Answer extraction and evaluation protocol for MATH dataset

[soumyapal91]

Hello, thanks for this great work! Are the code and prompts the same that is used to generate the results on the MATH dataset? What is the evaluation protocol that is used in the paper? I ran this code on some questions from the MATH dataset and noticed quite a few bad answer extractions. I am wondering if you could provide the exact code for the answer extraction and accuracy calculation on the MATH dataset. Thanks in advance!