•
Citation:
Shouxin Liu, Chongyong Zhang, Chenhoo An, In-Kwon Lee, and Xiaowei Li, “Fake news detection with external entity expanding and multi-modal dynamic fusion,” Information Sciences, 718, November, 2025.
•
Abstract:
Current fake news detection methods rarely consider the impact of news-related external entities on fake news identification. Moreover, the feature fusion process generally follows the inherent hand-designed fusion module, which makes it difficult to dynamically adjust the fusion process adaptively. Therefore, we propose a new fake news detection model that constructs an external entity extension network to comprehensively improve the richness of news-related external text entities and use these extended entities to deeply enhance text information. We design a bidirectional multi-scale fine-grained interaction mechanism to ensure deep collaboration between entities and image patches and capture rich local correlation information. Simultaneously, we introduce a multi-stream attention interaction mechanism between the enhanced text and image to further learn global coarse-grained correlation. Additionally, we utilize a dynamic routing fusion mechanism to dynamic fuse the fine-grained features and coarse-grained features, realize dynamic adjustment of the fusion process, and improve the robustness of the model. Compared with the state-of-the-art baseline models, we improve the detection accuracy by 2.8% and 8.5% on the mainstream public real-world datasets Weibo and Twitter, respectively.