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

Perception Test

Benchmarks

Task NameDataset NameSOTA ResultTrend
Video Action RecognitionPerception Test
Top-1 Accuracy60.9
16
Short Video Question AnsweringPerception Test (val)
Accuracy74.3
9
Visual Entity TrackingPerception Test Filtered Subset (test)
Accuracy42
6
Video Question AnsweringPerception Test zero-shot few-shot
Accuracy50.2
6
Video ReasoningPerception Test
Accuracy67.2
5
Sound LocalisationPerception Test (val)
AP@0.137.5
2
Action LocalisationPerception Test (val)
AP@0.133.5
2
Showing 7 of 7 rows