VideoChat-T (ZS)

Reported on 7 benchmarks across 3 tasks · 1 paper · 1 SOTA

Note: results are matched by exact model name. Different papers may use the same name for different model variants.

Computer Vision5 results

Moment RetrievalonCharades-STA
mIoU· uses extra data· 2024-10-25
45.43
best: 53.44 (LD-DETR)
SOTA
TimeSuite: Improving MLLMs for Long Video Understanding via Grounded Tuning arXiv:2410.19702
Moment RetrievalonCharades-STA
R@1 IoU=0.5· uses extra data· 2024-10-25
48.7
best: 71.1 (SG-DETR (w/ PT))
TimeSuite: Improving MLLMs for Long Video Understanding via Grounded Tuning arXiv:2410.19702
Moment RetrievalonCharades-STA
R@1 IoU=0.7· uses extra data· 2024-10-25
24
best: 52.8 (SG-DETR (w/ PT))
TimeSuite: Improving MLLMs for Long Video Understanding via Grounded Tuning arXiv:2410.19702
Highlight DetectiononQVHighlights
Hit@1
54.1
best: 71.01 (FlashVTG)
Highlight DetectiononQVHighlights
mAP
26.5
best: 44.7 (SG-DETR (w/ PT))

Methodology2 results

16konQVHighlights
Hit@1
54.1
best: 71.01 (FlashVTG)
16konQVHighlights
mAP
26.5
best: 44.7 (SG-DETR (w/ PT))