57. Robot foundation models and vision-language-action systems

Connect language, vision, and action with multimodal foundation models and vision-language-action policies. Ground instructions in robot state, constrain outputs, evaluate generalization, and guard against unsafe or invented actions.