1 paper
Yu Chen, Ruihang Liu, Yangguang Xu +5
Spatial audio-visual understanding requires models to recognize not only what is present, but also where events occur and how they relate across modalities. Existing benchmarks oft…