TimeChat: A Time-sensitive Multimodal Large Language Model for Long Video Understanding

Benchmark Model Rank Results
video-text-retrieval-on-test-of-timeTime-Chat#22-Class Accuracy: 76.67
video-question-answering-on-ovbenchTimeChat (7B)#15AVG: 12.8
zero-shot-video-question-answer-on-egoschema-fullsetTimeChat (7B)#50Accuracy: 33.0
video-question-answering-on-mvbenchTimeChat#85Avg.: 38.5