Plug, Play, and Fortify: A Low-Cost Module for Robust Multimodal Image Understanding Models

Siqi Lu, Wanying Xu, Yongbin Zheng +3

Missing modalities present a fundamental challenge in multimodal models, often causing catastrophic performance degradation. Our observations suggest that this fragility stems from…