Embodied AI and VLA Research
Long-horizon robot behavior with vision-language-action systems.
Research on embodied AI systems that combine computer vision, language-conditioned planning, and robot action generation. Current work includes diffusion-based VLA systems, robustness evaluation, and long-horizon embodied benchmarks.
Highlights:
- Technical Lead for The NorthStar Team, ranked 4th worldwide in the NeurIPS 2025 BEHAVIOR-1K Challenge.
- Research on robustness and safety evaluation for open-world VLA systems.
- Work on world action model and VLA policy comparison across robotics benchmarks.
Project links: BEHAVIOR-1K, winning challenge page, challenge video.