Binary Classification on MNIST averaged across 45 binary pairs
94Clean AccuracyStandardly pretrained model
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Standardly pretrained modelArchitecture=single-layer linear transformer, Pre-training strategy=Standard2025.05 | 94 | 4 | |
| Adversarially pretrained modelArchitecture=single-layer linear transformer, Pre-training strategy=Adversarial2025.05 | 93 | 72 |