Toxicity Auditing on RealToxicityPrompts (hold-out)
4.97Detoxify Identity Attack Score
Text-only Baseline
Updated 2mo ago
Evaluation Results
| Method | Links | ||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
2025.03 | 4.97 | 60.31 | 2.53 | 35.27 | 2.53 | 70.35 | 70.54 | 6.74 | 61.38 | 5.43 | 21.36 | 3.84 | 68.22 | 70 | |
2025.03 | 5.44 | 66.22 | 3 | 40.15 | 3.47 | 77.39 | 77.57 | 7.59 | 66.63 | 6.56 | 23.14 | 5.34 | 75.16 | 77.6 | |
2025.03 | 5.53 | 66.32 | 3.28 | 40.52 | 4.69 | 78.42 | 78.7 | 7.31 | 66.91 | 6.46 | 23.33 | 6.84 | 76.1 | 78.16 | |
2025.03 | 5.9 | 66.69 | 3.37 | 39.77 | 3.37 | 78.23 | 78.33 | 8.15 | 67.29 | 7.12 | 23.89 | 5.62 | 75.35 | 77.5 | |
2025.03 | 6 | 65.19 | 3.09 | 39.02 | 4.12 | 76.54 | 76.64 | 7.87 | 65.04 | 5.99 | 22.68 | 5.9 | 73.19 | 75.63 | |
2025.03 | 6 | 67.44 | 3.09 | 42.02 | 4.22 | 79.36 | 79.45 | 8.05 | 68.13 | 7.12 | 24.36 | 5.9 | 77.31 | 79.28 | |
2025.03 | 6.09 | 66.69 | 3.18 | 40.43 | 3.84 | 79.36 | 79.36 | 7.68 | 66.91 | 6.65 | 23.43 | 5.99 | 76.47 | 78.81 |