Offline Policy Evaluation
Simulate photorealistic policy rollouts in GWM-1 from your policy's action sequences and camera observations. Surface failure modes, evaluate edge cases across diverse scenarios and lighting conditions, and iterate on your policies without touching a physical robot.
Test policies without testing hardware.
Catch failures early
Submit action sequences and observations to simulate full rollouts. Surface collisions, missed grasps, and edge cases before deploying to your robot.
Photorealistic simulation
GWM-1 generates visually accurate rollouts that reflect real-world physics, materials, and lighting — giving you high-fidelity signal on policy behavior.
Iterate faster
Evaluate policy changes in minutes instead of hours. Run hundreds of simulated rollouts in the time it takes to do a single hardware test.
Sign up to access the Offline Policy Evaluation SDK and start simulating policy rollouts with GWM-1.
Get Access