lightonai/gpqa_diamond_multilingual

GPQA Diamond Multilingual 是基准数据集 GPQA Diamond 的多语言版本，涵盖六种语言：英语、法语、德语、西班牙语、中文和斯瓦希里语。每个样本是生物学、物理学或化学领域的研究生水平多项选择题，由领域专家编写和验证，并翻译成五种目标语言。此版本是修正版，修复了翻译中的错误和伪影。数据集与论文Rethinking the Multilingual Reasoning Gap with Layer Swap一同发布。

GPQA Diamond Multilingual is the multilingual variant of the benchmark dataset GPQA Diamond, covering six languages: English, French, German, Spanish, Chinese, and Swahili. Each sample consists of graduate-level multiple-choice questions in biology, physics, or chemistry, which were authored and validated by domain experts, and translated into the five target languages. This is a corrected version that fixes translation errors and artifacts. The dataset was released alongside the paper *Rethinking the Multilingual Reasoning Gap with Layer Swap*.