metadata
dataset_info:
features:
- name: type
dtype: string
- name: question
dtype: string
- name: answer
dtype: string
splits:
- name: train
num_bytes: 1890281
num_examples: 7473
download_size: 1068227
dataset_size: 1890281
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
To reproduce
from datasets import load_dataset, Dataset
from tqdm import tqdm
ds = load_dataset("allenai/RLVR-GSM")
out = []
for item in tqdm(ds["train"]):
text = item['messages'][0]['content']
text = text.split("\n\nQuestion:")[-1].strip()
out.append({
"type": "math",
"question": text,
'answer': item['ground_truth']
})
ds = Dataset.from_list(out)