CASTER: Zero-Shot Transfer of Synthetic Video to Robot Manipulation

Researchers from Virginia Tech and Johns Hopkins released CASTER, which formalizes zero-shot transfer of synthetic videos from generative models to robot manipulation as a constrained optimization problem, requiring no robot interaction data.

2026-10-20 ~ 2026-10-20 · 2 related posts

• What makes human video generation good for robot manipulation? A constrained-optimization analysis — mangahomanga · 2026-10-20
• CASTER formalizes zero-shot robot manipulation from synthetic videos as constrained optimization — mangahomanga · 2026-10-20