I submmit the csv file by using VLMEvalKit to [MMBench Submission](https://mmbench.opencompass.org.cn/mmbench-submission). But got very bad results compared with the results in the paper for Qwen2.5-VL 7B. The logs says "ChatGPT API is not working, use the exact matching policy. The evaluation successfully finished". How can I get the evaluation results by ChatGPT API .