huggingface / huggingface/datasets
Add QuALITY dataset
- Dominant language
- Python
- Stars
- 22k
- Forks
- 3.4k
- Avg merge
- 5d 7h
- Merged PRs (30d)
- 17
Description
## Adding a Dataset
- **Name:** QuALITY
- **Description:** A challenging question answering with very long contexts (Twitter [thread](https://twitter.com/sleepinyourhat/status/1471225421794529281?s=20))
- **Paper:** No ArXiv link yet, but draft is [here](https://github.com/nyu-mll/quality/blob/main/quality_preprint.pdf)
- **Data:** GitHub repo [here](https://github.com/nyu-mll/quality)
- **Motivation:** This dataset would serve as a nice way to benchmark long-range Transformer models like BigBird, Longformer and their descendants. In particular, it would be very interesting to see how the S4 model fares on this given it's impressive performance on the Long Range Arena
Instructions to add a new dataset can be found [here](https://github.com/huggingface/datasets/blob/master/ADD_NEW_DATASET.md).
Contributor guide
Assessment
This issue has not been assessed yet.