Human-like behavior evaluation on OOD Human-like behavior benchmarks (test)
0.075Accuracy DiffBC pruning
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| BC pruningPruning strategy=BC, Pruning fraction=0.32022.06 | 0.075 | 0.671 | 0.194 | 1.667 | |
| ResNet-50Backbone=ResNet-50, Pre-training source=VISSL2022.06 | 0.077 | 0.671 | 0.184 | 3.667 | |
| BC pruningPruning strategy=BC, Pruning fraction=0.52022.06 | 0.077 | 0.666 | 0.19 | 4 | |
| BC pruningPruning strategy=BC, Pruning fraction=0.42022.06 | 0.078 | 0.67 | 0.187 | 4.333 | |
| BC pruningPruning strategy=BC, Pruning fraction=0.12022.06 | 0.079 | 0.671 | 0.182 | 5.667 | |
| WC pruningPruning strategy=WC, Pruning fraction=0.12022.06 | 0.082 | 0.666 | 0.191 | 4.667 | |
| BC pruningPruning strategy=BC, Pruning fraction=0.22022.06 | 0.084 | 0.66 | 0.175 | 8.333 | |
| WC pruningPruning strategy=WC, Pruning fraction=0.22022.06 | 0.086 | 0.658 | 0.186 | 7.667 | |
| ResNet-50Backbone=ResNet-50, Pre-training source=torchvision2022.06 | 0.087 | 0.665 | 0.208 | 5.667 | |
| WC pruningPruning strategy=WC, Pruning fraction=0.32022.06 | 0.088 | 0.652 | 0.183 | 9.333 | |
| WC pruningPruning strategy=WC, Pruning fraction=0.42022.06 | 0.099 | 0.635 | 0.169 | 11 | |
| WC pruningPruning strategy=WC, Pruning fraction=0.52022.06 | 0.108 | 0.623 | 0.165 | 12 |