Lihan Zha
Lihan Zha is a robotics PhD student at Princeton University, advised by Anirudha Majumdar and Dhruv Shah. He is currently a research scientist intern at Physical Intelligence. He is best known for his work on Language-Action Pre-training (LAP), which enables vision-language-action models to transfer zero-shot to new robot embodiments by representing robot actions as natural language tokens, and PlayWorld, a system for training video-based robot world models through autonomous self-play.
“EgoLAP introduces a method to train robots using large-scale egocentric human video data, bypassing the expensive process of collecting robot-specific demonstrations.”
Source→“reaches 80.1% mean real-world task progress, a 2.3x performance gain over alternative action representations”
Source→AI-extracted from podcast / newsletter / paper summaries. May contain errors.