Kaizhi Zheng

Papers

Total Citations

H-Index

About

Kaizhi Zheng is a researcher advancing the frontier of embodied AI, with a focus on bridging vision, language, and robotic manipulation. His work centers on enabling robots to understand and execute complex, language-guided tasks in unstructured environments. Zheng’s most influential contribution is **VLMbench**, a compositional benchmark for vision-and-language manipulation that systematically evaluates an agent’s ability to follow natural language commands for object manipulation—a critical step toward general-purpose household robots. This work, garnering 15 citations, addresses the “last mile” of embodied agents by testing compositional language understanding in physical tasks. Additionally, in his work on **Manipulation-Oriented Object Perception in Clutter**, Zheng introduced Affordance Coordinate Frames, a novel framework that allows robots to generalize manipulation actions—such as pouring or serving—to novel objects by recognizing task-relevant affordances. This research tackles the core challenge of robust operation in cluttered, unstructured environments. Through these contributions, Zheng is shaping how robots perceive and interact with the world, making him a notable voice in the growing field of vision-and-language robotics.