1 paper · 1 filter
Jihan Yao, Yushi Hu, Yujie Yi +9
Automatically evaluating multimodal generation presents a significant challenge, as automated metrics often struggle to align reliably with human evaluation, especially for complex…