List your company

Explanation Faithfulness on Jailbreaking DAN (test)

24Label Change RateICL
21.5638.0354.570.97Mar 31, 2026

Evaluation results

MethodLinks
ICLDeletion Ratio=10%2026.0324
ICLDeletion Ratio=20%2026.0327
ShapleyDeletion Ratio=10%2026.0333
ShapleyDeletion Ratio=20%2026.0333
LIMEDeletion Ratio=20%2026.0338
LIMEDeletion Ratio=10%2026.0354
EnsembleSHAPDeletion Ratio=20%2026.0374
EnsembleSHAPDeletion Ratio=10%2026.0385