Toxicity Mitigation on REALTOXICITYPROMPTS
21.24Toxicity
EIGENSHIFT + Targeted Subspace Intervention
Updated 4d ago
Evaluation Results
| Method | Links | |
|---|---|---|
| 21.24 | ||
2026.02 | 24.49 | |
| 24.61 | ||
2026.02 | 24.64 | |
| 25.94 | ||
| 26.24 | ||
| 28.3 | ||
2026.02 | 28.73 | |
| 29.68 | ||
| 31.39 | ||
| 33.99 | ||
2026.02 | 34.6 | |
2026.02 | 34.78 | |
| 36.21 | ||
| 38.41 | ||
2026.02 | 39.29 | |
2026.02 | 42.33 | |
2026.02 | 44.82 | |
2026.02 | 50.39 | |
2026.02 | 50.89 | |
| 53.93 | ||
| 54.06 | ||
2026.02 | 54.25 | |
2026.02 | 54.9 |