Demetri Terzopoulos

Kuafu's ability to synthesize robust policies from language and task semantics achieves over 82% success in complex manipulation tasks, revolutionizing how we approach control-to-learning paradigms.

RoboEdit transforms abundant human manipulation videos into high-fidelity training data for robots, unlocking new avenues for scalable robot learning.

Counterintuitively, distilling LLMs is more effective when you only use the first few tokens of a student's rollout, surpassing full-trajectory distillation while saving compute.