Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

MVBench

Benchmarks

Task NameDataset NameSOTA ResultTrend
Video UnderstandingMVBench
Accuracy100
635
Video UnderstandingMVBench (test)
Accuracy100.3
201
Video Question AnsweringMVBench
Accuracy81.3
90
Multi-modal Video UnderstandingMVBench
Score79.4
84
Multi-choice Video Question AnsweringMVBench
Avg Accuracy67.6
73
Video Question AnsweringMVBench
Accuracy77.1
72
Video ReasoningMVBench
MVBench Score84.3
56
Video Question AnsweringMVBench (test)
Accuracy73.6
45
Video UnderstandingMVBench
MVBench Score67.5
39
General Video UnderstandingMVBench Overall
Accuracy86.03
39
Adversarial AttackMVBench
ASR83.84
37
Short video understandingMVBench
Accuracy76.4
28
Video UnderstandingMVBench zero-shot
Accuracy69.7
25
Video Question AnsweringMVBench 1.0 (test)
AS Score77
25
Fine-grained Video UnderstandingMVBench
Accuracy73.8
22
Video Question AnsweringMVBench
Average Score61.2
20
Video understandingMVBench (val)
Accuracy54.9
20
Multiple-choice question answeringMVBench
Accuracy58.54
15
Video GeneralizationMVBench
MVBench Score72.3
14
Exocentric Video UnderstandingMVBench
Score73.8
13
Text GenerationMVBench
MVBench Score75.4
12
Event-level Video UnderstandingMVBench
Average Accuracy58.3
12
General spatiotemporal perceptionMVbench
Score49
11
Video Question AnsweringMVBench
Overall Score59.13
10
Video Question AnsweringMVBench
Accuracy72
10
Showing 25 of 37 rows