Image Classification on MNIST (unseen classes 5-9)
89.47AccuracyULHM
Evaluation Results
| Method | Links | |
|---|---|---|
| ULHMmode=zero-shot2026.01 | 89.47 | |
| Contrastive Learningmode=zero-shot2026.01 | 87.9 | |
| Mixup Augmentationmode=zero-shot2026.01 | 87.79 | |
| Variational Autoencodersmode=zero-shot2026.01 | 85.76 | |
| Triplet Networksmode=zero-shot2026.01 | 77.25 | |
| Correlation Alignmentmode=zero-shot2026.01 | 75.58 | |
| Prototypical Networksmode=zero-shot2026.01 | 67.76 | |
| Max. Classifier Discrepancymode=zero-shot2026.01 | 62.75 | |
| Self-Supervised Rotationmode=zero-shot2026.01 | 38.26 |