List your company

Utility Evaluation on MATH500

93.6Pass@1 AccuracyRealSafe-R1
30.26446.70763.1579.593Aug 6, 2025

Evaluation results

MethodLinks
RealSafe-R1Backbone=R1-Qwen-7B2025.0893.6
SAFEPATH-FTBackbone=R1-Qwen-7B2025.0893.6
SafeKeyBackbone=R1-Qwen-7B2025.0893.6
No DefenseBackbone=R1-Qwen-7B2025.0892.9
SafeDecodingBackbone=R1-Qwen-7B2025.0891.8
SAFEPATH-FTBackbone=R1-Llama-8B2025.0890.9
SAFEPATH-ZSBackbone=R1-Qwen-7B2025.0890
RealSafe-R1Backbone=R1-Llama-8B2025.0886.4
ReasoningGuardBackbone=R1-Llama-8B2025.0886.4
ParaphraseBackbone=R1-Qwen-7B2025.0885.5
No DefenseBackbone=R1-Llama-8B2025.0885.5
ThinkingIBackbone=R1-Llama-8B2025.0885.5
ReasoningGuardBackbone=R1-Qwen-7B2025.0884.5
ThinkingIBackbone=R1-Qwen-7B2025.0883.6
SafeKeyBackbone=R1-Llama-8B2025.0883.6
SafeDecodingBackbone=R1-Llama-8B2025.0881.8
ParaphraseBackbone=R1-Llama-8B2025.0881.8
SAFEPATH-ZSBackbone=R1-Llama-8B2025.0877.3
Self-ReminderBackbone=R1-Qwen-7B2025.0874.5
Self-ReminderBackbone=R1-Llama-8B2025.0871.8
SmoothLLMBackbone=R1-Qwen-7B2025.0852.7
SmoothLLMBackbone=R1-Llama-8B2025.0832.7