Bridging VideoQA and Video-Guided Agentic Tasks via Generalized Keyframe Extraction

Benchmark Model Rank Results
zero-shot-video-question-answer-on-next-qaTASKER (GPT-4o)#4Accuracy: 78.1
zero-shot-video-question-answer-on-next-qaTASKER (GPT-4)#5Accuracy: 77.4
zero-shot-video-question-answer-on-egoschema-fullsetTASKER (GPT-4o)#19Accuracy: 63.6
zero-shot-video-question-answer-on-egoschema-fullsetTASKER (GPT-4)#21Accuracy: 63.1