Jailbreak Attack Transferability on Gemma-7b-it finetuned variants v1 (test)
65.4TSRPGP
Evaluation Results
| Method | Links | |
|---|---|---|
| PGPAttacker Knowledge=fully black-box2025.12 | 65.4 | |
| AutoDanAttacker Knowledge=white-box2025.12 | 63.4 | |
| TUJAAttacker Knowledge=pretrain–to–finetune setting2025.12 | 41.2 | |
| Guiding-GCGAttacker Knowledge=pretrain–to–finetune setting2025.12 | 39.4 | |
| SCAVAttacker Knowledge=pretrain–to–finetune setting2025.12 | 37.2 | |
| GCGAttacker Knowledge=white-box2025.12 | 36 | |
| AutoDanAttacker Knowledge=pretrain–to–finetune setting2025.12 | 35 | |
| PAIRAttacker Knowledge=fully black-box2025.12 | 34.4 | |
| SEAAttacker Knowledge=pretrain–to–finetune setting2025.12 | 25.6 | |
| DI-GCGAttacker Knowledge=pretrain–to–finetune setting2025.12 | 14 | |
| GCGAttacker Knowledge=pretrain–to–finetune setting2025.12 | 12.6 | |
| PIFAttacker Knowledge=pretrain–to–finetune setting2025.12 | 11.6 | |
| GCG EnsembleAttacker Knowledge=fully black-box2025.12 | 9 | |
| DANAttacker Knowledge=fully black-box2025.12 | 6.4 | |
| LSGM_LILAAttacker Knowledge=pretrain–to–finetune setting2025.12 | 4.4 | |
| L4AAttacker Knowledge=pretrain–to–finetune setting2025.12 | 0 |