#Denver #XPENG #Physical AI #CVPR

XPENG's Key Contributions at CVPR 2026

XPENG, a leading technology enterprise based in China, is making waves at the prestigious CVPR 2026 (The IEEE/CVF Conference on Computer Vision and Pattern Recognition) conference. This marks the company's third appearance at this notable event, reinforcing its commitment to pushing the boundaries of physical AI technology. Dr. Xianming Liu, the head of the General Intelligence Center at XPENG, delivered an insightful presentation during the inaugural Workshop on Deployment of Foundation Models for Embodied AI (WDFM-EAI), joining forces with global tech giants like Tesla, NVIDIA, and Waymo.

Breakthrough Achievements in Physical AI

During his speech, Dr. Liu outlined the evolution of XPENG's physical AI technology, detailing the journey from initial concept validation and technical refinements to full-scale production. This year, the company achieved a significant milestone with the launch of VLA2.0, its latest foundation model, which has now entered formal mass production. Originally showcased at CVPR 2025, VLA2.0 has already set an industry record, accounting for over 50% of assisted driving mileage in its first month alone — a remarkable achievement that sets new standards in China’s assisted driving sector.

Advancing the Concept of World Models

One of the standout elements of Dr. Liu's presentation was the introduction of the world model, a pivotal component of XPENG's foundation model system. This new model aims to facilitate Deliberative Reasoning, Controllable Generation, and Long-Horizon Forecasting. Rather than viewing the world model and VLA2.0 as competitors, Dr. Liu emphasized their complementary roles — where VLA2.0 teaches