Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

GRIT

Benchmarks

Task NameDataset NameSOTA ResultTrend
General Robust Image Task (GRIT) multi-task evaluationGRIT ablation set (same)
Categorization Accuracy85
38
Referring Expression ComprehensionGRIT refexp
Accuracy78.61
15
Multi-task Vision and Language EvaluationGRIT (test)
Overall Score67
14
Multi-task vision and language evaluationGRIT (General Robustness and Information Transfer) unrestricted track (test)
Captioning Acc55.1
2
Showing 4 of 4 rows