List your company

Agentic Attack on WhisperBench Background execution (held-out)

96.4ISRMEMGHOST
-3.85622.17248.274.228Jul 6, 2026

Evaluation results

MethodLinks
MEMGHOSTTarget Model=GPT-5.4, Evaluation Cases=562026.0796.410087.587.5
MEMGHOSTTarget Model=Grok 4.1 Fast, Evaluation Cases=562026.0796.410091.191.1
MEMGHOSTTarget Model=DeepSeek v4 Flash, Evaluation Cases=562026.0796.487.585.775
MEMGHOSTTarget Model=MiniMax-M2.5, Evaluation Cases=562026.0794.694.682.178.6
MEMGHOSTTarget Model=Sonnet 4.5, Evaluation Cases=562026.0792.994.682.180.4
MEMGHOSTTarget Model=Sonnet 4.6, Evaluation Cases=562026.0773.273.271.471.4
RGSTarget Model=GPT-5.4, Evaluation Cases=202026.0740504040
MAP-EliteTarget Model=GPT-5.4, Evaluation Cases=202026.07205200
Naive AttackTarget Model=GPT-5.4, Evaluation Cases=562026.070000
Escape CharactersTarget Model=GPT-5.4, Evaluation Cases=562026.0705.400
Context IgnoreTarget Model=GPT-5.4, Evaluation Cases=562026.0702500
Fake CompletionTarget Model=GPT-5.4, Evaluation Cases=562026.0705.400
Combined AttackTarget Model=GPT-5.4, Evaluation Cases=562026.07019.600