My research interests primarily lie in Humanities Intelligence, Model Adaptation, and Kernel Learning.

In Humanities Intelligence, I am interested in developing AI methods for the understanding, restoration, and analysis of cultural heritage, particularly Chinese historical documents and ancient inscriptions. Our recent work spans multimodal document understanding, inscription and text image restoration, and AI-assisted analysis of historical language and phonology.

In Model Adaptation, my research focuses on improving the generalization and adaptability of learning systems under limited supervision and distribution shifts, including few-shot learning, cross-domain learning, open-world recognition, and multimodal retrieval.

I also have a long-standing interest in Kernel Learning, including spectral, indefinite, and hyperbolic kernel methods, and their connections to modern representation learning.