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

WorldSense

Benchmarks

Task NameDataset NameSOTA ResultTrend
Audio-visual understandingWorldSense
Accuracy66.4
42
Video UnderstandingWorldSense
Score52.01
25
Multimodal Fact-Level AttributionWorldSense 1.0 (sampled examples)
Accuracy71.4
24
Long Audio-Video Question AnsweringWorldSense
Average Accuracy61.2
18
Audio-visual Question AnsweringWorldSense
Accuracy50
18
Audio-Visual PerceptionWorldSense
Score47.4
8
Video ReasoningWorldSense
Accuracy40.4
7
Commonsense ReasoningWorldSense
Overall Score42.6
7
Audio-Visual QuestionWorldSense
Accuracy (Clean)59.7
6
Video Grounded ReasoningWorldSense
Original Score45.4
6
Common Sense ReasoningWorldSense
Accuracy0.637
6
Video Question AnsweringWorldSense
Accuracy49.2
5
Visual Question AnsweringWorldSense sampled examples 1.0
Accuracy60
4
Showing 13 of 13 rows