MM-IQ: Benchmarking Human-Like Abstraction and Reasoning in Multimodal Models Paper โข 2502.00698 โข Published Feb 2 โข 24 โข 2