1 paper · 1 filter
Yida Yin, Harish Krishnakumar, Chung Peng Lee +9
In real-world applications, models are expected to perform reliably across diverse settings. Yet, many existing multimodal benchmarks expand task types without capturing the visual…