Harmful Content Detection on PHEME Known Attacks: DeepWordBug, TFAdjusted, TREPAT (test)
85.59AccuracyLLM-SGA/ARHOCD
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| LLM-SGA/ARHOCDapproximation=Monte Carlo2025.12 | 85.59 | 84.42 | 85.54 | 84.85 | 10.59 | |
| LLM-SGA/ARHOCDapproximation=Lower-bound2025.12 | 85.49 | 84.32 | 85.45 | 84.76 | 10.59 | |
| LLM-purimode=Enhanced2025.12 | 83.66 | 82.73 | 82.09 | 82.38 | 12.55 | |
| MRATmode=Enhanced2025.12 | 83.54 | 82.74 | 81.71 | 82.15 | 13.96 | |
| EnsSelmode=Enhanced2025.12 | 81.81 | 81.32 | 79.13 | 79.93 | 15.35 | |
| ATmode=Enhanced2025.12 | 81.59 | 80.82 | 79.23 | 79.85 | 15.03 | |
| Det&Resmode=Enhanced2025.12 | 81.38 | 80.42 | 79.27 | 79.74 | 14.99 | |
| OutRegmode=Enhanced2025.12 | 81.15 | 80.6 | 78.36 | 79.16 | 15.26 | |
| ARTextmode=Enhanced2025.12 | 80.95 | 79.62 | 79.82 | 79.71 | 12.79 | |
| ARDELmode=Enhanced2025.12 | 80.78 | 80.1 | 78.08 | 78.81 | 15.03 |