15 datasets
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
15 datasets (average)
95.7Coverage
39
15 datasets (average)
83.5Identification Accuracy (8% Budget)
12
15 datasets Average (test)
45.39Average Robust Accuracy
12
15 datasets Zero-shot (test)
44.34Robust Accuracy (PGD-20, eps=1/255)
5