List your company

Attack Success Rate Evaluation on SKILL-INJECT Obvious

93ASRSkillAttack
-0.623.74872.3Apr 5, 2026

Evaluation results

MethodLinks
SkillAttackModel=kimi-k2.52026.0493
SkillAttackModel=doubao-seed-1-8-2512282026.0490
SkillAttackModel=hunyuan-2.0-instruct-202511112026.0490
SkillAttackModel=gpt-5.42026.0487
SkillAttackModel=gemini-3.0-pro-preview2026.0487
SkillAttackModel=gpt-5.4-nano2026.0483
SkillAttackModel=MiniMax-M2.52026.0483
SkillAttackModel=claude-sonnet-4-5-202509292026.0477
SkillAttackModel=glm-52026.0477
SkillAttackModel=Qwen3.5-Plus2026.0473
SKILL-INJECTModel=hunyuan-2.0-instruct-202511112026.0443
SKILL-INJECTModel=doubao-seed-1-8-2512282026.0440
SKILL-INJECTModel=gpt-5.4-nano2026.0427
SKILL-INJECTModel=kimi-k2.52026.0427
SKILL-INJECTModel=MiniMax-M2.52026.0427
SKILL-INJECTModel=gemini-3.0-pro-preview2026.0420
SKILL-INJECTModel=Qwen3.5-Plus2026.0420
SKILL-INJECTModel=glm-52026.0420
SKILL-INJECTModel=claude-sonnet-4-5-202509292026.0417
SKILL-INJECTModel=gpt-5.42026.0413
Direct AttackModel=gemini-3.0-pro-preview2026.0413
Direct AttackModel=doubao-seed-1-8-2512282026.0413
Direct AttackModel=hunyuan-2.0-instruct-202511112026.0413
Direct AttackModel=gpt-5.4-nano2026.0410
Direct AttackModel=Qwen3.5-Plus2026.0410
Direct AttackModel=glm-52026.0410
Direct AttackModel=kimi-k2.52026.047
Direct AttackModel=MiniMax-M2.52026.047
Direct AttackModel=gpt-5.42026.043
Direct AttackModel=claude-sonnet-4-5-202509292026.043