Hi there 👋

• 🔭 My research interests focus on discovering the underlying potential of modern models that can empower an increasingly broad range of downstream tasks, spanning perception, reconstruction, generation. I believe that pretrained foundation models possess remarkable intelligence, yet they often lack the specialized capabilities required for real-world downstream applications. This motivates my work toward bridging this gap directly. Recently, I have been deeply inspired by the impressive progress of generative video models. These models appear to encode rich world knowledge, such as physical dynamics and multi-image interactions, capturing what I would describe as the abstract relationships that govern the world. I believe such models have the potential to evolve into \emph{world models} capable of simulating complex real-world behaviors and powering a wide variety of intelligent tasks. I am actively exploring this direction and welcome discussions or collaboration ideas.
• 📫 How to reach me: haosen.yang.6@gmail.com