Catatan episode
A novel 3D-aware VLA pre-training approach that aligns visual embeddings with 3D perception, trained on 1.7M episodes and 14k+ hours of robot data, achieving top performance on the RoboDojo real-world leaderboard. The spatial-language-action alignment represents a meaningful advance in grounding robot policies in 3D scene understanding.
Situs: share.transistor.fm

