X-Square Robot Unveils WALL-WM, the World's First Event-Level Prediction Embodied AI World Model

X-Square Robot launched WALL-WM, an embodied AI world model that predicts actions by semantic events instead of fixed time steps.

The system is built to better align text, vision, and action, using a three-layer design with DMuon optimization and multi-event packing.

X-Square says WALL-WM outperforms several competing models on video generation and robot manipulation benchmarks.

The event-based approach could improve robustness and generalization in real-world robot tasks as scenes, objects, and timing change.