Image Classification on CIFAR10 (Task and Concept Accuracy)
91.57Task AccuracyDNN
Evaluation results
| Method | Links | ||
|---|---|---|---|
| DNNBackbone=ResNet2026.05 | 91.57 | — | |
| SynCBMBackbone=ResNet2026.05 | 91.49 | 76.85 | |
| SynCEMBackbone=ResNet2026.05 | 91.12 | 76.28 | |
| CEMBackbone=ResNet2026.05 | 90.48 | 79.69 | |
| CBMBackbone=ResNet2026.05 | 90.46 | 79.31 | |
| HP-CBMBackbone=ResNet2026.05 | 87.13 | 52.03 | |
| ProbCBMBackbone=ResNet2026.05 | 78.97 | 77.57 | |
| MixCEMBackbone=ResNet2026.05 | 76.61 | 73.99 |