Jailbreak Attack Transferability on Llama-3-8b-Instruct finetuned variants v1 (test)
51.2TSRPGP
Evaluation Results
| Method | Links | |
|---|---|---|
| PGPAttacker Knowledge=fully black-box2025.12 | 51.2 | |
| AutoDanAttacker Knowledge=white-box2025.12 | 48.2 | |
| GCGAttacker Knowledge=white-box2025.12 | 38.8 | |
| SCAVAttacker Knowledge=pretrain–to–finetune setting2025.12 | 31.8 | |
| Guiding-GCGAttacker Knowledge=pretrain–to–finetune setting2025.12 | 27.4 | |
| TUJAAttacker Knowledge=pretrain–to–finetune setting2025.12 | 16.4 | |
| LSGM_LILAAttacker Knowledge=pretrain–to–finetune setting2025.12 | 13 | |
| GCGAttacker Knowledge=pretrain–to–finetune setting2025.12 | 12.4 | |
| AutoDanAttacker Knowledge=pretrain–to–finetune setting2025.12 | 12.2 | |
| SEAAttacker Knowledge=pretrain–to–finetune setting2025.12 | 10.8 | |
| DANAttacker Knowledge=fully black-box2025.12 | 8.6 | |
| DI-GCGAttacker Knowledge=pretrain–to–finetune setting2025.12 | 5.8 | |
| L4AAttacker Knowledge=pretrain–to–finetune setting2025.12 | 0.8 | |
| GCG EnsembleAttacker Knowledge=fully black-box2025.12 | 0.8 | |
| PIFAttacker Knowledge=pretrain–to–finetune setting2025.12 | 0 | |
| PAIRAttacker Knowledge=fully black-box2025.12 | 0 |