1 paper · 1 filter
Weijie Tu, Weijian Deng, Dylan Campbell +4
Can the relative performance of a pre-trained large multimodal model (LMM) be predicted without access to labels? As LMMs proliferate, it becomes increasingly important to develop…