CIFAR-100
Benchmarks
TaskDatasetSOTA resultTrendResults
CIFAR-100 high-C holdout
0.081RMSE (log space)
6
CIFAR-100 rho=0 (test)
84.62Accuracy
6
CIFAR-100 (test)
80.1FPR (CIFAR-10 OOD)
6
CIFAR-100 10+10 classes (test)
74.3Retain Accuracy
6
CIFAR-100 (test)
71.45Baseline Accuracy
6
CIFAR-100 In-Dist NearOOD Aggregate
80.18AUROC
6
CIFAR-100 10-shot
68.9Accuracy
6
CIFAR-100 5-shot
58.22Accuracy
6
CIFAR-100 (D_t)
73.59Accuracy
6
CIFAR-100 Dr (retain)
73.66Accuracy
6
CIFAR-100 25-shot (test)
89.27Top-1 Accuracy
6
CIFAR-100 1-shot (test)
63.74Top-1 Accuracy
6
CIFAR-100 SCOOD benchmark (test)
81.99ID Accuracy
6
CIFAR-100 1.0 (test)
1ToW
6
CIFAR-100 RobustBench standard (val)
73.85Clean Accuracy
6
CIFAR-100 (test)
0.98Poison Success Rate
6
CIFAR-100 16 (test)
81.425-way Acc
6
CIFAR-100 (train)
56.1Accuracy
6
CIFAR-100 (test)
0.67Spearman's rho
6
CIFAR-100 32x32x3 (train)
18Model Size Overhead (MB)
6
CIFAR-100 SCOOD Benchmark Average (6 OOD datasets)
41.05FPR@95%
6
CIFAR-100 (test)
81.99Local Accuracy
6
CIFAR-100 (test)
59.39Type-I Accuracy (70%)
6
CIFAR-100 Type-I 35% Polynomial Margin Diminishing Noise 1.0 (test)
70.13Accuracy
6
CIFAR-100 MI scenario (test)
4.74Mf
6