Multimodal Adaptive Expert Selection with Text Routing and Ordinal Prototype Optimization for Sentiment Analysis
This work introduces Multimodal Adaptive Expert Selection with Text Routing and Ordinal prototype optimization (MAESTRO), a novel framework designed to dynamically orchestrate and refine multimodal representations and proposes an Ordinal-aware Prototype Contrastive Learning (O-PCL), a structured latent space that prese...
Xiao-De Chen, Jiakang Yu, Hong-Tao Deng et al.
· 0 citations