SHAPE: An Unified Approach to Evaluate the Contribution and Cooperation of Individual Modalities

Pengbo Hu, Xingyu Li, Yi Zhou

As deep learning advances, there is an ever-growing demand for models capable of synthesizing information from multi-modal resources to address the complex tasks raised from real-l…