Label Association Fairness on Casual Conversations 1.0 (test)
0Association Magnitude (Female Darker)SEER
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| SEERData=IG-1B, Arch.=RG-10B, Association=Non-Human, Confidence threshold=0.1, Fine-tuning dataset=ImageNet-22K subset2022.02 | 0 | 0.1 | 0.2 | 0 | 0.1 | 0 | 0.1 | 0 | |
| SEERData=IG-1B, Arch.=RG-10B, Association=Crime, Confidence threshold=0.1, Fine-tuning dataset=ImageNet-22K subset2022.02 | 0 | 0 | 0.2 | 0.1 | 0 | 0.1 | 0 | 1.6 | |
| SwAVData=INet-1K, Arch.=RG-128Gf, Association=Non-Human, Confidence threshold=0.1, Fine-tuning dataset=ImageNet-22K subset2022.02 | 0.1 | 0.2 | 0.3 | 0.1 | 0.1 | 0.2 | 0.2 | 0.1 | |
| SwAVData=INet-1K, Arch.=RG-128Gf, Association=Crime, Confidence threshold=0.1, Fine-tuning dataset=ImageNet-22K subset2022.02 | 0.1 | 0.1 | 0.3 | 0.1 | 0.1 | 0.3 | 0.1 | 0.1 | |
| SEERData=IG-1B, Arch.=RG-128Gf, Association=Non-Human, Confidence threshold=0.1, Fine-tuning dataset=ImageNet-22K subset2022.02 | 0.1 | 0.6 | 0.7 | 0.7 | 0.8 | 0.1 | 0.5 | 3.2 | |
| SEERData=IG-1B, Arch.=RG-128Gf, Association=Crime, Confidence threshold=0.1, Fine-tuning dataset=ImageNet-22K subset2022.02 | 0.1 | 0.1 | 0.2 | 0.1 | 0.1 | 0.1 | 0.2 | 0.1 | |
| SupervisedData=INet-1K, Arch.=RG-128Gf, Association=Crime, Confidence threshold=0.1, Fine-tuning dataset=ImageNet-22K subset2022.02 | 1.2 | 0.2 | 0.7 | 0.4 | 0.6 | 0.9 | 0.1 | 3.2 | |
| SupervisedData=INet-1K, Arch.=RG-128Gf, Association=Non-Human, Confidence threshold=0.1, Fine-tuning dataset=ImageNet-22K subset2022.02 | 2.3 | 6 | 2 | 1.8 | 2.1 | 2.4 | 5.4 | 4.9 | |
| SEERData=IG-1B, Arch.=RG-10B, Association=Possibly-Human, Confidence threshold=0.1, Fine-tuning dataset=ImageNet-22K subset2022.02 | 20.2 | 27.9 | 72.6 | 65.1 | 44.9 | 48.3 | 39.5 | 22.6 | |
| SEERData=IG-1B, Arch.=RG-128Gf, Association=Possibly-Human, Confidence threshold=0.1, Fine-tuning dataset=ImageNet-22K subset2022.02 | 23.8 | 21.8 | 56.4 | 40.6 | 38.7 | 38.6 | 24.8 | 6.45 | |
| SupervisedData=INet-1K, Arch.=RG-128Gf, Association=Possibly-Human, Confidence threshold=0.1, Fine-tuning dataset=ImageNet-22K subset2022.02 | 24.3 | 41.4 | 50.1 | 54 | 43.9 | 43.7 | 39.7 | 22.7 | |
| SupervisedData=INet-1K, Arch.=RG-128Gf, Association=Human, Confidence threshold=0.1, Fine-tuning dataset=ImageNet-22K subset2022.02 | 37.4 | 18.5 | 29.5 | 17.5 | 26.9 | 25.7 | 22.8 | 21 | |
| SwAVData=INet-1K, Arch.=RG-128Gf, Association=Human, Confidence threshold=0.1, Fine-tuning dataset=ImageNet-22K subset2022.02 | 58.7 | 58.2 | 32.2 | 43.1 | 46.6 | 44.7 | 57.9 | 46.8 | |
| SwAVData=INet-1K, Arch.=RG-128Gf, Association=Possibly-Human, Confidence threshold=0.1, Fine-tuning dataset=ImageNet-22K subset2022.02 | 66.9 | 66.4 | 82.5 | 70.4 | 70.8 | 73.4 | 69.1 | 53.2 | |
| SEERData=IG-1B, Arch.=RG-128Gf, Association=Human, Confidence threshold=0.1, Fine-tuning dataset=ImageNet-22K subset2022.02 | 78.7 | 73.3 | 40 | 43.3 | 58.4 | 57.4 | 66.1 | 67.7 | |
| SEERData=IG-1B, Arch.=RG-10B, Association=Human, Confidence threshold=0.1, Fine-tuning dataset=ImageNet-22K subset2022.02 | 93 | 87.3 | 57.2 | 59.8 | 73.3 | 72.7 | 82.4 | 79 |