List your company

Utility Evaluation on MMLU (pass@1 Accuracy)

78Accuracy (pass@1)SafeDecoding
28.0841.045466.96Aug 6, 2025

Evaluation results

MethodLinks
SafeDecodingBackbone=R1-Llama-8B2025.0878
ReasoningGuardBackbone=R1-Llama-8B2025.0874
No DefenseBackbone=R1-Llama-8B2025.0873
SafeKeyBackbone=R1-Llama-8B2025.0873
Self-ReminderBackbone=R1-Llama-8B2025.0873
ThinkingIBackbone=R1-Llama-8B2025.0873
SAFEPATH-FTBackbone=R1-Llama-8B2025.0872
No DefenseBackbone=R1-Qwen-7B2025.0869
SafeDecodingBackbone=R1-Qwen-7B2025.0869
RealSafe-R1Backbone=R1-Qwen-7B2025.0869
SAFEPATH-ZSBackbone=R1-Llama-8B2025.0869
SafeKeyBackbone=R1-Qwen-7B2025.0867
Self-ReminderBackbone=R1-Qwen-7B2025.0867
ThinkingIBackbone=R1-Qwen-7B2025.0866
RealSafe-R1Backbone=R1-Llama-8B2025.0866
ReasoningGuardBackbone=R1-Qwen-7B2025.0865
SAFEPATH-ZSBackbone=R1-Qwen-7B2025.0863
SAFEPATH-FTBackbone=R1-Qwen-7B2025.0859
SmoothLLMBackbone=R1-Qwen-7B2025.0843
SmoothLLMBackbone=R1-Llama-8B2025.0842
ParaphraseBackbone=R1-Llama-8B2025.0838
ParaphraseBackbone=R1-Qwen-7B2025.0830