Mirasol3B: A Multimodal Autoregressive model for time-aligned and contextual modalities

Benchmark Model Rank Results
action-classification-on-kinetics-soundsMirasol3BTop 1 Accuracy: 90.1
audio-classification-on-vggsoundMirasol3BTop 1 Accuracy: 69.8
video-question-answering-on-activitynet-qaMirasol3BAccuracy: 51.13
video-question-answering-on-msrvtt-qaMirasol3BAccuracy: 50.42
video-question-answering-on-next-qaMirasol3BAccuracy: 72