My primary research focus is on efficient video representations, for both video understanding and generation. I have worked on fine-grained activity localization, long-video reasoning with large multimodal models, and diffusion-based video editing/generation.