Are Multimodal Large Language Models Ready for Omnidirectional Spatial Reasoning? | Read Paper on Bytez