Adversarial Detection on AG News
95.7F1 ScoreWDR
Evaluation results
| Method | Links | ||
|---|---|---|---|
| WDRModel=DistilBERT, Attack=TextFooler, Detector Trained On=DistilBERT, AG News, PWWS2022.04 | 95.7 | 97.3 | |
| WDRModel=LSTM, Attack=PWWS, Detector Trained On=DistilBERT, AG News, PWWS2022.04 | 94 | 94.2 | |
| WDRModel=DistilBERT, Attack=PWWS, Detector Trained On=DistilBERT, AG News, PWWS2022.04 | 93.6 | 94.8 | |
| WDRModel=DistilBERT, Attack=PWWS2022.04 | 93.1 | 96.1 | |
| WDRModel=BERT, Attack=PWWS, Detector Trained On=DistilBERT, AG News, PWWS2022.04 | 92.5 | 93 | |
| WDRModel=CNN, Attack=PWWS, Detector Trained On=DistilBERT, AG News, PWWS2022.04 | 91.1 | 91.2 | |
| FGWSModel=CNN, Attack=PWWS, Detector Trained On=DistilBERT, AG News, PWWS2022.04 | 90.6 | 87.6 | |
| FGWSModel=DistilBERT, Attack=PWWS, Detector Trained On=DistilBERT, AG News, PWWS2022.04 | 89.5 | 84.6 | |
| FGWSModel=DistilBERT, Attack=PWWS2022.04 | 89.5 | 84.6 | |
| FGWSModel=LSTM, Attack=PWWS, Detector Trained On=DistilBERT, AG News, PWWS2022.04 | 88.9 | 84.9 | |
| FGWSModel=BERT, Attack=PWWS, Detector Trained On=DistilBERT, AG News, PWWS2022.04 | 88.7 | 83.2 | |
| FGWSModel=DistilBERT, Attack=TextFooler, Detector Trained On=DistilBERT, AG News, PWWS2022.04 | 87 | 79.4 | |
| WDRModel=DistilBERT, Attack=IGA, Detector Trained On=DistilBERT, AG News, PWWS2022.04 | 86.7 | 93.6 | |
| WDRModel=DistilBERT, Attack=BAE, Detector Trained On=DistilBERT, AG News, PWWS2022.04 | 86.4 | 94.5 | |
| WDRModel=DistilBERT, Attack=BAE2022.04 | 81 | 95.4 | |
| FGWSModel=DistilBERT, Attack=IGA, Detector Trained On=DistilBERT, AG News, PWWS2022.04 | 68.6 | 58.3 | |
| FGWSModel=DistilBERT, Attack=BAE, Detector Trained On=DistilBERT, AG News, PWWS2022.04 | 55.8 | 44 | |
| FGWSModel=DistilBERT, Attack=BAE2022.04 | 55.8 | 44 |