SOURCE-LINKED INTELLIGENCE

Mitigating Strong-Modality Collapse in Multimodal Learning via Inverted Asymmetric Fusion

arXiv · AI, language, vision and robotics · article · Aug 27, 2026 · UTC

Fusing multiple modalities is expected to improve model performance. However, on the MultiHuSE dataset, early, late, and symmetric attention fusion often fail to outperform the best unimodal baseline (text). Pathway isolation of a symmetric attention fusion model reveals that the text-pathway accuracy drops from 74.9% to 56.4% after fusion in one such setting, indicating that the dominant modality can be degraded during integration. We term this strong-modality collapse and argue that it helps explain why some multimodal models fail to surpass unimodal baselines. We propose Inverted Asymmetric

Read original source ↗ Open in workspace

recordType

paper

region

Global