Text Classification on SST-2 (Harmful Score Evaluation)
55.7Harmful ScoreNA-SFT
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| NA-SFTPoison Ratio (p)=0.2, Fine-tuning Sample Number (n)=50002024.05 | 55.7 | 95.64 | |
| NA-SFTPoison Ratio (p)=0.1, Fine-tuning Sample Number (n)=50002024.05 | 55 | 95.41 | |
| NA-SFTPoison Ratio (p)=0.05, Fine-tuning Sample Number (n)=50002024.05 | 54.9 | 95.07 | |
| SFTPoison Ratio (p)=0.3, Fine-tuning Sample Number (n)=50002024.05 | 53.6 | 95.32 | |
| NA-SFTPoison Ratio (p)=0.3, Fine-tuning Sample Number (n)=50002024.05 | 53.5 | 95.07 | |
| Vaccine-SFTPoison Ratio (p)=0.2, Fine-tuning Sample Number (n)=50002024.05 | 53.5 | 94.5 | |
| Vaccine-SFTPoison Ratio (p)=0.3, Fine-tuning Sample Number (n)=50002024.05 | 53 | 94.38 | |
| Vaccine-SFTPoison Ratio (p)=0.1, Fine-tuning Sample Number (n)=50002024.05 | 52.7 | 94.27 | |
| SFTPoison Ratio (p)=0.2, Fine-tuning Sample Number (n)=50002024.05 | 52.5 | 95.3 | |
| SFTPoison Ratio (p)=0.1, Fine-tuning Sample Number (n)=50002024.05 | 51.6 | 95.76 | |
| BSOPoison Ratio (p)=0.2, Fine-tuning Sample Number (n)=50002024.05 | 50.7 | 95.3 | |
| BSOPoison Ratio (p)=0.3, Fine-tuning Sample Number (n)=50002024.05 | 50.6 | 95.18 | |
| VlguardPoison Ratio (p)=0.3, Fine-tuning Sample Number (n)=50002024.05 | 49.3 | 95.53 | |
| SFTPoison Ratio (p)=0.05, Fine-tuning Sample Number (n)=50002024.05 | 49.1 | 94.95 | |
| BSOPoison Ratio (p)=0.1, Fine-tuning Sample Number (n)=50002024.05 | 49 | 95.18 | |
| Vaccine-SFTPoison Ratio (p)=0.05, Fine-tuning Sample Number (n)=50002024.05 | 48.5 | 93.92 | |
| VlguardPoison Ratio (p)=0.2, Fine-tuning Sample Number (n)=50002024.05 | 47.4 | 95.64 | |
| EWCPoison Ratio (p)=0.3, Fine-tuning Sample Number (n)=50002024.05 | 46.5 | 11.81 | |
| EWCPoison Ratio (p)=0.2, Fine-tuning Sample Number (n)=50002024.05 | 46.4 | 8.72 | |
| BSOPoison Ratio (p)=0.05, Fine-tuning Sample Number (n)=50002024.05 | 46 | 94.72 | |
| VlguardPoison Ratio (p)=0.1, Fine-tuning Sample Number (n)=50002024.05 | 43.9 | 95.18 | |
| LisaPoison Ratio (p)=0.3, Fine-tuning Sample Number (n)=50002024.05 | 43.6 | 94.04 | |
| LisaPoison Ratio (p)=0.2, Fine-tuning Sample Number (n)=50002024.05 | 42.6 | 94.61 | |
| VlguardPoison Ratio (p)=0.05, Fine-tuning Sample Number (n)=50002024.05 | 42 | 94.72 | |
| EWCPoison Ratio (p)=0.1, Fine-tuning Sample Number (n)=50002024.05 | 41.8 | 11.58 | |
| EWCPoison Ratio (p)=0.05, Fine-tuning Sample Number (n)=50002024.05 | 41.7 | 13.88 | |
| LisaPoison Ratio (p)=0.1, Fine-tuning Sample Number (n)=50002024.05 | 40.2 | 94.84 | |
| EWCPoison Ratio (p)=clean, Fine-tuning Sample Number (n)=50002024.05 | 38.3 | 45.18 | |
| LisaPoison Ratio (p)=0.05, Fine-tuning Sample Number (n)=50002024.05 | 36.6 | 95.18 | |
| LisaPoison Ratio (p)=clean, Fine-tuning Sample Number (n)=50002024.05 | 34.9 | 95.07 | |
| SFTPoison Ratio (p)=clean, Fine-tuning Sample Number (n)=50002024.05 | 34.6 | 95.3 | |
| BSOPoison Ratio (p)=clean, Fine-tuning Sample Number (n)=50002024.05 | 34.3 | 95.53 | |
| VlguardPoison Ratio (p)=clean, Fine-tuning Sample Number (n)=50002024.05 | 33.8 | 95.64 | |
| Vaccine-SFTPoison Ratio (p)=clean, Fine-tuning Sample Number (n)=50002024.05 | 26.6 | 95.3 | |
| NA-SFTPoison Ratio (p)=clean, Fine-tuning Sample Number (n)=50002024.05 | 24.1 | 94.84 |