1 paper · 1 filter
Jihan Yao, Yushi Hu, Yujie Yi +10
Automatically evaluating multimodal generation presents a significant challenge, as automated metrics often struggle to align with human evaluation, especially for complex tasks th…