List your company

Hidden objective detection on 9 objectives out-of-domain code-manipulation A2

96.67Exit RateGemini-3-Flash
11.733233.784155.83577.8859May 12, 2026

Evaluation results

MethodLinks
Gemini-3-FlashGenerator (M)=Qwen3-14B, Attack Setting=A2 (code-manipulation), Evaluation Scope=CoT+Code2026.0596.6710010004.910063.8394.2298.8273.16
CoT-Guard 4BGenerator (M)=Qwen3-14B, Attack Setting=A2 (code-manipulation), Evaluation Scope=CoT+Code2026.0594.494.91001.829.410022.2217.3498.2262.03
Qwen3-32BGenerator (M)=Qwen3-14B, Attack Setting=A2 (code-manipulation), Evaluation Scope=CoT+Code2026.0591.6794.8799.619.099.7697.3810.1129.8849.4154.64
Qwen3-8BGenerator (M)=Qwen3-14B, Attack Setting=A2 (code-manipulation), Evaluation Scope=CoT+Code2026.0553.394.997.406.994.742.130.5850.344.47
GPT-5.4Generator (M)=Qwen3-14B, Attack Setting=A2 (code-manipulation), Evaluation Scope=CoT+Code2026.0523.3310098.50092.11000.5934.95
Qwen3-4BGenerator (M)=Qwen3-14B, Attack Setting=A2 (code-manipulation), Evaluation Scope=CoT+Code2026.051592.357.91.81.585.961.61.1615.9830.36
GPT-5-miniGenerator (M)=Qwen3-14B, Attack Setting=A2 (code-manipulation), Evaluation Scope=CoT+Code2026.051584.6275.640089.4701.732.9629.94