LiteFrame: Efficient Vision Encoders Unlock Frame Scaling in Video LLMs

Benchmark Model Rank Results
video-question-answering-on-video-mme-with-subtitlesLiteFrame (w/ subtitles)#28Accuracy: 68.5
video-question-answering-on-mlvuLiteFrame#32Accuracy: 70.7
video-question-answering-on-lvbenchLiteFrame#46Accuracy: 43.9