Interpretable Chinese Fake News Detection with Chain-of-Thought and In-context Learning

The spread of fake news through online platforms poses a significant threat to public trust and social stability, especially in resource-constrained Chinese domains. Existing Fake News Detection (FND) models often lack transparency, functioning as “black boxes" that limit their practical utility. To address this challenge, we propose a novel framework for Chinese Fake News Detection that enhances both accuracy and explainability. Our approach integrates Chain-of-Thought reasoning, retrieval-based in-context learning, and knowledge distillation, allowing for interpretable model predictions. We also introduce an “undetermined” label to handle ambiguous cases and construct the Chinese Explainable Fake News Detection Dataset, which provides detailed annotations for training and evaluation. Experiments show that our model improves both detection accuracy and interpretability, advancing the field of explainable AI in fake news detection.