Large Language Models with visual capabilities can use symbolic reasoning by interpreting images and forming a mental map of the environment, similar to how humans process spatial information.