toolevalxm/Fusion-CodeEval-Raw

Fusion: Enhancing Code Understanding through Structured Evaluation

📑 Paper    |    🌐 Project Page    |    💾 Released Resources    |    📦 Repo

We release the raw data for our processed CodeEval dataset, derived from the original dataset maintained by the BigCode team.

The data format for each line in the code_samples_filtered_v3.jsonl is as follows:

{
  "code_snippet": <the original code sample>,
  "language": <programming language identifier>,
  "complexity_score": <computed complexity metric>,
  "test_cases": [
    {
      "input": <test input>,
      "expected_output": <expected result>
    },
    ...
  ],
  "source_repo": <original repository reference>,
  "annotations": <human-verified annotations>
}

Some entries may have empty test_cases due to execution timeout or resource constraints during automated test generation.

*Note: The complexity scoring algorithm is documented in our technical appendix. Future releases will include improved annotation coverage.

Citation

If you use this dataset, please cite:

@article{Kocetkov2022TheStack,
  title={The Stack: 3 TB of permissively licensed source code},
  author={Kocetkov, Denis and Li, Raymond and Ben Allal, Loubna and Li, Jia and Mou,Chenghao and Muñoz Ferrandis, Carlos and Jernite, Yacine and Mitchell, Margaret and Hughes, Sean and Wolf, Thomas and Bahdanau, Dzmitry and von Werra, Leandro and de Vries, Harm},
  journal={Preprint},
  year={2022}
}