Safety Evaluation on XSTest (F1)
97F1 ScoreSafeKey
Evaluation Results
| Method | Links | |
|---|---|---|
| SafeKeyModel=R1-Llama-8B, Output Component=Final Answer2025.08 | 97 | |
| ReasoningGuardModel=R1-Llama-8B, Output Component=Reasoning2025.08 | 97 | |
| SafeKeyModel=R1-Llama-8B, Output Component=Reasoning2025.08 | 95 | |
| Self-ReminderModel=R1-Qwen-7B, Output Component=Reasoning2025.08 | 94 | |
| Self-ReminderModel=R1-Llama-8B, Output Component=Reasoning2025.08 | 93 | |
| SafeKeyModel=R1-Qwen-7B, Output Component=Reasoning2025.08 | 92 | |
| Self-ReminderModel=R1-Qwen-7B, Output Component=Final Answer2025.08 | 89 | |
| SafeDecodingModel=R1-Llama-8B, Output Component=Reasoning2025.08 | 88 | |
| RealSafe-R1Model=R1-Llama-8B, Output Component=Reasoning2025.08 | 88 | |
| SafeDecodingModel=R1-Qwen-7B, Output Component=Reasoning2025.08 | 87 | |
| SafeKeyModel=R1-Qwen-7B, Output Component=Final Answer2025.08 | 87 | |
| RealSafe-R1Model=R1-Qwen-7B, Output Component=Reasoning2025.08 | 86 | |
| No DefenseModel=R1-Llama-8B, Output Component=Reasoning2025.08 | 86 | |
| SAFEPATH-ZSModel=R1-Qwen-7B, Output Component=Reasoning2025.08 | 85 | |
| Self-ReminderModel=R1-Llama-8B, Output Component=Final Answer2025.08 | 85 | |
| ReasoningGuardModel=R1-Llama-8B, Output Component=Final Answer2025.08 | 85 | |
| No DefenseModel=R1-Qwen-7B, Output Component=Reasoning2025.08 | 84 | |
| ParaphraseModel=R1-Qwen-7B, Output Component=Reasoning2025.08 | 84 | |
| ParaphraseModel=R1-Llama-8B, Output Component=Reasoning2025.08 | 83 | |
| No DefenseModel=R1-Qwen-7B, Output Component=Final Answer2025.08 | 81 | |
| RealSafe-R1Model=R1-Llama-8B, Output Component=Final Answer2025.08 | 81 | |
| SafeDecodingModel=R1-Qwen-7B, Output Component=Final Answer2025.08 | 80 | |
| SAFEPATH-ZSModel=R1-Qwen-7B, Output Component=Final Answer2025.08 | 80 | |
| ReasoningGuardModel=R1-Qwen-7B, Output Component=Reasoning2025.08 | 80 | |
| SafeDecodingModel=R1-Llama-8B, Output Component=Final Answer2025.08 | 80 | |
| SAFEPATH-ZSModel=R1-Llama-8B, Output Component=Final Answer2025.08 | 80 | |
| SmoothLLMModel=R1-Qwen-7B, Output Component=Reasoning2025.08 | 79 | |
| SmoothLLMModel=R1-Llama-8B, Output Component=Reasoning2025.08 | 79 | |
| No DefenseModel=R1-Llama-8B, Output Component=Final Answer2025.08 | 78 | |
| ParaphraseModel=R1-Llama-8B, Output Component=Final Answer2025.08 | 78 | |
| ParaphraseModel=R1-Qwen-7B, Output Component=Final Answer2025.08 | 77 | |
| SmoothLLMModel=R1-Qwen-7B, Output Component=Final Answer2025.08 | 77 | |
| SAFEPATH-ZSModel=R1-Llama-8B, Output Component=Reasoning2025.08 | 77 | |
| SmoothLLMModel=R1-Llama-8B, Output Component=Final Answer2025.08 | 76 | |
| ThinkingIModel=R1-Llama-8B, Output Component=Final Answer2025.08 | 74 | |
| SAFEPATH-FTModel=R1-Qwen-7B, Output Component=Final Answer2025.08 | 73 | |
| ReasoningGuardModel=R1-Qwen-7B, Output Component=Final Answer2025.08 | 64 | |
| RealSafe-R1Model=R1-Qwen-7B, Output Component=Final Answer2025.08 | 63 | |
| SAFEPATH-FTModel=R1-Qwen-7B, Output Component=Reasoning2025.08 | 62 | |
| ThinkingIModel=R1-Qwen-7B, Output Component=Final Answer2025.08 | 58 | |
| SAFEPATH-FTModel=R1-Llama-8B, Output Component=Final Answer2025.08 | 56 | |
| SAFEPATH-FTModel=R1-Llama-8B, Output Component=Reasoning2025.08 | 50 | |
| ThinkingIModel=R1-Qwen-7B, Output Component=Reasoning2025.08 | 21 | |
| ThinkingIModel=R1-Llama-8B, Output Component=Reasoning2025.08 | 12 |