SafeVLA: Towards Safety Alignment of Vision-Language-Action Model via Constrained Learning

Open in new window