Kaiyu Yang
Papers
1
Total Citations
13
H-Index
1
About
Kaiyu Yang is a researcher at the forefront of spatial reasoning and grounded language understanding in 3D environments. Their work centers on bridging the gap between natural language and physical space, with a particular focus on how machines can interpret spatial relations—such as “laptop on table”—from visual data. Yang’s key contribution, the Rel3D benchmark, introduced a minimally contrastive, large-scale 3D dataset that provides high-quality ground truth for training and evaluating models on spatial relation comprehension. This work, which has garnered 13 citations, addresses a critical limitation in prior datasets that lacked robust 3D annotations, enabling more accurate and generalizable learning for embodied AI and robotics. By emphasizing minimal contrastive examples, Yang’s approach sharpens models’ ability to distinguish subtle spatial distinctions, advancing the field of visual reasoning. Their research is instrumental for applications in human-robot interaction, autonomous navigation, and scene understanding, making Yang a notable figure in the intersection of computer vision, natural language processing, and 3D reasoning.
Research Focus
Key Achievements
Top Papers
- 1