Runway· Gen / AlephResearch paper

Runway tested GWM-Robotics, a world model for robot policy evaluation, across eight manipulation policies and found a 0.95 correlation between simulated and real-world scores. The model outperforms PolaRiS and can generate up to 30-second rollouts. Runway is building GWM-Robotics with partners including NVIDIA and Berkshire Grey.

From the source

Across all eight policies, simulated and real-world scores correlate at 0.95, demonstrating that our world model can serve as a reliable proxy for evaluating robot policies before committing to hardware deployment.

runway.com