A Multifaceted Investigation on Enhancing Fake News Detection

Abstract

In the age of digitalization, information overload exacerbates the spread of misinformation, the rapid development of the Internet and social media has reconstructed the information communication pattern, and the proliferation of false information has posed a severe threat to social stability and public safety. Against this background, the demand for fake news detection has increased, and improving the accuracy and efficiency of rumor detection has become an important issue to be solved urgently. This paper systematically reviews the methods of strengthening fake news detection with the help of large language models (LLMs), covering data enhancement, knowledge-driven prompting, BERT-LSTM hybrid models, and hybrid models. Although these methods have improved the detection performance to varying degrees, they also face challenges such as insufficient model interpretability, limited applicability in real scenarios, and difficulty in coping with evolving false information strategies. Future research can focus on combining explainable artificial intelligence, strengthening domain adaptation, and promoting multimodal fusion and cross-lingual expansion to further optimize the effect of fake news detection.