allenai/winogrande
WinoGrande is a dataset containing 44,000 problems, inspired by the Winograd Schema Challenge, but scaled up and adjusted to improve robustness against dataset‑specific biases. The task is cloze, providing two options; the goal is to select the correct option for the given sentence, requiring common‑sense reasoning.
Dataset description and usage context
Dataset Overview
Dataset Name: WinoGrande
Dataset Configurations:
- winogrande_xs
- winogrande_s
- winogrande_m
- winogrande_l
- winogrande_xl
- winogrande_debiased
Dataset Features:
- sentence: string
- option1: string
- option2: string
- answer: string
Dataset Splits:
| Configuration | Train | Validation | Test |
|---|---|---|---|
| winogrande_debiased | 9248 | 1267 | 1767 |
| winogrande_l | 10234 | 1267 | 1767 |
| winogrande_m | 2558 | 1267 | 1767 |
| winogrande_s | 640 | 1267 | 1767 |
| winogrande_xl | 40398 | 1267 | 1767 |
| winogrande_xs | 160 | 1267 | 1767 |
Dataset Size:
- Download size: 3,395,492 bytes
- Dataset size varies by configuration, e.g.:
- winogrande_xs: 412,552 bytes
- winogrande_s: 474,156 bytes
- winogrande_m: 720,849 bytes
- winogrande_l: 1,711,424 bytes
- winogrande_xl: 5,577,680 bytes
- winogrande_debiased: 1,595,268 bytes
Citation:
@InProceedings{ai2:winogrande, title = {WinoGrande: An Adversarial Winograd Schema Challenge at Scale}, author = {Keisuke Sakaguchi and Ronan Le Bras and Chandra Bhagavatula and Yejin Choi}, year = {2019} }
Pair the dataset with AI analysis and content workflows.
Once the source passes your review, move straight into summarization, transformation, report drafting, or presentation generation with the JuheAI toolchain.