VideoTree: Adaptive Tree-based Video Representation for LLM Reasoning on Long Videos

Benchmark Model Rank Results
zero-shot-video-question-answer-on-egoschema-subsetVideoTree (GPT4)#4Accuracy: 66.2
zero-shot-video-question-answer-on-intentqaVideoTree (GPT4)#4Accuracy: 66.9
zero-shot-video-question-answer-on-next-qaVideoTree (GPT4)#7Accuracy: 73.5
zero-shot-video-question-answer-on-egoschema-fullsetVideoTree (GPT4)#27Accuracy: 61.1