himel05/fake-news-roberta

This model is a RoBERTa-base transformer fine-tuned on the WELFake Dataset for binary text classification (Real/Fake news).

It was trained using class weight balancing and a weighted loss function to ensure robust and balanced performance.

Important Note on Labels

This specific model uses reversed labels:

Label 0 corresponds to REAL news.

Label 1 corresponds to FAKE news.