Vision-Language Reward Model Evaluation