TimeViper: A Hybrid Mamba-Transformer Vision-Language Model for Efficient Long Video Understanding

Benchmark Model Rank Results
zero-shot-video-question-answer-on-longvideobenchTimeViper#8Accuracy (% ): 54.1
video-question-answering-on-mlvuTimeViper#47Accuracy: 65.6
zero-shot-video-question-answer-on-video-mme-w-o-subsTimeViper#48Accuracy (%): 58.8
video-question-answering-on-lvbenchTimeViper (w/ TransV)#52Accuracy: 35.6
video-question-answering-on-mvbenchTimeViper#74Avg.: 57.2