VitaminC

该数据集名为VitaminC，是一个大规模的数据集，其中包含了对比性的例子：证据对在语言和内容上几乎完全相同，唯一的区别在于一个证据支持某个论点，而另一个证据则反驳该论点。有研究表明，在VitaminC数据集上训练的事实核查模型能够提高分类器对证据中微妙变化的敏感度。任务的目的是进行事实核查。This dataset, named VitaminC, is a large-scale dataset containing contrastive examples. Each evidence pair is nearly identical in both language and content, with the sole difference that one piece of evidence supports a given claim while the other refutes it. Studies have shown that fact-checking models trained on the VitaminC dataset can enhance classifiers' sensitivity to subtle changes in evidence. The objective of this task is fact-checking.

搜集汇总

背景与挑战

背景概述

VitaminC是一个大规模的事实核查数据集，其核心特点是包含语言和内容几乎完全相同的对比性证据对，一个支持论点、另一个反驳论点，旨在训练模型对证据微妙变化敏感。该数据集用于提高事实核查模型的分类性能。