Research Interests

I am someone who likes to understand what models actually learn, rather than just whether they work. I am curious about how vision & language and understanding & generation come together inside a model, and why some forms of learning lead to reasoning and generalization while others do not.

My work spans multimodal understanding as well as image and video generation. Instead of treating these as separate problems, I see them as different views of the same question: how learning across modalities shapes internal representations.

Click on a tag to filter publications Clear