Leveraging LLM-Generated Explanations for Detecting Emotionally Rewritten Fake News

2026-09-29

AI Summary

This study addresses the degradation in robustness of fake news detection models caused by sentiment rewriting. To tackle this issue, we propose a Gated Cross-Attention (GCA) framework that leverages large language models to generate stable explanations and adaptively fuses sentiment-rewritten texts with background knowledge through a gating mechanism. This approach effectively mitigates explanation mismatches induced by sentiment reconstruction, enabling the model to focus on critical discriminative features. Experimental results demonstrate that GCA significantly improves detection performance under diverse sentiment conditions on the PolitiFact and LUN datasets. The associated code and data have been made publicly available.

Abstract

The spread of fake news may cause severe social consequences. Existing fake news detection methods mainly focus on stylistic variations or incorporate external information such as explanations. However, news articles are often rewritten under different emotional backgrounds while preserving their underlying factual claims, which may affect the robustness of detection models. In this work, we investigate fake news detec- tion under fact-preserving emotional variations. To study this problem, we construct emotion-rewritten test sets and generate explanations from the original news articles as stable background knowledge. We then propose a Gated Cross Attention (GCA) framework that adaptively integrates emotionally rewritten news with the corresponding explanations, enabling the model to focus on informative explanation content while reducing potential mismatches caused by emotional reframing. Experiments on PolitiFact, GossipCop, and LUN demonstrate that the proposed method achieves notable improvements under multiple emotional conditions on PolitiFact and LUN, while maintaining competitive performance on GossipCop. We further analyze the effects of explanation guidance and gating mechanisms under different emotional conditions. Our code and data are available at: https://github.com/Flulike/fakenews gca .

Problem

• fake news detection
• emotional rewriting
• robustness
• fact-preserving variations

Innovation

• Fake News Detection
• Gated Cross Attention
• LLM-Generated Explanations
• Emotional Rewriting
• Robustness