Face4RAG

该数据集是一个全面的事实一致性评估（FCE）基准，它不仅包含基于事实不一致性错误类型的合成数据集，还包含从六个常用的大型语言模型（LLMs）构建的真实世界数据集。该基准测试使得评估不同FCE方法在特定错误类型或真实世界错误分布上的表现成为可能，同时突显了现有FCE方法在检测逻辑谬误方面的不足。这一综合基准涉及多个LLMs，其任务是进行事实一致性评估。

This dataset is a comprehensive Factuality Consistency Evaluation (FCE) benchmark. It includes not only synthetic datasets based on factual inconsistency error types, but also real-world datasets constructed from six widely used Large Language Models (LLMs). This benchmark enables the evaluation of the performance of different FCE methods on specific error types or real-world error distributions, while also highlighting the limitations of existing FCE methods in detecting logical fallacies. This comprehensive benchmark involves multiple LLMs, with the core task of conducting factuality consistency evaluation.