RoboRefer: Towards Spatial Referring with Reasoning in Vision-Language Models for Robotics | Read Paper on Bytez