This research presents a novel DIJD (Detect, Investigate, Judge, Determine) framework that enables large language models to detect fake news with minimal training examples.

• Leverages LLMs' prior knowledge and in-context learning for enhanced detection capabilities
• Implements a structured, knowledge-guided approach that mimics human fact-checking processes
• Achieves superior performance in extremely low-resource scenarios compared to existing methods
• Addresses critical security concerns around misinformation that threatens social stability

As fake news continues to pose significant security risks on social media platforms, this framework offers a practical solution for rapid detection with minimal training data—particularly valuable when combating emerging misinformation trends.

Detect, Investigate, Judge and Determine: A Knowledge-guided Framework for Few-shot Fake News Detection