Text Adversarial Example Detection on Yelp
100TPR@10WDR
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| WDRAttack Method=TF-adj2026.03 | 100 | 96.2 | 96.6 | |
| GradMaskAttack Method=TF-adj2026.03 | 100 | 95.8 | 98.2 | |
| RTD-GuardAttack Method=TF-adj2026.03 | 100 | 97.1 | 98.9 | |
| RTD-GuardAttack Method=TextFooler2026.03 | 99.1 | 95.8 | 98.5 | |
| RTD-GuardAttack Method=BAE2026.03 | 98.9 | 96 | 98.3 | |
| GradMaskAttack Method=TextFooler2026.03 | 98.8 | 95.7 | 97.9 | |
| WDRAttack Method=TextFooler2026.03 | 98.2 | 95.6 | 97.5 | |
| RTD-GuardAttack Method=PWWS2026.03 | 97.6 | 95.2 | 98.6 | |
| GradMaskAttack Method=BAE2026.03 | 96.7 | 94.2 | 97.4 | |
| WDRAttack Method=BAE2026.03 | 96.6 | 94.8 | 97.2 | |
| RDEAttack Method=TextFooler2026.03 | 95.8 | 93.1 | 96.3 | |
| GradMaskAttack Method=PWWS2026.03 | 94.8 | 92.8 | 97.3 | |
| WDRAttack Method=PWWS2026.03 | 94.5 | 92.7 | 96.7 | |
| RDEAttack Method=TF-adj2026.03 | 92.1 | 91.4 | 94.9 | |
| RDEAttack Method=BAE2026.03 | 90.4 | 90.3 | 95.3 | |
| FGWSAttack Method=PWWS2026.03 | 88.9 | 89.5 | 91 | |
| FGWSAttack Method=TextFooler2026.03 | 84.7 | 87.3 | 87.6 | |
| RDEAttack Method=PWWS2026.03 | 83.8 | 86.5 | 94.4 | |
| FGWSAttack Method=TF-adj2026.03 | 79.7 | 85.5 | 82.3 | |
| FGWSAttack Method=BAE2026.03 | 62.4 | 72.5 | 70.3 | |
| MLEAttack Method=PWWS2026.03 | 41.2 | 54.5 | 66.3 | |
| MLEAttack Method=BAE2026.03 | 41.2 | 54.5 | 66.3 | |
| PPLAttack Method=TextFooler2026.03 | 39.8 | 53.1 | 76.8 | |
| PPLAttack Method=PWWS2026.03 | 37.6 | 51 | 74.8 | |
| MLEAttack Method=TextFooler2026.03 | 33.8 | 47.1 | 65.1 | |
| MLEAttack Method=TF-adj2026.03 | 27.8 | 40.4 | 66 | |
| PPLAttack Method=BAE2026.03 | 17.8 | 27.8 | 66.5 | |
| PPLAttack Method=TF-adj2026.03 | 13.9 | 22.5 | 62.7 |