Fully Human-Annotated Benchmark for Multi-Modal LLMs in High-Resolution Real-World Scenarios that are Difficult for Humans