projecte-aina/hhh_alignment_ca

hhh_alignment_ca（Helpful, Honest, & Harmless - a Pragmatic Alignment Evaluation - Catalan）是一个加泰罗尼亚语的问答数据集，旨在评估语言模型在实用性、诚实性、无害性等方面的对齐性。数据集包含61个诚实性、59个实用性、58个无害性和43个其他类别的实例。每个实例包含一个输入提示和两个可能的目标答案。数据集由BSC-CNS的语言技术部门策划，并由ILENIA项目资助。数据集的语言为加泰罗尼亚语，采用CC-BY 4.0许可证。数据集的结构为JSONL格式，每个行包含一个实例标识符、问题、可能的答案及其对应的标签。数据集仅包含原始数据集的测试部分。

hhh_alignment_ca is a question answering dataset in Catalan, professionally translated from the English version of the hhh_alignment dataset. The dataset is designed to evaluate language models on alignment, specifically in the categories of helpfulness, honesty/accuracy, harmlessness, and other. The dataset comprises 61 honesty instances, 59 helpfulness instances, 58 harmlessness instances, and 43 other categorized instances. Each instance contains an input prompt with two possible target answers. The dataset is provided in JSONL format, where each row corresponds to a question with an instance identifier, the question, possible answers, and the corresponding label for the correct answer.