SimpleVQA: Multimodal Factuality Evaluation for Multimodal Large Language Models