Skip to main content
LingBot-VA predicts robot actions from camera observations and a language task, conditioning later predictions on the actions already executed. Reactor hosts inference; your application runs the simulator or robot controller. The hosted variant uses LIBERO-Long post-training. The upstream LingBot-VA project also includes other embodiment configurations. Those configurations are not selectable through this endpoint and do not share a hardware action contract automatically. Start with Get your first actions, then run LIBERO or review the robot integration requirements.