1 paper · 1 filter
Tao Yu, yiming ding, Shenghua Chai +16
Current omni-modal benchmarks mainly evaluate models under settings where multiple modalities are provided simultaneously, while the ability to start from audio alone and actively…