Reward AI's OM-1 shows emergent behaviors: arms compensating for each other's mistakes

zipengfu · x · 2026-09-15

Reward AI showcases several emergent behaviors in its OM-1 robot model: each arm learns to compensate for the other's mistakes, and the model knows when to retry, adapt to adversarial perturbations, or stop when the environment changes too drastically. All footage is shown in real time.

Related event: Reward AI Unveils OM-1, a Robot Foundation Model Trained Only on Human Data(12 posts)→

Original post →

More from Embodied

Embodied channel →