List your company

Phishing Detection on HTML-based PI (test)

0.3ASRInjectDefuser
-3.07619.71242.565.288Feb 5, 2026

Evaluation results

MethodLinks
InjectDefuserBackbone=GPT-5, Model Scale=Highest performance, Weight Visibility=closed-weight2026.020.3
AdvancedBackbone=GPT-5, Model Scale=Highest performance, Weight Visibility=closed-weight2026.0210.1
InjectDefuserBackbone=Grok 4 Fast, Model Scale=Medium-scale, Weight Visibility=closed-weight, Reasoning Capability=Non-Reasoning2026.0226.1
InjectDefuserBackbone=Gemma 3 27B, Model Scale=Lightweight, Weight Visibility=open-weight2026.0236.7
StandardBackbone=GPT-5, Model Scale=Highest performance, Weight Visibility=closed-weight2026.0239.9
AdvancedBackbone=Gemma 3 27B, Model Scale=Lightweight, Weight Visibility=open-weight2026.0255.4
InjectDefuserBackbone=Llama 4 Maverick, Model Scale=Medium-scale, Weight Visibility=open-weight2026.0261.7
StandardBackbone=Gemma 3 27B, Model Scale=Lightweight, Weight Visibility=open-weight2026.0264.7
StandardBackbone=Grok 4 Fast, Model Scale=Medium-scale, Weight Visibility=closed-weight, Reasoning Capability=Non-Reasoning2026.0265.1
AdvancedBackbone=Llama 4 Maverick, Model Scale=Medium-scale, Weight Visibility=open-weight2026.0275.6
AdvancedBackbone=Grok 4 Fast, Model Scale=Medium-scale, Weight Visibility=closed-weight, Reasoning Capability=Non-Reasoning2026.0276.2
StandardBackbone=Llama 4 Maverick, Model Scale=Medium-scale, Weight Visibility=open-weight2026.0284.7