Large Language Model Based Machine Learning Approach for Fake News Detection: Invited Paper

Fake news detection has continued to be a major concern for everyone as content creators are enormously in-creasing. It becomes challenging for humans, and sometimes experts, to classify a news as either fake or real. Although, machine learning (ML) has paved the way for solving many challenging classification tasks related to computer vision, speech recognition, and email filtering, recent advances have exploited pretrained large language models (LLMs) to solve the fake news detection problem. Despite the capability of LLMs, the problem of generalization still remains evident when LLMs are leveraged for fake news detection. In this paper, we propose an LLM-based multi-domain fake news detection (MDFND) framework that integrates smart data augmentation with robust feature learning to facilitate improvement on accuracy and generalization. We created the MDFND dataset and conducted several experiments to compare our proposed method with other natural language processing (NLP) models and evaluate their performance using appropriate metrics.