Flash-VStream: Memory-Based Real-Time Understanding for Long Video Streams

Benchmark Model Rank Results
question-answering-on-next-qa-open-ended-videoqaFlash-VStream#1Accuracy: 61.6Confidence Score: 3.4
zeroshot-video-question-answer-on-msrvtt-qaFlash-VStream#1Accuracy: 72.4Confidence Score: 3.4
zeroshot-video-question-answer-on-msvd-qaFlash-VStream#1Accuracy: 80.3Confidence Score: 3.9
zeroshot-video-question-answer-on-activitynet-qaFlash-VStream#9Accuracy: 51.9Confidence Score: 3.4
video-question-answering-on-ovbenchFlash-Vstream (7B)#12AVG: 31.2