Defense against adaptive adversarial attacks on JigsawToxicity
73.58AccuracyPre
Evaluation Results
| Method | Links | |
|---|---|---|
| PreDefense Protocol=Relabeling2024.05 | 73.58 | |
| PreDefense Protocol=Reweighting2024.05 | 73.58 | |
| LiSSADefense Protocol=Relabeling2024.05 | 73.3 | |
| Self-LiSSADefense Protocol=Reweighting2024.05 | 71.41 | |
| Self-TracInDefense Protocol=Reweighting2024.05 | 71.37 | |
| DataInfDefense Protocol=Reweighting2024.05 | 70.82 | |
| IPDefense Protocol=Reweighting2024.05 | 70.82 | |
| IP EnsembleDefense Protocol=Reweighting2024.05 | 70.67 | |
| IP EnsembleDefense Protocol=Relabeling2024.05 | 70.51 | |
| IPDefense Protocol=Relabeling2024.05 | 70.43 | |
| DataInfDefense Protocol=Relabeling2024.05 | 70.39 | |
| LiSSADefense Protocol=Reweighting2024.05 | 70.13 | |
| EKFACDefense Protocol=Reweighting2024.05 | 70.05 | |
| EKFACDefense Protocol=Relabeling2024.05 | 70.04 | |
| Self-LiSSADefense Protocol=Relabeling2024.05 | 69.18 | |
| Self-TracInDefense Protocol=Relabeling2024.05 | 68.31 | |
| PostDefense Protocol=Relabeling2024.05 | 66.18 | |
| PostDefense Protocol=Reweighting2024.05 | 66.18 |