metadata

license: mit
task_categories:
  - text-generation
tags:
  - code-generation
  - unit-testing
  - reinforcement-learning
  - llm

Co-Evolving LLM Coder and Unit Tester via Reinforcement Learning Datasets

This repository contains datasets used in the CURE framework for co-evolving coding and unit testing capabilities in LLMs. The data is formatted using Stdio input/output.

Data Format:

For tasks like calculating the sum of a list, the input/output is formatted as follows:

input = "5
1 2 3 4 5
"
output = "15"

Note that some datasets were originally in a functional format (e.g., assert sum_function([1, 2, 3, 4, 5]) == 15) and have been converted to the Stdio format for consistency.

Datasets Included:

CodeContests
LiveBench
LiveCodeBench
CodeForces
MBPP
CodeContests_train (training data)

Downloading Datasets:

Use the following commands to download the datasets:

cd data
python download_dataset.py --dataset LiveBench  # Example: Download LiveBench dataset
python download_dataset.py --dataset CodeContests_train # Example: Download training data

Paper | Code

Citation

@article{wang2025cure,
  title={Co-Evolving LLM Coder and Unit Tester via Reinforcement Learning},
  author={Wang, Yinjie and Yang, Ling and Tian, Ye and Shen, Ke and Wang, Mengdi},
  journal={arXiv preprint arXiv:2506.03136},
  year={2025}
}


@misc{penedo2025codeforces,
      title={CodeForces}, 
      author={Guilherme Penedo and Anton Lozhkov and Hynek Kydlíček and Loubna Ben Allal and Edward Beeching and Agustín Piqueres Lajarín and Quentin Gallouédec and Nathan Habib and Lewis Tunstall and Leandro von Werra},
      year={2025},
      publisher = {Hugging Face},
      journal = {Hugging Face repository},
      howpublished = {\url{https://huggingface.co/datasets/open-r1/codeforces}}
}