# D_ASAP2

This is the **train, test, and validation split** of the ASAP 2.0 dataset, prepared for use with the 
[S-GRADES benchmark](https://sgrades.eng.unt.edu). Ground truth labels have 
been removed to prevent leakage during evaluation.

## Original Dataset

🔗 [ASAP 2.0 on Kaggle](https://www.kaggle.com/datasets/lburleigh/asap-2-0)

## Citation

If you use this dataset, please cite the original:
```bibtex
@article{crossley2025asap2,
  title={A large-scale corpus for assessing source-based writing quality: ASAP 2.0},
  author={Crossley, Scott A. and Baffour, Perpetual and Burleigh, L. and King, Jules},
  journal={Assessing Writing},
  volume={65},
  pages={100954},
  year={2025},
  publisher={Elsevier},
  doi={10.1016/j.asw.2025.100954}
}
```

If used as part of S-GRADES, also cite:
```bibtex
@inproceedings{seuti2026sgrades,
  title={S-GRADES: Studying Generalization of Student Response Assessments in Diverse Evaluative Settings},
  author={Seuti, Tasfia and Ray Choudhury, Sagnik},
  booktitle={Proceedings of the 15th International Conference on Language Resources and Evaluation (LREC 2026)},
  year={2026}
}
```