List your company
ResearchFormerly WizWand

Model Efficiency on AdvBench

1.37ATGRSmoothLLM
0.98521.08511.1851.2849Aug 6, 2025
Updated 5mo ago

Evaluation results

MethodLinks
SmoothLLMBackbone=R1-Qwen-7B2025.081.37
SmoothLLMBackbone=R1-Llama-8B2025.081.35
ParaphraseBackbone=R1-Llama-8B2025.081.29
ParaphraseBackbone=R1-Qwen-7B2025.081.22
SafeDecodingBackbone=R1-Llama-8B2025.081.1
SafeDecodingBackbone=R1-Qwen-7B2025.081.09
ReasoningGuardBackbone=R1-Llama-8B2025.081.09
ReasoningGuardBackbone=R1-Qwen-7B2025.081.05
ThinkingIBackbone=R1-Qwen-7B2025.081.03
ThinkingIBackbone=R1-Llama-8B2025.081.02
SAFEPATH-ZSBackbone=R1-Qwen-7B2025.081.02
SAFEPATH-ZSBackbone=R1-Llama-8B2025.081.02
Self-ReminderBackbone=R1-Llama-8B2025.081.01
RealSafe-R1Backbone=R1-Qwen-7B2025.081
RealSafe-R1Backbone=R1-Llama-8B2025.081
SAFEPATH-FTBackbone=R1-Qwen-7B2025.081
SAFEPATH-FTBackbone=R1-Llama-8B2025.081
SafeKeyBackbone=R1-Qwen-7B2025.081
SafeKeyBackbone=R1-Llama-8B2025.081
Self-ReminderBackbone=R1-Qwen-7B2025.081