Out-of-Distribution Detection on SVHN (AUROC vs C10, CelebA, C100, Tex)
97.3AUROC vs C10Unified Few-shot
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Unified Few-shotTraining Mode=Few-shot, Training Set=CelebA, Samples per task=80-1502026.05 | 97.3 | 99.9 | 97.4 | 96.9 | |
| DoSTraining Mode=Full-data, Training Set=CelebA2026.05 | 96.7 | 76.9 | 96.5 | 88.6 | |
| NLLTraining Mode=Full-data, Training Set=CelebA2026.05 | 96.4 | 79.2 | 96 | 87.6 | |
| LMDTraining Mode=Full-data, Training Set=CelebA2026.05 | 95.8 | 94.4 | 95.6 | 91.1 | |
| DiffPathTraining Mode=Full-data, Training Set=CelebA2026.05 | 93.9 | 97.9 | 95.3 | 98.1 | |
| DDPM-OODTraining Mode=Full-data, Training Set=CelebA2026.05 | 93.7 | 76.7 | 90.9 | 72.5 | |
| VAEBMTraining Mode=Full-data, Training Set=CelebA2026.05 | 49.7 | 15.8 | 46.6 | 44.9 | |
| IGEBMTraining Mode=Full-data, Training Set=CelebA2026.05 | 37.2 | 24.9 | 33.4 | 41 | |
| MSMATraining Mode=Full-data, Training Set=CelebA2026.05 | 35.5 | 38.1 | 34.4 | 32.9 |