Video-RAG: Visually-aligned Retrieval-Augmented Long Video Comprehension

Benchmark Model Rank Results
zero-shot-video-question-answer-on-video-mmeVideo-RAG (Based on LLaVA-Video)#2Accuracy (%): 77.4
zero-shot-video-question-answer-on-longvideobenchVideo-RAG (based on LLaVA-Video)#4Accuracy (% ): 65.4
zero-shot-video-question-answer-on-video-mme-w-o-subsVideo-RAG (based on LLaVA-Video)#9Accuracy (%): 77.4
zero-shot-video-question-answer-on-egoschema-fullsetVideo-RAG (Based on LLaVA-Video)#15Accuracy: 66.7