I am an M.Phil. student at the School of Data Science, The Chinese University of Hong Kong, Shenzhen, advised by Prof. Li Jiang. Before that, I studied at the College of Computer Science and Electronic Engineering, Hunan University.
My research interests include vision-language-action (VLA) models, AI agents, and 3D computer vision.
Feel free to contact me for discussion or collaboration!
Conference on Neural Information Processing Systems (NeurIPS), 2026, Poster
VersaCamVLA enables pretrained VLA policies to handle varying camera counts and unseen camera poses through a unified scene-token interface. Learned from posed RGB views, its compact scene tokens improve manipulation robustness on RoboTwin, LIBERO, and real robots without explicit 3D reconstruction or novel-view rendering at deployment.
IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Highlight
GeoPredict improves 3D reasoning in VLA manipulation by predicting robot trajectories and future workspace geometry. Its training-only geometric supervision boosts simulated and real-world performance without adding 3D decoding at inference.
Conference on Neural Information Processing Systems (NeurIPS), 2026, Poster
Memory Forcing equips autoregressive video diffusion with spatio-temporal memory, balancing exploration of new Minecraft scenes with consistent reconstruction of revisited regions for more coherent long-horizon generation.
AAAI Conference on Artificial Intelligence (AAAI), Poster
PointSLAM++ combines hierarchical neural Gaussians, progressive pose optimization, and adaptive Gaussian density to improve camera tracking and structural consistency under noisy RGB-D input, producing accurate real-time reconstruction and photorealistic rendering.
College of Computer Science and Electronic Engineering
B.Eng. in Computer Science
Advisor: Prof. Ruihui Li