.
2024-01 Our paper analyzing randomized defense for black-box attacks is accepted at ICLR 2024.

selected publications

  1. NeurIPS
    Unveiling Concept Attribution in Diffusion Models
    Quang H Nguyen, Hoang Phan, and Khoa D Doan
    In The Thirty-Ninth Annual Conference on Neural Information Processing Systems, 2025
  2. ICLR
    Wicked Oddities: Selectively Poisoning for Effective Clean-Label Backdoor Attacks
    Quang H Nguyen, Nguyen Ngoc-Hieu, The-Anh Ta, and 4 more authors
    In International Conference on Learning Representations, 2025
  3. ACL Findings
    Fooling the Textual Fooler via Randomizing Latent Representations
    Duy C Hoang, Quang H Nguyen, Saurav Manchanda, and 3 more authors
    In Findings of the Association for Computational Linguistics: ACL 2024, 2024
  4. ICLR
    Understanding the Robustness of Randomized Feature Defense Against Query-Based Adversarial Attacks
    Quang H Nguyen, Yingjie Lao, Tung Pham, and 2 more authors
    In The Twelfth International Conference on Learning Representations, 2024