Iok Tong Lei
Iok Tong Lei is a researcher in the Department of Computer Science and Technology at Tsinghua University in Beijing, China. He is best known for his work on XS-VLA, a lightweight vision-language-action model for robotic manipulation that couples coarse-grained spatial distillation with latent flow matching. His research focuses on embodied AI, robotic control, and spatial grounding for efficient edge-deployable models. He has also contributed to publications on agentic control for embodied manipulation and environment-transferable robotic spatial pose memory.
“XS-VLA achieves 90.3% average success with a 0.25B backbone, outperforming Vanilla SmolVLA-0.25B by 7.5 percentage points under the same evaluation protocol. It also surpasses Vanilla SmolVLA-2.25B despite using roughly one-ninth of the parameters”
Source→“XS-VLA achieves 90.3% average success with a 0.25B backbone, outperforming Vanilla SmolVLA-0.25B by 7.5 percentage points under the same evaluation protocol. It also surpasses Vanilla SmolVLA-2.25B despite using roughly one-ninth of the parameters”
Source→“Department of Computer Science and Technology, Tsinghua University... Wuxi Dexteroushands Robotic Technology Co.”
Source→AI-extracted from podcast / newsletter / paper summaries. May contain errors.