My research interests include machine learning, large language models, pattern recognition, and computer vision.

My research focuses on a novel training and deployment paradigm for large models based on Learngene. Inspired by biological genetics, we propose a framework to extract and encapsulate meta-learning capabilities from foundation models. By modularizing this core attribute into a learngene, this paradigm achieves highly efficient and dynamic matching between specific target tasks and diverse hardware constraints, facilitating the rapid, on-demand deployment of edge-side models.

I am also working on Label Distribution Learning (LDL). Label Distribution Learning is a novel machine learning paradigm. A label distribution covers a certain number of labels, representing the degree to which each label describes the instance. LDL is a general learning framework which includes both single-label and multi-label learning as its special cases. The right is the top-100 word cloud generated from my recent published papers.

For future students（2026级硕士/博士研究生报名）