MeToM: Metadata-Guided Token Merging for Efficient Video LLMs
Open paper
Benchmark
Model
Rank
Results
video-question-answering-on-mlvu
LLaVA-Video-7B + MeToM
–
Accuracy: 70.4
zero-shot-video-question-answer-on-egoschema-fullset
LLaVA-OneVision-7B + MeToM
–
Accuracy: 62.4
Rank counts only results with a code link.