Text Classification on OLID (test)
1.79Delta CACCAttDef w/o ELECTRA
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| AttDef w/o ELECTRAAttacks=BadNLm2023.05 | 1.79 | — | — | 8.21 | |
| AttDef w/o ELECTRAAttacks=Benign2023.05 | 1.69 | — | — | — | |
| AttDefAttacks=BadNLm2023.05 | 1.56 | — | — | 10.99 | |
| AttDef w/o ELECTRAAttacks=BadNLl2023.05 | 1.47 | — | — | 20.19 | |
| AttDefAttacks=InSent2023.05 | 1.47 | — | — | 30.47 | |
| AttDefAttacks=Benign2023.05 | 1.34 | — | — | — | |
| AttDef w/o ELECTRAAttacks=Avg2023.05 | 1.25 | — | — | 22.54 | |
| AttDef w/o ELECTRAAttacks=BadNLh2023.05 | 1.21 | — | — | 38.68 | |
| AttDefAttacks=Avg2023.05 | 1.18 | — | — | 24.37 | |
| ONIONAttacks=BadNLh2023.05 | 1.17 | — | — | 68.56 | |
| ONIONAttacks=Benign2023.05 | 0.93 | — | — | — | |
| AttDefAttacks=BadNLh2023.05 | 0.86 | — | — | 35.28 | |
| AttDefAttacks=BadNLl2023.05 | 0.67 | — | — | 20.74 | |
| ONIONAttacks=BadNLm2023.05 | 0.56 | — | — | 77.16 | |
| ONIONAttacks=Avg2023.05 | 0.54 | — | — | 63.5 | |
| AttDef w/o ELECTRAAttacks=InSent2023.05 | 0.23 | — | — | 23.07 | |
| ONIONAttacks=BadNLl2023.05 | 0.21 | — | — | 63.13 | |
| ONIONAttacks=InSent2023.05 | 0.21 | — | — | 45.17 | |
| Poisoned ModelAttacks=Benign2023.05 | — | — | 81.82 | — | |
| Poisoned ModelAttacks=BadNLl2023.05 | — | 100 | 81.23 | — | |
| Poisoned ModelAttacks=BadNLm2023.05 | — | 100 | 81.3 | — | |
| Poisoned ModelAttacks=BadNLh2023.05 | — | 97.19 | 81.42 | — | |
| Poisoned ModelAttacks=InSent2023.05 | — | 100 | 80.91 | — | |
| Poisoned ModelAttacks=Avg2023.05 | — | 99.31 | 81.22 | — |