This project was interesting enough to quickly skim through. Regarding the actual verification, they use this dataset:

https://arxiv.org/pdf/2004.14974

To progress in this space, many more datasets would be likely required. The problem with the above dataset also is that it categorizes things into true/false/maybe. Not many things fit into such a coarse categorization.