RLHF Backdoor Attack on Anthropic Helpful Harmless prompts (train test)
28.1UHR Rate
Random
Updated 1d ago
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
2025.10 | 28.1 | 82.6 | 63.4 | |
2025.10 | 28.1 | 86.1 | 85.6 | |
2025.10 | 28 | 87.4 | 51.2 | |
2025.10 | 27.8 | 79.8 | 53.3 | |
2025.10 | 27.5 | 83.2 | 66.1 | |
2025.10 | 26.8 | 68 | 48 | |
2025.10 | 25.3 | 74.9 | 58.5 | |
2025.10 | 25 | 72 | 56.8 | |
2025.10 | 24.3 | 67.3 | 49.1 | |
2025.10 | 23.8 | 70.5 | 50.3 | |
2025.10 | 23.7 | 69.3 | 71.7 | |
2025.10 | 23.6 | 66.6 | 62.4 | |
2025.10 | 23.5 | 75.5 | 76 | |
2025.10 | 23.5 | 71 | 47.8 | |
2025.10 | 23.4 | 68.9 | 60.7 | |
2025.10 | 22.7 | 67.5 | 38.8 | |
2025.10 | 22.1 | 69 | 50 | |
2025.10 | 21.5 | 70 | 44 | |
2025.10 | 21.3 | 68 | 61.7 | |
2025.10 | 20.5 | 60.5 | 55.3 | |
2025.10 | 17.3 | 77.2 | 70.5 | |
2025.10 | 17 | 81.2 | 32 | |
2025.10 | 16.3 | 84.5 | 40.3 | |
2025.10 | 15.1 | 35 | 41 | |
2025.10 | 14.5 | 42 | 40.5 | |
2025.10 | 14.5 | 43 | 42.2 | |
2025.10 | 14 | 81.4 | 32 | |
2025.10 | 13.3 | 80 | 77.5 | |
2025.10 | 13.1 | 38.4 | 24 | |
2025.10 | 12 | 45 | 20 |