CogniRoute: Learning to Route Social Evidence in Omni-Modal Models

Benchmark Model Rank Results
audio-visual-question-answering-on-dailyomniCogniRoute–Accuracy: 74.3
audio-visual-question-answering-on-omnibenchCogniRoute–Accuracy: 56.7
audio-visual-question-answering-on-omnivideobenchCogniRoute–Accuracy: 38.9
audio-visual-question-answering-on-worldsenseCogniRoute–Accuracy: 53.8
video-question-answering-on-mmvuCogniRoute–Accuracy: 60.7
video-question-answering-on-mvbenchCogniRoute–Avg.: 70.2