Our training dataset consists of approximately 40,000 unique mathematics problem-answer pairs compiled from:
AIME (American Invitational Mathematics Examination) problems (1984-2023)
AMC (American Mathematics Competition) problems (prior to 2023)
Omni-MATH dataset
Still dataset
problem: The mathematical question text, formatted with LaTeX notation.
solution: Offical solution to the problem, including LaTeX formatting… See the full description on the dataset page:
https://huggingface.co/datasets/agentica-org/DeepScaleR-Preview-Dataset.
