Where can I find the COMIEVAL dataset mentioned in the paper

[zwangsyc]

Can you also share the COMIEVAL dataset? I am looking into the topic data contamination as well, I found this project is very interesting and inspiring. But I notice that I can not find the COMIEVAL dataset in this repo. Can you please tell me how to get access to this dataset:)

[YihongDong]

Thanks for your attention to our work! You can find them in Google Drive, the links are as follows:
https://drive.google.com/file/d/1FTakU4HIXz00rg8GQQjM7jRw2hBWMsHu/view
https://drive.google.com/file/d/1uBsnVoVCqT8VXA-4REClR6FS8AOab2X9/view

[zwangsyc]

Hi, Yihong, thanks for your sharing. I am running the TED experiment, the main function mentions the

def evaluate_pass_at_k(Occurrence=0, path = '', duplicates = True, top_percent_exclusion = None, dataset_name = 'humaneval'):

if dataset_name == 'humaneval':
    humaneval_data = load_from_disk('datasets/humaneval')['test']#load_dataset("openai_humaneval")['test']

INPUT_PATH = os.path.join(path, f'Occurrence_{Occurrence}.jsonl')
assert os.path.exists(INPUT_PATH)

Might I know where the 'Occurrence_{Occurrence}.jsonl' file is?

[YihongDong]

They are included in Google Drive as well.