XSPARK AI

Sparking Trusted Physical AI

News

Cross-Hand Complementarity Offers a New Solution for Tactile Representation Learning! Xspark AI and Universities Jointly Propose BiView-Touch

Cover art for the BiView-Touch cross-hand tactile representation research

ECCV 2026 Xspark AI, Tsinghua, and HKU MMLab take first place at the DexGraspMotion Challenge

ECCV 2026 DexGraspMotion Challenge championship announcement

MoPA Xspark AI and university partners unveil coordinated mobile manipulation via subsystem-specific perception alignment

MoPA paper cover for coordinated mobile manipulation

Reading Full-Hand Force from Video and Training World Models via Human Hand Interaction! Xspark AI and Universities Jointly Release Two Tactile Research Achievements

Cover art for DexTouch-WM and TouchSight tactile research
More
SparkLab embodied AI laboratory and robot testing environment

SparkLab

SparkLab is a research group whose founding team comes from Tsinghua University and the University of Hong Kong, united by a shared commitment to advancing trustworthy physical intelligence. Guided by a research-driven vision, we pursue foundational work across algorithms, evaluation, hardware, and data, building a full-stack research team where ideas are not only conceived, but rigorously tested, embodied in real systems, and made reliable in the physical world.

Learn More
X-Eval simulation and real-robot evaluation environment

Customized Simulation & Real-Robot Evaluation Platform

X-Eval

Powered by generative physical simulation as its core engine, the platform delivers end-to-end solutions for robotics AI researchers and engineering teams — from data generation and scientific evaluation to scene asset management.

Learn More

X-Data

X-Data delivers multimodal data services spanning real-world environments and virtual simulations, providing high-quality "fuel" for embodied AI model training and bridging the gap between laboratory validation and real-world deployment.

X-Data

X-Ego-Pro

Our platform captures high-quality multimodal human operation data by synchronizing first-person visuals, dual wrist-mounted camera feeds, hand kinematics, and tactile signals with precise spatiotemporal alignment.