Li Yi

Li Yi is a Professor in the Department of Computer Science at Tsinghua University's School of Information Science and Technology, where they lead cutting-edge research at the intersection of computer vision, 3D graphics, and robotics. Their work focuses on advancing neural rendering, point cloud processing, and embodied AI with applications in human-object interaction and robotic manipulation.

Research interests span Computer Vision, 3D Graphics, Robotics, Point Cloud Processing, Neural Rendering, and Human-Object Interaction. Recent work explores language-grounded spatial reasoning, dexterous manipulation, and 4D dynamic content generation, with publications appearing in top venues like CVPR, ICCV, and NeurIPS. Their research bridges theoretical advances with practical applications in embodied AI systems.

The publication trends reveal a strong focus on neural rendering techniques (particularly NeRF variants), embodied AI for robotic manipulation, and multimodal understanding integrating vision, language, and action. Recent work increasingly incorporates large language models and focuses on generalizable approaches that transfer from simulation to real-world settings.

As an advisor, Professor Li has mentored numerous students including Yunze Liu, Xueyi Liu, Zekun Qi, and Runpei Dong, who frequently appear as first authors on collaborative publications. Their research has been supported by significant grants enabling work on human-robot interaction, 3D scene understanding, and embodied AI systems.

Professor Li leads a research group focused on developing comprehensive frameworks for spatial reasoning, object manipulation, and dynamic scene understanding. The team works on creating benchmarks like TACO for tool-action-object understanding and developing systems like MobileH2R for human-robot handover tasks. Current work emphasizes real-world applicability with a focus on generalizable solutions that work across diverse settings.