I have encountered with the evaluation datasets on TMMLU.
Specifically, in the 分科測驗化學 and many more topics dataset, I observed that only question 1 includes choices in the query, whereas the ground truth provides corresponding answers.
I believe addressing this discrepancy is crucial to maintaining the integrity and accuracy of the TMMLU evaluation process. Your prompt attention to this matter would be highly appreciated.
I have encountered with the evaluation datasets on TMMLU.
Specifically, in the 分科測驗化學 and many more topics dataset, I observed that only question 1 includes choices in the query, whereas the ground truth provides corresponding answers.
I believe addressing this discrepancy is crucial to maintaining the integrity and accuracy of the TMMLU evaluation process. Your prompt attention to this matter would be highly appreciated.
Here's the topic with the same issues:
How and Why does that cause an issue? Because the query prompt doesn't include the choices.