List your company

Hidden Objective Monitoring on Prompt-Manipulation A1 Monitor Notice

100Exit Detection RateGemini-1.5-Flash
-4235077May 12, 2026

Evaluation results

MethodLinks
Gemini-1.5-FlashMonitor Scope=Code, Exact objective awareness=Aware τhid2026.0510010010020.1888.1596.8610010010089.47
Gemini-1.5-FlashMonitor Scope=CoT+Code2026.05100100100062.7598.79910010084.49
Gemini-1.5-FlashMonitor Scope=CoT+Code, Exact objective awareness=Aware τhid2026.0510010010049.9399.0298.7410010010094.19
CoT-Guard 4BMonitor Scope=Code2026.0592.8984.6298.441.6315.8599.1276.5463.4598.2270.08
CoT-Guard 4BMonitor Scope=CoT+Code2026.0592.1594.87100018.210068.999.298.2274.62
Gemini-1.5-FlashMonitor Scope=Code2026.0588.8210010001.9696.8678.7693.0698.9373.15
Qwen2-1.5BMonitor Scope=Code, Training Protocol=RL Only2026.0580.2692.3197.670.816.0498.1172.999.8688.9260.77
GPT-4oMonitor Scope=CoT+Code2026.0574.21001000083.811.78897.3361.67
Qwen2-1.5BMonitor Scope=CoT+Code, Training Protocol=RL Only2026.0562.5687.1899.612.4622.198.222.895.6597.6365.35
Qwen2-72BMonitor Scope=Code2026.055097.4489.5101.1784.281.6316.9714.8439.54
Qwen2-1.5BMonitor Scope=Code, Training Protocol=SFT Only2026.0538.8210098.1500.3989.311.31.079.0737.57
GPT-4o-miniMonitor Scope=CoT+Code2026.0536.189079.9300.5940.911.470.447.241.84
Qwen2-7BMonitor Scope=Code2026.0532.8997.4490.120.821.9686.161.620.2715.236.28
Qwen2-72BMonitor Scope=CoT+Code2026.0528.2994.8789.7701.3684.835.678.4876.6154.42
Llama3.1-8B-InstructMonitor Scope=Code2026.051938.336.17.71737.520.611.932.524.51
Qwen2-7BMonitor Scope=CoT+Code2026.0517.7685.3879.320.820.9847.20.759.250.9338.03
Qwen2-1.5BMonitor Scope=CoT+Code, Training Protocol=SFT Only2026.0512.7779.4954.921.732.163.81.755.9242.0634.94
Qwen2-1.5BMonitor Scope=Code2026.055.267017.760054.860.3201.0616.58
Qwen2-1.5BMonitor Scope=CoT+Code2026.053.297025.93008.802414.616.29
GPT-4o-miniMonitor Scope=Code2026.051.978526.540075.47000.2721.03
Llama3.1-8B-InstructMonitor Scope=CoT+Code2026.050.6000.90.41.40.41.82.40.88
GPT-4oMonitor Scope=Code2026.05010087.10083.3300030.05