Code bug in text-only GPT-4 evaluation

[sudanl]

Hello, Thanks for the interesting work and benchmark.

I found a variable new_json_dataset in the following code that was defined but nothing was added (should be appended the data), causing a bug in for i in range(0, len(new_json_dataset), 10):

https://github.com/OpenBMB/OlympiadBench/blob/5e794b52c394e5132f288fb8d1e5e5ced022cb2c/inference/code/evaluators/text_only_gpt_4.py#L131C1-L131C24