Data-centric AI opens the door to Artificial General Intelligence (AGI). However, data biases inevitably exist in large-scale data, like data imbalance, spurious correlations among classes, and annotation noise. I’m interested in fundamental research of model generalization and robustness on different data distributions, realizing effective usage of large-scale data, and ensuring that the trained models do not suffer from data biases. It usually involves problems in machine learning and computer vision, like distribution shifts, adversarial robustness, fairness, and out-of-distribution generalization.

Additionally, I aim to understand the behaviors of large models, such as Stable Diffusion models, LLMs, and VLMs, and their applications in downstream tasks.

I am excited to explore new machine learning problems, including AI for science and 3D modeling challenges.