A Simulation-Based VQA Dataset for Evaluating Intrinsic Physical Property Inference Capabilities of Vision-Language Models | Synapse