AraLingBench A Human-Annotated Benchmark for Evaluating Arabic Linguistic Capabilities of Large Language Models
Paper โข 2511.14295 โข Published โข 74
Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
๐ Paper: arXiv:2511.14295
๐ป GitHub: hammoudhasan/AraLingBench
AraLingBench is a 150-question Arabic multiple-choice benchmark that tests core linguistic competence of language models across five pillars:
All questions are human-authored and validated, with a single correct answer and a difficulty label: Easy, Medium, or Hard.
Each example has:
label (str) โ linguistic categorycontext (str) โ optional supporting text (may be empty)question (str) โ question in Arabicoptions (List[str]) โ answer choicesanswer (str) โ correct choice (matches one of options)difficulty (str) โ one of Easy, Medium, HardSingle split:
train โ 150 examples (use as an evaluation set)from datasets import load_dataset
ds = load_dataset("hammh0a/AraLingBench")
example = ds["train"][0]
print(example["label"])
print(example["question"])
print(example["options"])
print(example["answer"])
If you use AraLingBench, please cite:
@article{zbib2025aralingbench,
title = {AraLingBench: A Human-Annotated Benchmark for Evaluating Arabic Linguistic Capabilities of Large Language Models},
author = {Mohammad Zbib and Hasan Abed Al Kader Hammoud and Sina Mukalled and Nadine Rizk and Fatima Karnib and Issam Lakkis and Ammar Mohanna and Bernard Ghanem},
journal = {arXiv preprint arXiv:2511.14295},
year = {2025},
url = {https://arxiv.org/abs/2511.14295}
}