MM-CoT:A Benchmark for Probing Visual Chain-of-Thought Reasoning in Multimodal Models