Agent TrustCan Large Language Model Agents Simulate Human Trust Behaviors?CRABCross-platform Agent Benchmark for Multimodal Embodied Language Model Agents