chaosc/math1.2k

Math 1.2K Dataset

This dataset is a sampled version of hiyouga/math12k, containing:

• Train set: 1,200 randomly sampled examples from the original 12,000 training examples
• Test set: All 500 test examples (unchanged from original)

Dataset Creation

The original dataset was converted from https://github.com/openai/prm800k.

This sampled version was created using the following script:

from datasets import load_dataset, DatasetDict
from huggingface_hub import HfApi
import random

def main():
    # Load the original dataset
    dataset = load_dataset("hiyouga/math12k")
    
    # Get train and test sets
    train_data = dataset["train"]
    test_data = dataset["test"]
    
    # Sample 1200 examples from train set
    train_indices = list(range(len(train_data)))
    random.seed(42)  # Set seed for reproducibility
    sampled_indices = random.sample(train_indices, 1200)
    sampled_train = train_data.select(sampled_indices)
    
    # Create new dataset dict
    new_dataset = DatasetDict({
        "train": sampled_train,
        "test": test_data
    })
    
    # Push to huggingface hub
    api = HfApi()
    username = api.whoami()['name']
    new_dataset.push_to_hub(f"{username}/math1.2k")

if __name__ == "__main__":
    main()

Usage

from datasets import load_dataset

dataset = load_dataset("chaosc/math1.2k")
train_data = dataset["train"]  # 1,200 examples
test_data = dataset["test"]    # 500 examples

Dataset Structure

Each example contains:

• problem: A math problem statement (string)
• answer: The answer to the problem (string)

Citation

If you use this dataset, please cite the original PRM800K dataset:

@article{lightman2023let,
  title={Let's Verify Step by Step},
  author={Lightman, Hunter and Kosaraju, Vineet and Burda, Yura and Edwards, Harri and Baker, Bowen and Lee, Teddy and Leike, Jan and Schulman, John and Sutskever, Ilya and Cobbe, Karl},
  journal={arXiv preprint arXiv:2305.20050},
  year={2023}
}