README.md

---
language:
- gl
pretty_name: GlobalPIQA_gl
task_categories:
- question-answering
task_ids:
- multiple-choice-qa
tags:
- galician
- commonsense-reasoning
- physical-reasoning
- multiple-choice
- evaluation
- multilingual
- low-resource-nlp
size_categories:
- 1K<n<10K
configs:
- config_name: default
  data_files:
  - split: train
    path: '*.tsv'
license: cc-by-sa-4.0
---

# GlobalPIQA_gl

**Related paper:** [Global PIQA: Evaluating Physical Commonsense Reasoning Across 100+ Languages and Cultures](https://arxiv.org/abs/2510.24081)

## Dataset Summary

GlobalPIQA_gl is the Galician subset of **GlobalPIQA**, a multilingual benchmark for evaluating physical commonsense reasoning across more than 100 languages and cultural contexts. It is intended as an evaluation resource for models that must choose the most plausible solution to a practical physical situation.

The dataset follows the same general formulation as PIQA-style multiple-choice reasoning tasks, adapted here to Galician as part of the broader GlobalPIQA benchmark. The associated preprint presents GlobalPIQA as a large-scale multilingual benchmark for evaluating physical commonsense reasoning across languages and cultures.

## Dataset Description

GlobalPIQA_gl contains Galician multiple-choice examples focused on **physical commonsense reasoning**. Each item presents a short goal or situation and candidate solutions, with one correct answer.

The dataset is intended primarily for **evaluation**, especially in multilingual settings and for Galician language models.

## Dataset Structure

The dataset is distributed in **TSV format**.

Each row contains the following columns:

- `goal`: the question, goal, or practical situation to solve
- `solution0`: first candidate solution
- `solution1`: second candidate solution
- `label`: index of the correct solution
- `topic`: thematic category of the example

### Example

| goal | solution0 | solution1 | label | topic |
|------|-----------|-----------|------:|-------|
| Como se prepara o sangue do porco para facer filloas? | Cando o porco comece a sangrar, débese desbotar o sangue inicial e recoller o resto nun recipiente. Débese comezar a bater inmediatamente, mentres estea quente, e sempre na mesma dirección para que non calle. É importante bater o sangue ata que arrefríe, podendo engadir un pouco de auga fría para acelerar o proceso, pero non demasiada, para non diluír o produto en exceso. Para que o sangue se conserve durante máis tempo, é recomendable engadir sal gordo antes de que arrefríe de todo. Por último, cómpre coar o líquido para asegurarse de que non quedan grumos. | Cando o porco comece a sangrar, débese desbotar o sangue inicial e recoller o resto nun recipiente. Débese comezar a bater inmediatamente, mentres estea quente, e sempre na mesma dirección para que non calle. É importante bater o sangue ata que arrefríe, podendo engadir un pouco de auga fría para acelerar o proceso, pero non demasiada, para non diluír o produto en exceso. Para que o sangue se conserve durante máis tempo, é recomendable engadir azucre antes de que arrefríe de todo. Por último, cómpre coar o líquido para asegurarse de que non quedan grumos. | 0 | cooking |


## Supported Tasks and Leaderboards

This dataset is suitable for:

- multiple-choice question answering
- physical commonsense reasoning
- evaluation of multilingual language models
- low-resource reasoning benchmarks for Galician

## Languages

The dataset is in **Galician**.

## Intended Uses

GlobalPIQA_gl can be used for:

- evaluation of physical commonsense reasoning in Galician
- benchmarking multilingual transfer to Galician
- comparison of multilingual and language-specific models
- low-resource evaluation of multiple-choice reasoning systems

## Limitations

- The dataset is intended primarily for evaluation rather than large-scale training.
- As a translated or adapted multilingual benchmark subset, some items may reflect translation choices in addition to reasoning difficulty.
- Performance on the benchmark may depend on both commonsense reasoning and sensitivity to the phrasing of the answer options.
- The Galician subset should be interpreted within the broader multilingual design of GlobalPIQA rather than as a standalone monolingual benchmark.

## Usage

Example with `datasets`:

```python
from datasets import load_dataset

ds = load_dataset("csv", data_files="GlobalPIQA_gl.tsv", delimiter="\t")
print(ds["train"][0])
```

## Citation

If you use this dataset, please cite:

```bibtex
@misc{chang2025globalpiqa,
  title={Global PIQA: Evaluating Physical Commonsense Reasoning Across 100+ Languages and Cultures},
  author={Tyler A. Chang and Catherine Arnett and many others},
  year={2025},
  eprint={2510.24081},
  archivePrefix={arXiv},
  primaryClass={cs.CL},
  url={https://arxiv.org/abs/2510.24081}
}
```

## Acknowledgements

This work is funded by the Ministerio para la Transformación Digital y de la Función Pública - Funded by EU – NextGenerationEU within the framework of the project Desarrollo de Modelos ALIA. (Esta publicación del proyecto Desarrollo de Modelos ALIA está financiada por el Ministerio para la Transformación Digital y de la Función Pública y por el Plan de Recuperación, Transformación y Resiliencia – Financiado por la Unión Europea – NextGenerationEU)