List your company

Prompt Injection on URL-based PI (200-sample dataset)

33.5ASRStandard
-1.347.70516.7525.795Feb 5, 2026

Evaluation results

MethodLinks
StandardModel=Grok 4 Fast2026.0233.5
StandardModel=Llama 42026.0222
StandardModel=GPT-52026.0220
AdvancedModel=Llama 42026.0216
InjectDefuserModel=Grok 4 Fast2026.028
InjectDefuserModel=Llama 42026.027
AdvancedModel=Grok 4 Fast2026.022
AdvancedModel=GPT-52026.021
StandardModel=Gemma 32026.020
AdvancedModel=Gemma 32026.020
InjectDefuserModel=GPT-52026.020
InjectDefuserModel=Gemma 32026.020