InternVideo-Next: Towards General Video Foundation Models without Video-Text Supervision

Benchmark Model Rank Results
action-classification-on-kinetics-400InternVideo-Next (s2 Large)–Acc@1: 88.4
action-recognition-in-videos-on-something-something-v2InternVideo-Next (s2 Large)–Top-1 Accuracy: 73.0
video-question-answering-on-mvbenchInternVideo-Next (Large)–Avg.: 50.6