RumorLLM: A Rumor Large Language Model-Based Fake-News-Detection Data-Augmentation Approach

Abstract

With the rapid development of the Internet and social media, false information, rumors, and misleading content have become pervasive, posing significant threats to public opinion and social stability, and even causing serious societal harm. This paper introduces a novel solution to address the challenges of fake news detection, presenting the “Rumor Large Language Models” (RumorLLM), a large language model finetuned with rumor writing styles and content. The key contributions include the development of RumorLLM and a data-augmentation method for small categories, effectively mitigating the issue of category imbalance in real-world fake-news datasets. Experimental results on the BuzzFeed and PolitiFact datasets demonstrate the superiority of the proposed model over baseline methods, particularly in F1 score and AUC-ROC. The model’s robust performance highlights its effectiveness in handling imbalanced datasets and provides a promising solution to the pressing issue of false-information proliferation.