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

VSR

Benchmarks

Task NameDataset NameSOTA ResultTrend
Visual Question AnsweringVSR
Top-1 Accuracy73.77
26
Spatial Relationship UnderstandingVSR
Overall Accuracy73.9
17
Spatial ReasoningVSR
LLM-Judge Accuracy84.3
12
2D Spatial ReasoningVSR
Accuracy75.6
10
Spatial ReasoningVSR (ood)
Accuracy84.8
10
Directional attributionVSR (n=240)
DAE96.8
8
Confidence estimationVSR (test)
AUROC67.4
6
Spatial ReasoningVSR zero-shot (test)
Accuracy (zero-shot)63.67
6
GeneralVSR
Score80.6
3
Showing 9 of 9 rows