NaturalBench: Evaluating Vision-Language Models on Natural Adversarial Samples Paper โข 2410.14669 โข Published Oct 18, 2024 โข 37