Lexing Zhang

Beijing Academy of Artificial Intelligence

Papers

Total Citations

H-Index

About

Lexing Zhang is a rising researcher in embodied AI and 3D scene understanding, whose work bridges the gap between visual reconstruction and physical robot interaction. His most-cited paper, "Part-level Scene Reconstruction Affords Robot Interaction" (2023, 8 citations), tackles a critical limitation in robotic perception: the inability to reconstruct interactive scenes with sufficient fidelity. While prior methods relied on replacing objects with limited CAD model databases—leading to large discrepancies between reconstructed and real-world scenes—Zhang introduced a novel part-level reconstruction framework that preserves geometric and functional details essential for manipulation. This contribution enables robots to perceive environments not just as static geometry, but as spaces composed of actionable components. By focusing on part-level granularity, Zhang’s work directly supports downstream tasks like grasping, pushing, and assembly, making it foundational for next-generation interactive robotics. His research sits at the intersection of computer vision, robotics, and scene understanding, with clear implications for autonomous systems operating in unstructured human environments. As his citation trajectory suggests, Zhang is establishing a reputation for rigorous, application-driven research that redefines how machines interpret and act upon the physical world.