Stanford and Caltech's HomeBody connects GPT-6 Astra straight to a humanoid robot's skills to tidy an unfamiliar kitchen

Researchers at Stanford and Caltech have built HomeBody, a system that lets a Unitree G1 humanoid robot find its way around an unfamiliar kitchen on its own, tidy up and fetch items from drawers. Most setups put a trained control layer between the language model and the robot. HomeBody drops that layer. A swappable vision-language model, OpenAI's GPT-6 Astra in these tests, calls directly into an extensible library of skills for grasping, navigating and opening drawers. The robot first explores the room and builds a digital twin in Nvidia's Isaac Sim. It records objects and their locations in a spatial memory, so it can find items after they leave its field of view. For open-ended instructions such as cleaning the kitchen, the model plans each step and corrects itself when something goes wrong. The project page shows the robot gathering coffee bags on an island and throwing out specified cartons, which takes repeated trips, grasps and placements. The researchers list several limitations: Astra's latency, finger servos that overheat, and high compute costs. The code is on GitHub. The Decoder notes that this is another test in which Astra works well with robots, while an earlier benchmark raised safety concerns when Astra controls robot hardware.