The paper "Heterogeneous Subgraph Transformer for Fake News Detection" addresses the challenge of detecting fake news by leveraging both structural and textual information. The authors construct a heterogeneous graph that connects news articles, entities, and topics to capture the complex relationships among them. They propose a novel model called Heterogeneous SubGraph Transformer (HETERO-SGT) to exploit these subgraphs for fake news detection. HETERO-SGT integrates a pre-trained dual-attention module to extract word-level and sentence-level semantics from news articles. Random walks with restart (RWR) are used to extract subgraphs centered on each news article, which are then fed into a subgraph Transformer to quantify the authenticity of the news. Extensive experiments on five real-world datasets demonstrate that HETERO-SGT outperforms five baselines in terms of accuracy, macro-precision, macro-recall, macro-F1, and ROC. The paper also includes case studies and ablation studies to validate the effectiveness of the proposed method and its key components.The paper "Heterogeneous Subgraph Transformer for Fake News Detection" addresses the challenge of detecting fake news by leveraging both structural and textual information. The authors construct a heterogeneous graph that connects news articles, entities, and topics to capture the complex relationships among them. They propose a novel model called Heterogeneous SubGraph Transformer (HETERO-SGT) to exploit these subgraphs for fake news detection. HETERO-SGT integrates a pre-trained dual-attention module to extract word-level and sentence-level semantics from news articles. Random walks with restart (RWR) are used to extract subgraphs centered on each news article, which are then fed into a subgraph Transformer to quantify the authenticity of the news. Extensive experiments on five real-world datasets demonstrate that HETERO-SGT outperforms five baselines in terms of accuracy, macro-precision, macro-recall, macro-F1, and ROC. The paper also includes case studies and ablation studies to validate the effectiveness of the proposed method and its key components.