Label source
#2
by
matous-volf
- opened
How were the labels determined? I only found a mention of this dataset in the paper. Thanks!
Labels come from the original datasets. This dataset was compiled from several other entailment datasets that are explicitly about political topics, and combined with other news datasets that have a mix of political and non-political documents. Every document from datasets about political topics gets the political label, and documents from the news dataset get the political label if the original dataset had them in that category.
Labels were all validated by GPT-4o as well.