metadata
dataset_info:
features:
- name: problem
dtype: string
- name: answer
dtype: int64
- name: messages
list:
- name: content
dtype: string
- name: role
dtype: string
splits:
- name: train
num_bytes: 2622112
num_examples: 3516
- name: test
num_bytes: 27748
num_examples: 40
download_size: 1158566
dataset_size: 2649860
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: test
path: data/test-*
1750 challenging math QA combining historical AIME and AMC12, plus a subset of AMC10 filtered by pass@1 failure from Deekseek-Distill-Qwen-7B COT inference.
Each problem is prompted with both English and Chinese templates to enhance diversity for GRPO training.
Test set consists of open reference problem from AIMO1 and AIMO2 competitions.