Backdoor Trigger Detection on OLID
0.43PrecisionAttDef
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| AttDefAttack=BadNLl2023.05 | 0.43 | 0.92 | |
| BFClassAttack=BadNLl2023.05 | 0.38 | 1 | |
| BFClassAttack=BadNLm2023.05 | 0.38 | 0.71 | |
| BFClassAttack=BadNLh2023.05 | 0.38 | 1 | |
| AttDefAttack=BadNLh2023.05 | 0.34 | 0.88 | |
| AttDefAttack=BadNLm2023.05 | 0.32 | 0.64 | |
| AttDefAttack=Avg2023.05 | 0.32 | 0.77 | |
| BFClassAttack=Avg2023.05 | 0.31 | 0.73 | |
| AttDefAttack=InSent2023.05 | 0.19 | 0.64 | |
| BFClassAttack=InSent2023.05 | 0.11 | 0.2 |