X-Fact

X-FACT 是一个大型的公开可用的多语言数据集，用于对自然存在的现实世界主张进行事实验证。该数据集包含 25 种语言的简短陈述，并由专家事实核查人员标记为真实性。该数据集包括一个多语言评估基准，用于测量多语言模型的域外泛化和零样本能力。

X-FACT is a large, publicly available multilingual dataset designed for fact verification of naturally occurring real-world claims. It contains short statements across 25 languages, annotated with veracity labels by expert fact-checkers. The dataset also includes a multilingual evaluation benchmark for measuring the out-of-domain generalization and zero-shot capabilities of multilingual models.

搜集汇总

背景与挑战

背景概述

X-FACT是一个大型多语言事实验证数据集，包含25种语言的专家标记陈述，用于评估多语言模型的泛化能力。该数据集由犹他大学于2021年发布，支持对现实世界主张的真实性验证。