LMMs-Eval: Reality Check on the Evaluation of Large Multimodal Models