GVLA uses gripper-aware representations and adapter routing to adapt vision-language-action policies across different gripper types, supported by the MiGA multi-gripper dataset.
Jun 24, 2026