Tag
This paper proposes Expert-Guided Mutual Distillation (EGMD) to address domain bias and semantic misalignment in multimodal fake news detection, achieving state-of-the-art accuracy and reducing domain bias by up to 57.3% across four datasets.