three-system-architecture

GigaBrain-0.7 is an embodied foundation model that uses a three-system architecture and large-scale heterogeneous pretraining to improve generalization across diverse robot embodiments, achieving enhanced zero-shot capabilities and task success rates.