Datasets:
metadata
language: el
license: cc-by-4.0
multilinguality: monolingual
size_categories: n<1K
task_categories:
- question-answering
pretty_name: Greek physical commonsense reasoning
dataset_info:
features:
- name: id
dtype: int64
- name: prompt
dtype: string
- name: solution0
dtype: string
- name: solution1
dtype: string
- name: label
dtype: int64
- name: culturally specific
dtype: int64
- name: domain
dtype: string
splits:
- name: default
num_bytes: 91539
num_examples: 208
download_size: 49794
dataset_size: 91539
configs:
- config_name: default
data_files:
- split: default
path: data/default-*
Dataset Card for Greek physical commonsense reasoning
The GPCR (Greek physical commonsense reasoning) resource is a manually-annotated physical commonsense reasoning evaluation dataset for Greek. GPRC includes 208 PIQA-style examples consisting of a prompt with two candidate completions ("solutions"). The "solution" pairs are as similar as possible, only differing in one or two words in most of the cases. Approximately 40% of the examples are culturally specific, i.e. they may not be easily translatable into English, or may require regional and/or cultural commonsense.
Dataset Details
Dataset Description
- Curated by: ILSP/Athena RC
- Language(s) (NLP): el
- License: cc-by-4.0
Citation
If you use this dataset in your research or project, please consider including the following citation:
@misc{gprc_greek,
title={Greek physical commonsense reasoning dataset},
author={Vacalopoulou, Anna and Sofianopoulos, Sokratis and Prokopidis, Prokopis},
url=https://huggingface.co/datasets/ilsp/greek_pcr/,
year={2025}
}