WueNLP/mHallucination_Detection
收藏Hugging Face2025-02-20 更新2025-04-12 收录
下载链接：
https://hf-mirror.com/datasets/WueNLP/mHallucination_Detection
下载链接
链接失效反馈官方服务：
资源简介：
多语言虚构检测数据集（mFAVA）是一个用于虚构检测任务的数据集，它包含了来自不同语言（包括阿拉伯语、德语、中文、土耳其语和俄语）的150个提示，这些提示来源于FAVA数据集和Google自然问题测试集。数据集分为黄金注释和银色（合成）注释，可用于评估大型语言模型在不同语言环境下的虚构生成能力。
The Multilingual Hallucination Detection Dataset (mFAVA) is a dataset for the hallucination detection task, which includes 150 prompts from various languages (including Arabic, German, Chinese, Turkish, and Russian). These prompts are sourced from the FAVA dataset and the Google Natural Questions test set. The dataset is divided into gold annotations and silver (synthetic) annotations, which can be used to evaluate the hallucination generation capabilities of large language models in different linguistic environments.
提供机构：
WueNLP
