VILA: On Pre-training for Visual Language Models

Benchmark Model Rank Results
zeroshot-video-question-answer-on-msvd-qaVILA1.5-40B#4Accuracy: 80.1
zero-shot-video-question-answer-on-video-mmeVILA-1.5 (34B)#7Accuracy (%): 64.1
zero-shot-video-question-answer-on-video-mme-w-o-subsVILA-1.5 (34B)#46Accuracy (%): 61.4
visual-question-answering-on-mm-vetVILA-13B#79GPT-4 score: 45.7