List your company

Jailbreak Attack on CommonsenseQA

1Attack Success Rate (ASR)H-COT
-2.7622.624873.38Apr 17, 2026

Evaluation results

MethodLinks
H-COTVictim LRM=Qwen2.5-Max2026.0412
H-COTVictim LRM=OpenAI o4-mini2026.0412
Cognitive OverloadVictim LRM=OpenAI o4-mini2026.04102.6
Cognitive OverloadVictim LRM=Qwen2.5-Max2026.04112.36
H-COTVictim LRM=DeepSeek R12026.04602.63
Cognitive OverloadVictim LRM=DeepSeek R12026.04722.58
PRJAVictim LRM=OpenAI o4-mini2026.04863.3
PRJAVictim LRM=DeepSeek R12026.04913.63
PRJAVictim LRM=Qwen2.5-Max2026.04953.42