My research interests span:

• Multimodal Foundation Model / World Model: Developing native multimodal foundation models which can perform understanding, reasoning, generation tasks from video, language, speech. These models will serve as the core intelligence—the "brain"—for Embodied AI, Robotics, and many other applications. (My Long-Term research interest and belief)
• Reinforcement Learning: I study reinforcement learning on top of pretrained and SFT-initialized models to move beyond imitation—unlocking new capabilities in generative modeling and robotics, including training agents inside learned world-model environments.
• Data-centric Machine Learning: Beyond the perspectives of models and algorithms, I am also enjoying to analysis and understand training data, improve data quality, compressdata for training efficiency, and build efficient / scalable data pipelines for curating or synthesizing high-quality training data for foundation models.