Breakthrough AI Framework Enables Robots to Learn from Single Human Demonstration

Cornell University researchers have developed RHyME, an AI system that allows robots to learn multi-step tasks by watching just one human demonstration—even if human and robot movements differ, Earth.com reported.

Traditional methods required thousands of controlled demonstrations, but RHyME uses a “memory” system to adapt actions creatively, improving task success rates by 50% with only 30 minutes of robot-specific data.

Led by doctoral student Kushal Kedia, the team compared the approach to “translating human tasks into robot language.” The innovation could accelerate robotics in fields like elder care and logistics, moving closer to practical home assistants. Findings will be presented at the IEEE International Conference on Robotics and Automation in Atlanta.