Hello 🎊
I am Jinxi He (何锦熙), a second-year master’s student in Robotics at Carnegie Mellon University
, advised by Prof. Katia Sycara.
My research focuses on video understanding, multimodal reasoning, vision-language models, visual memory, and contextual safety.
I want to build video models that:
- Retain meaningful context
- Efficiently build memories that preserve relevant information over time and connect past observations to new experiences.
- Reason beyond observation
- Infer relationships, dynamics, and intent from video to understand not only what happens, but how and why.
- Guide real-world decisions and actions
- Translate this understanding into informed judgments and purposeful actions in real-world situations.
News 🐝
New preprint StreamScout: Learning When to Look Deeper for Streaming Video Understanding is now available on arXiv! Paper