Attack Success Rate Evaluation on SKILL-INJECT Obvious
93ASRSkillAttack
Evaluation Results
| Method | Links | |
|---|---|---|
| SkillAttackModel=kimi-k2.52026.04 | 93 | |
| SkillAttackModel=doubao-seed-1-8-2512282026.04 | 90 | |
| SkillAttackModel=hunyuan-2.0-instruct-202511112026.04 | 90 | |
| SkillAttackModel=gpt-5.42026.04 | 87 | |
| SkillAttackModel=gemini-3.0-pro-preview2026.04 | 87 | |
| SkillAttackModel=gpt-5.4-nano2026.04 | 83 | |
| SkillAttackModel=MiniMax-M2.52026.04 | 83 | |
| SkillAttackModel=claude-sonnet-4-5-202509292026.04 | 77 | |
| SkillAttackModel=glm-52026.04 | 77 | |
| SkillAttackModel=Qwen3.5-Plus2026.04 | 73 | |
| SKILL-INJECTModel=hunyuan-2.0-instruct-202511112026.04 | 43 | |
| SKILL-INJECTModel=doubao-seed-1-8-2512282026.04 | 40 | |
| SKILL-INJECTModel=gpt-5.4-nano2026.04 | 27 | |
| SKILL-INJECTModel=kimi-k2.52026.04 | 27 | |
| SKILL-INJECTModel=MiniMax-M2.52026.04 | 27 | |
| SKILL-INJECTModel=gemini-3.0-pro-preview2026.04 | 20 | |
| SKILL-INJECTModel=Qwen3.5-Plus2026.04 | 20 | |
| SKILL-INJECTModel=glm-52026.04 | 20 | |
| SKILL-INJECTModel=claude-sonnet-4-5-202509292026.04 | 17 | |
| SKILL-INJECTModel=gpt-5.42026.04 | 13 | |
| Direct AttackModel=gemini-3.0-pro-preview2026.04 | 13 | |
| Direct AttackModel=doubao-seed-1-8-2512282026.04 | 13 | |
| Direct AttackModel=hunyuan-2.0-instruct-202511112026.04 | 13 | |
| Direct AttackModel=gpt-5.4-nano2026.04 | 10 | |
| Direct AttackModel=Qwen3.5-Plus2026.04 | 10 | |
| Direct AttackModel=glm-52026.04 | 10 | |
| Direct AttackModel=kimi-k2.52026.04 | 7 | |
| Direct AttackModel=MiniMax-M2.52026.04 | 7 | |
| Direct AttackModel=gpt-5.42026.04 | 3 | |
| Direct AttackModel=claude-sonnet-4-5-202509292026.04 | 3 |