Truth in the Age of AI: Fine-Tuning LLMs for Reliable Fake News Detection

With the rapid rise of social media platforms, people increasingly rely on them not only for entertainment but also as a primary source of news. However, the authenticity of information shared on these platforms is of paramount importance, as misinformation can have catastrophic societal consequences. In this paper, we focus on detecting fake news with high confidence using fine-tuned large language models (LLMs). Leveraging an extensive open-source news dataset, we fine-tuned both open-source and proprietary LLMs to classify news articles as real or fake. Our experiments demonstrate that models such as RoBERTa (open-source) and GPT (proprietary) achieve significantly improved performance after fine-tuning compared to their original pre-trained versions. The evaluation, conducted using standard metrics including accuracy, precision, recall, and F1-score, highlights the effectiveness of fine-tuning in enhancing model performance. The model training cost is also considered in the comparative performance analysis. The findings of this study provide a foundation for future research on LLM-based misinformation detection and offer valuable insights for advancing the development of reliable AI-driven solutions to combat fake news.