MenteeBot Executes Tasks via Verbal Instructions (No Teleoperation)

MenteeBot completed a can-retrieval task based on verbal instructions without teleoperation: it understood the request, identified the correct can, navigated to the counter, grasped it, and brought it back to the person. The demo integrates language understanding, scene perception, navigation, and manipulation into a single coherent behavior. Instructions can be given in any language, and the original text states this lowers the barrier to human-robot collaboration and humanoid robot deployment.