Physical RSI: Learning to Improve, in the Physical World
Helping robots learn new skills faster—and master increasingly complex skills. Turning every real-world deployment into a shorter adaptation cycle for the next …
Read more ›Research
Achieving 4× Inference Acceleration: Cross-Stack Optimization Practices for Heterogeneous Multimodal Models
We propose a Cross-Stack Inference Optimization Framework for Physical AI models, which achieves 4× inference acceleration through coordinated optimizations …
Read more ›
World Reward Model: The Scaling Path for Real-World RL
The World Reward Model (WRM) we are building transforms a world model’s physical understanding into learnable, transferable reward signals. It unifies the …
Read more ›