A classifier trained on around 70k news articles, aimed at distinguishing real news from fake news. It is an EmbeddingBag-based text classifier made with PyTorch. Accuracy is 99.205%
Data used is Gonzalo Álvarez's Fake News and Mohammad Javad Pirhadi's Fake News Detection Dataset (English). I could not find any specific citation to use.
Check HuggingFace for the trained model (3.03GB) and prebuilt vocabulary (3.6 MB).