Leveraging LLM-Generated Explanations for Detecting Emotionally Rewritten Fake News

查看原文 / View Original 编辑 / Edit

Computer Science > Computation and Language arXiv:2610.08835 (cs) [Submitted on 29 Sep 2026 (v1), last revised 8 Oct 2026 (this version, v2)] Title: Leveraging LLM-Generated Explanations for Detecting Emotionally Rewritten Fake News Authors: Yupei Guo, Jiajun He, Xiaohan Shi, Tomoki Toda, Zekun Yang, Bowen Wang, Yukinobu Taniguchi.

计算机科学 > 计算与语言 arXiv:2610.08835 (cs) [2026年9月29日提交 (v1)，2026年10月8日最后修订 (本版本, v2)] 标题：利用大语言模型生成的解释来检测情感重写后的虚假新闻作者：Yupei Guo, Jiajun He, Xiaohan Shi, Tomoki Toda, Zekun Yang, Bowen Wang, Yukinobu Taniguchi。

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.

摘要：虚假新闻的传播可能会造成严重的社会后果。现有的虚假新闻检测方法主要关注文体差异或结合外部信息（如解释）。然而，新闻文章经常在保持其基本事实主张的同时，在不同的情感背景下被重写，这可能会影响检测模型的鲁棒性。

In this work, we investigate fake news detection 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.

在这项工作中，我们研究了事实保持前提下的情感变体虚假新闻检测。为了研究这一问题，我们构建了情感重写测试集，并从原始新闻文章中生成解释作为稳定的背景知识。随后，我们提出了一种门控交叉注意力（GCA）框架，该框架能够自适应地将情感重写后的新闻与相应的解释相结合，使模型能够专注于信息量大的解释内容，同时减少由情感重构引起的潜在不匹配。

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: this https URL gca.

在 PolitiFact、GossipCop 和 LUN 上的实验表明，所提出的方法在 PolitiFact 和 LUN 的多种情感条件下取得了显著改进，同时在 GossipCop 上保持了具有竞争力的性能。我们进一步分析了在不同情感条件下解释引导和门控机制的影响。我们的代码和数据可在以下网址获取：this https URL gca。