Multimodal robots learn more efficiently

A vision‑language backbone can significantly reduce robot training data requirements, addressing the...