Multimodal Transformer-Based Framework for Fine-Grained Fake News Classification

Samiksha Nager , Himanshu Yadav , Ajit Shriastva

Multimodal Fake News Detection, Multimodal Transformers, Fine-Grained Classification, Vision-Language Models, Misinformation Analysis.

The proliferation of misinformation across social media platforms has necessitated the development of advanced automated detection systems capable of handling multimodal content. While early fake news detection approaches relied on handcrafted linguistic features and unimodal deep learning models, recent research has shifted toward multimodal transformer-based architectures that enable joint reasoning over textual and visual information. This survey provides a comprehensive review of fake news detection methods, with a particular emphasis on multimodal transformer frameworks developed recently. We present a structured taxonomy that categorizes detection approaches into traditional machine learning, deep neural text models, CNN-based multimodal systems, unimodal transformers, and state-of-the-art multimodal transformer architectures. The survey further examines the transition from binary classification to fine-grained misinformation detection, highlighting the importance of cross-modal alignment and semantic consistency modelling. Comparative analysis demonstrates that multimodal transformers consistently outperform earlier paradigms due to their cross-modal attention mechanisms and shared representation learning capabilities. In addition, we identify key research challenges, including dataset bias, annotation subjectivity, adversarial robustness, computational scalability, and explainability. Emerging directions such as multimodal large language models, causal reasoning frameworks, graph-enhanced architectures, and lightweight deployment strategies are discussed to outline future research opportunities. This survey synthesizes recent advances and provides a unified analytical framework and research roadmap for advancing multimodal transformer-based fake news detection toward robust, interpretable, and real-world deployment-ready systems.

Volume 14 Issue 3, March-2026

Pages : c333-c350