Single-turn Jailbreak Attack on JailbreakBench
9ASRRolebreakPersona
Evaluation Results
| Method | Links | |
|---|---|---|
| RolebreakPersonaModel=AVG, Variant=PlanX2026.06 | 9 | |
| RolebreakPersonaModel=AVG, Variant=base2026.06 | 15 | |
| CrescendoCondensedModel=AVG, Variant=base2026.06 | 22 | |
| DaggerHappyEndingModel=AVG, Variant=base2026.06 | 27 | |
| DaggerHappyEndingModel=AVG, Variant=PlanX2026.06 | 30 | |
| CrescendoCondensedModel=AVG, Variant=PlanX2026.06 | 31 | |
| DeepInceptionModel=AVG, Variant=base2026.06 | 34 | |
| AdvPoetryModel=AVG, Variant=base2026.06 | 34 | |
| DeepInceptionModel=AVG, Variant=PlanX2026.06 | 36 | |
| AdvPoetryModel=AVG, Variant=PlanX2026.06 | 41 | |
| MimicryNarrativeModel=AVG, Variant=base2026.06 | 42 | |
| VernacularModel=Gemini-2.5-Pro, Variant=Mean2026.06 | 51 | |
| VernacularModel=DeepSeek-V4-Pro, Variant=Mean2026.06 | 58 | |
| VernacularModel=Llama-3.2-3B, Variant=Mean2026.06 | 66 | |
| VernacularModel=Gemma-4-E4B, Variant=Mean2026.06 | 70 | |
| MimicryNarrativeModel=AVG, Variant=PlanX2026.06 | 72 | |
| VernacularModel=AVG, Variant=Mean2026.06 | 72 | |
| VernacularModel=Gemma-4-E2B, Variant=Mean2026.06 | 73 | |
| VernacularModel=Llama-3.1-8B, Variant=Mean2026.06 | 80 | |
| VernacularModel=Qwen3.5-4B, Variant=Mean2026.06 | 89 | |
| VernacularModel=Qwen3.5-9B, Variant=Mean2026.06 | 90 |