MiniGPT4-Video: Advancing Multimodal LLMs for Video Understanding with Interleaved Visual-Textual Tokens

Benchmark Model Rank Results
zero-shot-video-question-answer-on-tvqaMiniGPT4-video-7B#3Accuracy: 54.21
video-based-generative-performance-benchmarking-detail-orientation-on-videoinstructMiniGPT4-video-7B#7gpt-score: 3.02
video-based-generative-performance-benchmarking-contextual-understanding-on-videoinstructMiniGPT4-video-7B#8gpt-score: 3.57
video-based-generative-performance-benchmarking-correctness-of-information-on-videoinstructMiniGPT4-video-7B#8gpt-score: 3.08
video-based-generative-performance-benchmarking-consistency-on-videoinstructMiniGPT4-video-7B#9gpt-score: 2.67
video-based-generative-performance-benchmarking-temporal-understanding-on-videoinstructMiniGPT4-video-7B#9gpt-score: 2.65
zeroshot-video-question-answer-on-tgif-qaMiniGPT4-video-7B#9Accuracy: 72.22
zeroshot-video-question-answer-on-msvd-qaMiniGPT4-video-7B#13Accuracy: 73.92
zeroshot-video-question-answer-on-msrvtt-qaMiniGPT4-video-7B#14Accuracy: 59.73
zeroshot-video-question-answer-on-activitynet-qaMiniGPT4-video-7B#19Accuracy: 46.3