LENS: Adaptive Spatio-Temporal Zooming for Keyframe Sampling in Long-Form Videos
Open paper
Benchmark
Model
Rank
Results
zero-shot-video-question-answer-on-egoschema-fullset
LENS (Qwen2.5-VL 7B, 32 frames)
#16
Accuracy: 66.1
video-question-answering-on-mlvu
LENS (Qwen2.5-VL 7B, 32 frames)
#46
Accuracy: 66.7
Rank counts only results with a code link.