Explanation Faithfulness on Jailbreaking GCG (test)
0Rate of Label ChangesICL
Evaluation Results
| Method | Links | |
|---|---|---|
| ICLDeletion Ratio=10%2026.03 | 0 | |
| ICLDeletion Ratio=20%2026.03 | 0 | |
| ShapleyDeletion Ratio=10%2026.03 | 11 | |
| LIMEDeletion Ratio=10%2026.03 | 15 | |
| EnsembleSHAPDeletion Ratio=10%2026.03 | 15 | |
| ShapleyDeletion Ratio=20%2026.03 | 19 | |
| LIMEDeletion Ratio=20%2026.03 | 23 | |
| EnsembleSHAPDeletion Ratio=20%2026.03 | 24 |