Abstract: Vision Language Models (VLMs) offer promising solutions for robots to learn on both visual and linguistic data and adapt to dynamic real-world environments. However, the integration of depth ...