Grounded Action Model: 3D Grounding as a Foundation for Robotics
Manipulation policies must know which objects matter and where they are, yet the pretrained backbones that current robot foundation models build on, from language in vision-language-action models (VLAs) to video generation in world-action models (WAMs), do not directly require this metric grounding, leaving it to be learned implicitly from robot demonstrations. We propose Groun
Reading in Plain English mode — toggle below to switch
Manipulation policies must know which objects matter and where they are, yet the pretrained backbones that current robot foundation models build on, from language in vision-language-action models (VLAs) to video generation in world-action models (WAMs), do not directly require this metric grounding, leaving it to be learned implicitly from robot demonstrations. We propose Groun
This work introduces a new paradigm for robot foundation models by integrating 3D grounding, which is crucial for robust manipulation policies. ML engineers specializing in robotics and embodied AI will find this approach essential for developing more capable and reliable robotic systems.