Zoom-Zero: Reinforced Coarse-to-Fine Video Understanding via Temporal Zoom-in
Open paper
Benchmark
Model
Rank
Results
video-question-answering-on-mlvu
Zoom-Zero + Divide-and-conquer
#27
Accuracy: 73.4
video-question-answering-on-lvbench
Zoom-Zero + Divide-and-conquer
#40
Accuracy: 48.1
Rank counts only results with a code link.