Image Classification on CCC medium (test)
43.35Accuracy (1000 speed)KLSoft
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| KLSoftBackbone=ResNet-50, Pre-training=ImageNet, Evaluation Seeds=3, Sequence Length=1M steps, Update Parameters=BatchNorm affine, Reset Strategy=Soft, Variant=RDumb++2026.01 | 43.35 | 37.77 | 43.82 | 41.65 | |
| EntropyFullBackbone=ResNet-50, Pre-training=ImageNet, Evaluation Seeds=3, Sequence Length=1M steps, Update Parameters=BatchNorm affine, Reset Strategy=Full, Variant=RDumb++2026.01 | 43.13 | 44.25 | 45.22 | 44.2 | |
| KLFullBackbone=ResNet-50, Pre-training=ImageNet, Evaluation Seeds=3, Sequence Length=1M steps, Update Parameters=BatchNorm affine, Reset Strategy=Full, Variant=RDumb++2026.01 | 42.68 | 43.53 | 45.03 | 43.75 | |
| EntropySoftBackbone=ResNet-50, Pre-training=ImageNet, Evaluation Seeds=3, Sequence Length=1M steps, Update Parameters=BatchNorm affine, Reset Strategy=Soft, Variant=RDumb++2026.01 | 40.16 | 40.59 | 44.28 | 41.67 | |
| RDumbBackbone=ResNet-50, Pre-training=ImageNet, Evaluation Seeds=3, Sequence Length=1M steps, Update Parameters=BatchNorm affine2026.01 | 37.41 | 42.71 | 44.71 | 41.61 | |
| BaselineBackbone=ResNet-50, Pre-training=ImageNet, Evaluation Seeds=3, Sequence Length=1M steps, Adaptation=None2026.01 | 15.84 | 17.22 | 17.47 | 16.84 |