X-Square Robot Unveils WALL-WM, the World's First Event-Level Prediction Embodied AI World Model
X-Square Robot, the Chinese embodied AI company behind the GreatWall series of robotic foundation models, has introduced WALL-WM — the world's first event-level prediction world model for embodied intelligence. The breakthrough shifts the prediction unit of world models from fixed time frames to semantic events, fundamentally changing how robots understand and execute physical tasks. Conventional vision-language-action models operate by predicting fixed-length action chunks f
Chinese AI company X-Square Robot has launched WALL-WM, a new embodied AI world model that predicts actions based on events rather than fixed time intervals. This innovative approach allows robots to better understand and execute tasks by focusing on the desired outcome, such as grasping an object, instead of calculating movements frame by frame. By shifting from time-based predictions to event-centric ones, WALL-WM demonstrates improved generalization capabilities across different scenarios and objects.
The model utilizes a novel three-layer architecture and a unique training strategy to handle the complexities of integrating text, vision, and action modalities. Initial benchmarks show WALL-WM surpassing existing models in embodied video generation and robot task completion.
This development could lead to more adaptable and intelligent robots capable of performing complex physical tasks in real-world environments with greater reliability.
📌 Kaynak
Bu özet Pandaily kaynağından otomatik derlenmiştir. Tamamı için orijinal habere gidin.
Orijinal haberi oku →