Concept Accuracy on Broden simplified (test)
79.07Concept AccuracyLG-CAV + GA + CE + DSR
Evaluation results
| Method | Links | ||
|---|---|---|---|
| LG-CAV + GA + CE + DSRBackbone=DenseNet-121, Pre-trained=ImageNet2024.10 | 79.07 | — | |
| LG-CAV + GA + CEBackbone=DenseNet-121, Pre-trained=ImageNet2024.10 | 78.18 | — | |
| LG-CAV + GA + CE + DSRBackbone=ResNet-18, Pre-trained=ImageNet2024.10 | 77.45 | — | |
| LG-CAV + GA + CE + DSRBackbone=DenseNet-169, Pre-trained=ImageNet2024.10 | 77.25 | — | |
| LG-CAV + GA + CE + DSRBackbone=ResNet-50, Pre-trained=ImageNet2024.10 | 76.48 | — | |
| LG-CAV + GA + CEBackbone=ResNet-18, Pre-trained=ImageNet2024.10 | 76.41 | — | |
| LG-CAV + GABackbone=DenseNet-121, Pre-trained=ImageNet2024.10 | 76.28 | — | |
| LG-CAV + GA + CEBackbone=DenseNet-169, Pre-trained=ImageNet2024.10 | 76.12 | — | |
| LG-CAV + GA + CE + DSRBackbone=ResNet-34, Pre-trained=ImageNet2024.10 | 76.04 | — | |
| LG-CAV + GA + CEBackbone=ResNet-50, Pre-trained=ImageNet2024.10 | 75.63 | — | |
| OA-TCAVBackbone=DenseNet-169, Pre-trained=ImageNet2024.10 | 74.89 | — | |
| LG-CAV + GABackbone=ResNet-18, Pre-trained=ImageNet2024.10 | 74.89 | — | |
| LG-CAV + GABackbone=DenseNet-169, Pre-trained=ImageNet2024.10 | 74.7 | — | |
| LG-CAV + GA + CEBackbone=ResNet-34, Pre-trained=ImageNet2024.10 | 74.47 | — | |
| Text-to-ConceptBackbone=DenseNet-169, Pre-trained=ImageNet2024.10 | 74.19 | — | |
| LG-CAV + GABackbone=ResNet-50, Pre-trained=ImageNet2024.10 | 74.19 | — | |
| OA-TCAVBackbone=DenseNet-121, Pre-trained=ImageNet2024.10 | 73.9 | — | |
| Text-to-ConceptBackbone=DenseNet-121, Pre-trained=ImageNet2024.10 | 73.67 | — | |
| LG-CAV + GABackbone=ResNet-34, Pre-trained=ImageNet2024.10 | 73.47 | — | |
| LG-CAV + GA + CE + DSRBackbone=VGG-19, Pre-trained=ImageNet2024.10 | 73.47 | — | |
| OA-TCAVBackbone=ResNet-50, Pre-trained=ImageNet2024.10 | 73.24 | — | |
| LG-CAV + GA + CEBackbone=VGG-19, Pre-trained=ImageNet2024.10 | 72.92 | — | |
| Original CAVBackbone=DenseNet-169, Pre-trained=ImageNet2024.10 | 72.7 | — | |
| OA-TCAVBackbone=ResNet-18, Pre-trained=ImageNet2024.10 | 72.62 | — | |
| Original CAVBackbone=DenseNet-121, Pre-trained=ImageNet2024.10 | 72.46 | — | |
| Text-to-ConceptBackbone=ResNet-50, Pre-trained=ImageNet2024.10 | 72.4 | — | |
| LG-CAV + GABackbone=VGG-19, Pre-trained=ImageNet2024.10 | 72.31 | — | |
| OA-TCAVBackbone=ResNet-34, Pre-trained=ImageNet2024.10 | 72.2 | — | |
| Text-to-ConceptBackbone=ResNet-34, Pre-trained=ImageNet2024.10 | 71.35 | — | |
| Original CAVBackbone=ResNet-50, Pre-trained=ImageNet2024.10 | 71.34 | — | |
| OA-TCAVBackbone=VGG-19, Pre-trained=ImageNet2024.10 | 70.81 | — | |
| LG-CAV + GA + CE + DSRBackbone=VGG-13, Pre-trained=ImageNet2024.10 | 70.69 | — | |
| LG-CAV + GA + CE + DSRBackbone=ViT-B, Pre-trained=ImageNet2024.10 | 70.52 | — | |
| LG-CAV + GA + CEBackbone=VGG-13, Pre-trained=ImageNet2024.10 | 70.25 | — | |
| Text-to-ConceptBackbone=VGG-19, Pre-trained=ImageNet2024.10 | 70.24 | — | |
| LG-CAV + GA + CE + DSRBackbone=ViT-L, Pre-trained=ImageNet2024.10 | 70.09 | — | |
| Text-to-ConceptBackbone=ResNet-18, Pre-trained=ImageNet2024.10 | 70.04 | — | |
| LG-CAV + GABackbone=VGG-13, Pre-trained=ImageNet2024.10 | 69.63 | — | |
| Original CAVBackbone=VGG-19, Pre-trained=ImageNet2024.10 | 69.56 | — | |
| LG-CAVBackbone=DenseNet-121, Pre-trained=ImageNet2024.10 | 69.43 | — | |
| LG-CAV + GA + CEBackbone=ViT-B, Pre-trained=ImageNet2024.10 | 69.43 | — | |
| Original CAVBackbone=ResNet-34, Pre-trained=ImageNet2024.10 | 69.32 | — | |
| LG-CAV + GA + CEBackbone=ViT-L, Pre-trained=ImageNet2024.10 | 69.31 | — | |
| LG-CAV + GABackbone=ViT-B, Pre-trained=ImageNet2024.10 | 68.99 | — | |
| Original CAVBackbone=ResNet-18, Pre-trained=ImageNet2024.10 | 68.92 | — | |
| OA-TCAVBackbone=VGG-13, Pre-trained=ImageNet2024.10 | 68.69 | — | |
| LG-CAVBackbone=DenseNet-169, Pre-trained=ImageNet2024.10 | 68.46 | — | |
| LG-CAV + GABackbone=ViT-L, Pre-trained=ImageNet2024.10 | 68.41 | — | |
| Text-to-ConceptBackbone=VGG-13, Pre-trained=ImageNet2024.10 | 68.35 | — | |
| LG-CAVBackbone=VGG-19, Pre-trained=ImageNet2024.10 | 67.94 | — | |
| OA-TCAVBackbone=ViT-B, Pre-trained=ImageNet2024.10 | 67.83 | — | |
| LG-CAVBackbone=ResNet-50, Pre-trained=ImageNet2024.10 | 67.52 | — | |
| LG-CAVBackbone=ResNet-34, Pre-trained=ImageNet2024.10 | 67.48 | — | |
| Original CAVBackbone=VGG-13, Pre-trained=ImageNet2024.10 | 67.44 | — | |
| OA-TCAVBackbone=ViT-L, Pre-trained=ImageNet2024.10 | 67.35 | — | |
| LG-CAVBackbone=ResNet-18, Pre-trained=ImageNet2024.10 | 67.23 | — | |
| Text-to-ConceptBackbone=ViT-B, Pre-trained=ImageNet2024.10 | 67.22 | — | |
| Text-to-ConceptBackbone=ViT-L, Pre-trained=ImageNet2024.10 | 66.27 | — | |
| LG-CAVBackbone=VGG-13, Pre-trained=ImageNet2024.10 | 65.99 | — | |
| Original CAVBackbone=ViT-L, Pre-trained=ImageNet2024.10 | 65.85 | — | |
| Original CAVBackbone=ViT-B, Pre-trained=ImageNet2024.10 | 65.35 | — | |
| LG-CAVBackbone=ViT-L, Pre-trained=ImageNet2024.10 | 63.22 | — | |
| LG-CAVBackbone=ViT-B, Pre-trained=ImageNet2024.10 | 63.16 | — | |
| LG-CAVBackbone=ResNet-182024.10 | — | 4.72 | |
| LG-CAVBackbone=ResNet-342024.10 | — | 6.66 | |
| LG-CAVBackbone=ResNet-502024.10 | — | 5.99 | |
| LG-CAVBackbone=DenseNet-1212024.10 | — | 5.64 | |
| LG-CAVBackbone=DenseNet-1692024.10 | — | 6.02 | |
| LG-CAVBackbone=VGG-132024.10 | — | 4.02 | |
| LG-CAVBackbone=VGG-192024.10 | — | 4.49 | |
| LG-CAVBackbone=ViT-B2024.10 | — | 6.07 | |
| LG-CAVBackbone=ViT-L2024.10 | — | 6.38 | |
| LG-CAV (Full: GA + CE + DSR)Backbone=ResNet-182024.10 | — | 24.58 | |
| LG-CAV (Full: GA + CE + DSR)Backbone=ResNet-342024.10 | — | 25.61 | |
| LG-CAV (Full: GA + CE + DSR)Backbone=ResNet-502024.10 | — | 26.05 | |
| LG-CAV (Full: GA + CE + DSR)Backbone=DenseNet-1212024.10 | — | 23.93 | |
| LG-CAV (Full: GA + CE + DSR)Backbone=DenseNet-1692024.10 | — | 23.97 | |
| LG-CAV (Full: GA + CE + DSR)Backbone=VGG-132024.10 | — | 21.4 | |
| LG-CAV (Full: GA + CE + DSR)Backbone=VGG-192024.10 | — | 22.79 | |
| LG-CAV (Full: GA + CE + DSR)Backbone=ViT-B2024.10 | — | 26.12 | |
| LG-CAV (Full: GA + CE + DSR)Backbone=ViT-L2024.10 | — | 27.72 | |
| LG-CAV + GABackbone=ResNet-182024.10 | — | 16.72 | |
| LG-CAV + GABackbone=ResNet-342024.10 | — | 16.61 | |
| LG-CAV + GABackbone=ResNet-502024.10 | — | 16.92 | |
| LG-CAV + GABackbone=DenseNet-1212024.10 | — | 15.47 | |
| LG-CAV + GABackbone=DenseNet-1692024.10 | — | 15.81 | |
| LG-CAV + GABackbone=VGG-132024.10 | — | 14.55 | |
| LG-CAV + GABackbone=VGG-192024.10 | — | 15.04 | |
| LG-CAV + GABackbone=ViT-B2024.10 | — | 17.52 | |
| LG-CAV + GABackbone=ViT-L2024.10 | — | 17.83 | |
| LG-CAV + GA + CEBackbone=ResNet-182024.10 | — | 19.14 | |
| LG-CAV + GA + CEBackbone=ResNet-342024.10 | — | 20.1 | |
| LG-CAV + GA + CEBackbone=ResNet-502024.10 | — | 20.51 | |
| LG-CAV + GA + CEBackbone=DenseNet-1212024.10 | — | 18.78 | |
| LG-CAV + GA + CEBackbone=DenseNet-1692024.10 | — | 19 | |
| LG-CAV + GA + CEBackbone=VGG-132024.10 | — | 17.5 | |
| LG-CAV + GA + CEBackbone=VGG-192024.10 | — | 18.35 | |
| LG-CAV + GA + CEBackbone=ViT-B2024.10 | — | 21.52 | |
| LG-CAV + GA + CEBackbone=ViT-L2024.10 | — | 22.34 | |
| OA-TCAVBackbone=ResNet-182024.10 | — | 10.11 |