Discrepancy in the number of problems with test cases

[cai-jianfeng]

Hi, thank you for releasing this great dataset!
I've been working with the dataset and noticed a discrepancy regarding the number of problems that contain test cases. According to the paper, approximately 32.5k problems include test cases. However, when I processed the dataset, I found the following:

• Total number of problems: 47,136
• Problems with test cases (my count): 26,955
• Problems with test cases (as stated in the paper): ~32,500

There is a gap of roughly 5.5k between my count and the number reported in the paper. I wanted to confirm whether:

1. I might be using an incorrect method to identify problems with test cases, or
2. There is a specific version/subset of the dataset I should be using, or
3. The definition of "having test cases" differs from what I assumed.

Could you please clarify how the 32.5k figure was calculated? Any guidance on the correct way to filter for problems with test cases would be greatly appreciated.
Thanks in advance!