-
Galaxy released the world's first end-to-end body grasping base model, GraspVLA, with pre-training data of one billion frames of "vision-language-action" pairs.
Jan. 10 - Galaxy General yesterday (Jan. 9) announced the release of GraspVLA, the first fully generalized end-to-end embodied grasping base model, in conjunction with the Beijing Wisdom Source Institute of Artificial Intelligence (BAAI) and researchers at Peking University and the University of Hong Kong. (Note: "Embodied Intelligence" refers to the incorporation of Artificial Intelligence (AI) into Note: Embodied Intelligence refers to the integration of artificial intelligence into physical entities such as robots, giving them the ability to sense, learn, and dynamically interact with their environment. According to the introduction, the training of GraspVLA consists of two parts: pre-training and post-training. Among them, the pre-training is entirely based on synthetic big data, and the training data reaches the largest ever...- 9.1k