Nghia H. Nguyen

I am a PhD student in the Computer and Information Science Department at the University of Pennsylvania, advised by René Vidal. I did by B.S. in Computer Science at the University of Science, Vietnam National University – Ho Chi Minh City, advised by Minh-Triet Tran. Previously, I was in the AI Residency program at FPT Software AI Center, working with Thieu N. Vo and Tan M. Nguyen. In Summer 2020, I was a Research Intern at the Distributed Information Systems Laboratory (LSIR), EPFL, working on Graph Neural Networks for SQL query optimization.

My research focuses on leveraging optimization and algebraic structures (e.g., linear representations in LLMs) for trustworthy machine learning, specifically interpretability and adversarial robustness. I pursue two directions:

• Algebraic structures for interpretability: Given that linear and hierarchical structures exist in the latent space of neural networks, how can we design algorithms that exploit this prior knowledge to recover semantic concepts that are more reliable and faithful to the model?
• Emergence of structures via optimization: Why do these algebraic structures arise? I aim to prove their emergence from training objectives, connecting compositional structure in real-world data to geometry in representation space.