chrisji / fewglue

This repository contains the FewGLUE dataset for few-shot natural language understanding.

FewGLUE

This repository contains the FewGLUE dataset, consisting of a random selection of 32 training examples from the SuperGLUE training sets and up to 20,000 unlabeled examples for each SuperGLUE task.

🗂️ Structure

For each task t in SuperGLUE, the directory FewGLUE/t contains two files: train.jsonl, which contains the 32 training examples, and unlabeled.jsonl, which contains all unlabeled examples. The official development and test sets are not included as they can be found here.

📑 Format

All files follow the exact same format as the original SuperGLUE training files.

📕 Citation

If you make use of FewGLUE, please cite the following paper:

@article{schick2020small,
  title={It's Not Just Size That Matters: Small Language Models Are Also Few-Shot Learners},
  author={Timo Schick and Hinrich Schütze},
  journal={Computing Research Repository},
  volume={arXiv:2009.07118},
  url={http://arxiv.org/abs/2009.07118},
  year={2020}
}