1 paper · 1 filter
Jisheng Dang, Kai Wei, Ronghao Lin +6
Omni-modal models can answer questions about video, but their predictions rarely show which participant, event, or interval supports an answer. Moreover, answer-level supervision a…