[!IMPORTANT] Why this dataset is duplicated: This dataset actually repeats the AIME 2024 dataset for 32 times to help calculate metrics like Best-of-32.
How we are trying to fix: verl is supporting specifying sampling times for validation and we will fix it asap.
4 commits
[!IMPORTANT] Why this dataset is duplicated: This dataset actually repeats the AIME 2024 dataset for 32 times to help calculate metrics like Best-of-32.
How we are trying to fix: verl is supporting specifying sampling times for validation and we will fix it asap.
4 commits