About

About me

I study how long-horizon visual information should enter multimodal models, and how generation can help a model understand the world. My current questions are concrete: how to preserve event structure in continuous video, generate inspectable candidate explanations, and return to the original evidence to revise them.

I study at the Institute of Artificial Intelligence and MAC Lab, Xiamen University. I earned my bachelor’s degree in AI at Fuzhou University, where I began with generative vision and facial aesthetics before turning to information redundancy, event structure, and multimodal reasoning in long videos.

I began an internship at AMap, Alibaba Group, in May 2026 and enter the Ph.D. stage in September. My next questions ask how generation can turn an uncertain judgment into inspectable candidates, and how a model can maintain event memory over a continuing video stream while returning to the original evidence to verify them.