CoPE-VideoLM: Codec Primitives For Efficient Video Language Models

Benchmark Model Rank Results
question-answering-on-sqa3dCoPE 7B–AnswerExactMatch (Question Answering): 56.6
video-question-answering-on-lvbenchCoPE 7B–Accuracy: 46.4
video-question-answering-on-mvbenchCoPE 7B–Avg.: 61.9
video-question-answering-on-tempcompass-mcqCoPE 7B–Accuracy: 68.9
video-question-answering-on-tomatoCoPE 7B–Accuracy: 28.3
video-question-answering-on-video-mme-with-subtitlesCoPE 7B–Accuracy: 69.4
video-question-answering-on-video-mmmuCoPE 7B–Accuracy: 38.2