Real-to-sim-to-real is praised as a key scaling path for physical RL

DrJimFan · x · 2026-07-29

Real2sim2real is framed here as one of the best ways to scale environments for physical RL.

The reply argues that reinforcement learning is fundamentally about environments, and that building better simulated environments is a key lever for scaling physical robot training.

Related event: World Labs Unveils R2S2R Platform to Train Robots via Spatial Intelligence(5 posts)→

Original post →

More from Embodied

Embodied channel →