| Task Name | Dataset Name | SOTA Result | Trend | |
|---|---|---|---|---|
| Image Classification | EMNIST (test) | Accuracy94.63 | 234 | |
| Image Classification | EMNIST Balanced | Accuracy91.06 | 103 | |
| Image Classification | EMNIST | Accuracy98.9 | 82 | |
| Image Classification | EMNIST-L (test) | Top-1 Acc94.81 | 36 | |
| Image Classification | EMNIST Balanced (test) | Accuracy91.06 | 34 | |
| Image Classification | EMNIST Letters 4 (test) | Accuracy67.42 | 33 | |
| Image Classification | EMNIST Dir(0.1) (test) | Test Accuracy96.55 | 28 | |
| Clustering | E-MNIST | Accuracy95.5 | 25 | |
| Image Classification | M(EMNIST-20, F-EMNIST) | Accuracy97.12 | 22 | |
| Image Classification | EMNIST Dir(0.5) (test) | Test Accuracy91.82 | 18 | |
| Clustering | emnist | ARI41.2 | 17 | |
| Image Classification | EMNIST Letters (test) | Accuracy95.96 | 15 | |
| Handwritten Character Classification | EMNIST-Letters (test) | Accuracy95.88 | 15 | |
| Image Classification | EMNIST Private Task Finetuned (test) | Accuracy85 | 14 | |
| Image Classification | EMNIST Public Task Finetuned (test) | Accuracy88.5 | 14 | |
| Out-of-Distribution Detection | EMNIST (test) | AUROC0.917 | 14 | |
| Adversarial Reconstruction Defense | EMNIST (test) | FID (A)20.8 | 13 | |
| Image Reconstruction | EMNIST (test) | MSE0.302 | 12 | |
| Image Classification | EMNIST (Non-IID (4-Class partition)) | Accuracy86.88 | 12 | |
| Lossless Image Compression | EMNIST ByClass (test) | Bits Per Pixel (BPP)2.3 | 12 | |
| Handwritten Character Classification | EMNIST-Digits (test) | Accuracy99.82 | 12 | |
| Conformal Prediction | EMNIST ByClass (test) | Mean Prediction Set Size1.98 | 12 | |
| Conformal Prediction | EMNIST | Avg Prediction Set Size2.19 | 12 | |
| Conformal Prediction | EMNIST 10 different calib. (test) | Mean Prediction Set Size2.42 | 12 | |
| Classification | EMNIST (test) | Mean Prediction Set Size1.98 | 12 |