Harmful Content Detection on PHEME New Attacks: ExplainDrive (test)
82.91AccuracyLLM-SGA/ARHOCD
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| LLM-SGA/ARHOCDapproximation=Monte Carlo2025.12 | 82.91 | 81.67 | 82.61 | 82.03 | 13.16 | |
| LLM-SGA/ARHOCDapproximation=Lower-bound2025.12 | 82.88 | 81.64 | 82.58 | 82 | 13.24 | |
| LLM-purimode=Enhanced2025.12 | 82.41 | 81.41 | 80.65 | 81.02 | 14.24 | |
| MRATmode=Enhanced2025.12 | 82.05 | 81.22 | 79.91 | 80.44 | 15.38 | |
| ATmode=Enhanced2025.12 | 80.57 | 79.63 | 78.19 | 78.76 | 16.29 | |
| Det&Resmode=Enhanced2025.12 | 80.05 | 78.93 | 77.89 | 78.32 | 16.27 | |
| EnsSelmode=Enhanced2025.12 | 79.9 | 79.68 | 76.42 | 77.43 | 18.44 | |
| OutRegmode=Enhanced2025.12 | 79.88 | 79.36 | 76.71 | 77.59 | 17.52 | |
| ARTextmode=Enhanced2025.12 | 79.16 | 77.74 | 77.82 | 77.77 | 15.11 | |
| ARDELmode=Enhanced2025.12 | 79.04 | 78.29 | 75.94 | 76.73 | 16.29 | |
| FTMLmode=Enhanced2025.12 | 78.13 | 76.98 | 75.4 | 75.99 | 19.2 | |
| SAFERmode=Enhanced2025.12 | 77.27 | 75.78 | 75.2 | 75.45 | 17.54 | |
| MASKFilmode=Enhanced2025.12 | 77.15 | 75.77 | 74.61 | 75.06 | 18.69 | |
| FATmode=Enhanced2025.12 | 73.59 | 73.32 | 68.17 | 68.94 | 26.42 | |
| FIMmode=Enhanced2025.12 | 73.51 | 73.05 | 68.21 | 68.98 | 25.36 |