Data-centric AI plays a crucial role in advancing towards Artificial General Intelligence (AGI). My research focuses on developing novel algorithms and theoretical foundations — such as contrastive learning and generative learning — to more effectively leverage data for enhancing model generalization and robustness.

In addition, my team actively explores multi-modal alignment and robustness in large models, including LLMs and VLMs. Our work addresses critical challenges such as jailbreak attacks, adversarial robustness, and hallucination mitigation.

I am also broadly interested in emerging machine learning topics, including AI for science and 3D modeling, and always enthusiastic about tackling new challenges in these domains.