Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper addresses the poor generalization of multimodal fake news detectors across domains by introducing Expert-Guided Mutual Distillation (EGMD), which enhances trust in reliable evidence throughout the prediction pipeline. The method employs input-level calibration, expert-guided representation alignment, and prototype-anchored decision-making to mitigate the influence of domain-specific shortcuts and improve model robustness. Experimental results demonstrate that EGMD achieves state-of-the-art accuracy and reduces domain bias by up to 57.3% across four datasets in two languages, highlighting its effectiveness in closing the tail-domain gap.
EGMD slashes domain bias by over 57% while setting new accuracy benchmarks for multimodal fake news detection.
Multimodal fake news detectors often generalize poorly across domains because they learn to trust unreliable evidence: domain-specific shortcuts amplified by imbalanced data and semantically inconsistent text-image pairs that make cross-modal evidence unreliable. We propose Expert-Guided Mutual Distillation (EGMD), which learns what evidence to trust across the prediction pipeline. At the input level, input-level calibration encodes pair-level coherence as a shared gain before fusion. At the representation level, an expert-guided teacher aligns domain statistics and encourages domain-specific patterns to concentrate in specialized experts. At the decision level, prototype-anchored domain-specific students use mutual learning and dual-channel distillation to inherit the teacher's feature geometry and calibrated predictions while discouraging local domain priors. We further construct Weibo_Balanced, a domain-balanced benchmark that isolates the effect of imbalance on generalization. Across four datasets in two languages, EGMD achieves state-of-the-art accuracy while reducing domain bias by up to 57.3%.
Tsinghua AI