List your company

Prompt Injection Detection on NeuralExec (test)

98.8Detection Accuracy (Top-1)CONTEXTCITE
84.44888.17491.995.626Sep 1, 2024

Evaluation results

MethodLinks
CONTEXTCITEModel=Llama3-8B, Attack Type=Optimization-based2024.0998.8100
CONTEXTCITEModel=Phi-3-mini, Attack Type=Handcrafted2024.0990100
CONTEXTCITEModel=Phi-3-mini, Attack Type=Optimization-based2024.098585