Can Large Language Model Agents Simulate Human Trust Behaviors?