Safety Evaluation on XSTest Safe
4FCReasoningGuard
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| ReasoningGuardModel=R1-Qwen-7B, Output Component=Reasoning2025.08 | 4 | 2 | 94 | |
| ThinkingIModel=R1-Llama-8B, Output Component=Reasoning2025.08 | 6 | 82 | 12 | |
| ThinkingIModel=R1-Qwen-7B, Output Component=Reasoning2025.08 | 8 | 58 | 34 | |
| SAFEPATH-FTModel=R1-Llama-8B, Output Component=Reasoning2025.08 | 28 | 50 | 22 | |
| RealSafe-R1Model=R1-Qwen-7B, Output Component=Reasoning2025.08 | 32 | 10 | 58 | |
| SAFEPATH-FTModel=R1-Qwen-7B, Output Component=Reasoning2025.08 | 32 | 28 | 40 | |
| ReasoningGuardModel=R1-Qwen-14B, Evaluation Component=Reasoning2025.08 | 32 | 0 | 68 | |
| RealSafe-R1Model=R1-Qwen-7B, Output Component=Final Answer2025.08 | 34 | 40 | 26 | |
| ReasoningGuardModel=R1-Qwen-32B, Evaluation Component=Reasoning2025.08 | 34 | 2 | 64 | |
| ReasoningGuardModel=R1-Llama-8B, Output Component=Reasoning2025.08 | 38 | 0 | 62 | |
| SAFEPATH-ZSModel=R1-Qwen-7B, Output Component=Reasoning2025.08 | 40 | 8 | 52 | |
| SAFEPATH-FTModel=R1-Llama-8B, Output Component=Final Answer2025.08 | 40 | 40 | 20 | |
| ThinkingIModel=R1-Qwen-7B, Output Component=Final Answer2025.08 | 42 | 50 | 8 | |
| SAFEPATH-ZSModel=R1-Llama-8B, Output Component=Reasoning2025.08 | 44 | 10 | 46 | |
| RealSafe-R1Model=R1-Llama-8B, Output Component=Reasoning2025.08 | 46 | 10 | 44 | |
| RealSafe-R1Model=R1-Llama-8B, Output Component=Final Answer2025.08 | 50 | 24 | 26 | |
| SmoothLLMModel=R1-Qwen-7B, Output Component=Reasoning2025.08 | 52 | 8 | 40 | |
| ReasoningGuardModel=R1-Qwen-7B, Output Component=Final Answer2025.08 | 54 | 36 | 10 | |
| SafeKeyModel=R1-Qwen-7B, Output Component=Reasoning2025.08 | 60 | 8 | 32 | |
| SafeKeyModel=R1-Llama-8B, Output Component=Reasoning2025.08 | 60 | 6 | 34 | |
| ThinkingIModel=R1-Llama-8B, Output Component=Final Answer2025.08 | 60 | 30 | 10 | |
| ThinkingIModel=R1-Qwen-32B, Evaluation Component=Reasoning2025.08 | 60 | 30 | 10 | |
| SAFEPATH-FTModel=R1-Qwen-7B, Output Component=Final Answer2025.08 | 62 | 30 | 8 | |
| Self-ReminderModel=R1-Qwen-7B, Output Component=Reasoning2025.08 | 62 | 0 | 38 | |
| SAFEPATH-ZSModel=R1-Qwen-14B, Evaluation Component=Reasoning2025.08 | 64 | 0 | 36 | |
| SmoothLLMModel=R1-Llama-8B, Output Component=Reasoning2025.08 | 66 | 4 | 30 | |
| RealSafe-R1Model=R1-Qwen-14B, Evaluation Component=Reasoning2025.08 | 68 | 2 | 30 | |
| SAFEPATH-ZSModel=R1-Qwen-7B, Output Component=Final Answer2025.08 | 70 | 14 | 16 | |
| SafeKeyModel=R1-Qwen-7B, Output Component=Final Answer2025.08 | 72 | 14 | 14 | |
| ThinkingIModel=R1-Qwen-14B, Evaluation Component=Reasoning2025.08 | 72 | 18 | 10 | |
| SAFEPATH-ZSModel=R1-Qwen-32B, Evaluation Component=Reasoning2025.08 | 72 | 4 | 24 | |
| ParaphraseModel=R1-Qwen-7B, Output Component=Reasoning2025.08 | 74 | 4 | 22 | |
| SmoothLLMModel=R1-Qwen-7B, Output Component=Final Answer2025.08 | 74 | 8 | 18 | |
| RealSafe-R1Model=R1-Qwen-14B, Evaluation Component=Final Answer2025.08 | 74 | 22 | 4 | |
| RealSafe-R1Model=R1-Qwen-32B, Evaluation Component=Reasoning2025.08 | 74 | 2 | 24 | |
| SmoothLLMModel=R1-Llama-8B, Output Component=Final Answer2025.08 | 76 | 8 | 16 | |
| SafeKeyModel=R1-Qwen-14B, Evaluation Component=Reasoning2025.08 | 76 | 4 | 20 | |
| ReasoningGuardModel=R1-Qwen-32B, Evaluation Component=Final Answer2025.08 | 78 | 6 | 16 | |
| SAFEPATH-ZSModel=R1-Llama-8B, Output Component=Final Answer2025.08 | 80 | 8 | 12 | |
| ReasoningGuardModel=R1-Llama-8B, Output Component=Final Answer2025.08 | 80 | 8 | 12 | |
| Self-ReminderModel=R1-Qwen-7B, Output Component=Final Answer2025.08 | 82 | 6 | 12 | |
| No DefenseModel=R1-Llama-8B, Output Component=Reasoning2025.08 | 82 | 0 | 18 | |
| Self-ReminderModel=R1-Llama-8B, Output Component=Reasoning2025.08 | 82 | 0 | 18 | |
| SmoothLLMModel=R1-Qwen-14B, Evaluation Component=Reasoning2025.08 | 82 | 0 | 18 | |
| SafeKeyModel=R1-Llama-8B, Output Component=Final Answer2025.08 | 84 | 4 | 12 | |
| SmoothLLMModel=R1-Qwen-32B, Evaluation Component=Reasoning2025.08 | 84 | 0 | 16 | |
| Self-ReminderModel=R1-Qwen-14B, Evaluation Component=Reasoning2025.08 | 86 | 0 | 14 | |
| RealSafe-R1Model=R1-Qwen-32B, Evaluation Component=Final Answer2025.08 | 86 | 8 | 6 | |
| Self-ReminderModel=R1-Qwen-32B, Evaluation Component=Reasoning2025.08 | 86 | 0 | 14 | |
| SmoothLLMModel=R1-Qwen-14B, Evaluation Component=Final Answer2025.08 | 88 | 0 | 12 | |
| ThinkingIModel=R1-Qwen-14B, Evaluation Component=Final Answer2025.08 | 88 | 4 | 8 | |
| SmoothLLMModel=R1-Qwen-32B, Evaluation Component=Final Answer2025.08 | 88 | 2 | 10 | |
| No DefenseModel=R1-Qwen-7B, Output Component=Reasoning2025.08 | 90 | 2 | 8 | |
| SafeDecodingModel=R1-Qwen-7B, Output Component=Reasoning2025.08 | 90 | 0 | 10 | |
| ReasoningGuardModel=R1-Qwen-14B, Evaluation Component=Final Answer2025.08 | 90 | 8 | 2 | |
| No DefenseModel=R1-Qwen-32B, Evaluation Component=Reasoning2025.08 | 90 | 0 | 10 | |
| ParaphraseModel=R1-Qwen-7B, Output Component=Final Answer2025.08 | 92 | 4 | 4 | |
| SafeDecodingModel=R1-Llama-8B, Output Component=Reasoning2025.08 | 92 | 0 | 8 | |
| SAFEPATH-ZSModel=R1-Qwen-14B, Evaluation Component=Final Answer2025.08 | 92 | 0 | 8 | |
| ParaphraseModel=R1-Qwen-32B, Evaluation Component=Reasoning2025.08 | 92 | 0 | 8 | |
| ThinkingIModel=R1-Qwen-32B, Evaluation Component=Final Answer2025.08 | 92 | 2 | 6 | |
| No DefenseModel=R1-Qwen-14B, Evaluation Component=Reasoning2025.08 | 94 | 0 | 6 | |
| SafeKeyModel=R1-Qwen-14B, Evaluation Component=Final Answer2025.08 | 94 | 4 | 2 | |
| SAFEPATH-ZSModel=R1-Qwen-32B, Evaluation Component=Final Answer2025.08 | 94 | 2 | 4 | |
| No DefenseModel=R1-Qwen-7B, Output Component=Final Answer2025.08 | 96 | 2 | 2 | |
| ParaphraseModel=R1-Llama-8B, Output Component=Reasoning2025.08 | 96 | 0 | 4 | |
| Self-ReminderModel=R1-Qwen-14B, Evaluation Component=Final Answer2025.08 | 96 | 0 | 4 | |
| No DefenseModel=R1-Qwen-32B, Evaluation Component=Final Answer2025.08 | 96 | 0 | 4 | |
| Self-ReminderModel=R1-Qwen-32B, Evaluation Component=Final Answer2025.08 | 96 | 0 | 4 | |
| SafeDecodingModel=R1-Qwen-7B, Output Component=Final Answer2025.08 | 98 | 0 | 2 | |
| No DefenseModel=R1-Llama-8B, Output Component=Final Answer2025.08 | 98 | 0 | 2 | |
| SafeDecodingModel=R1-Llama-8B, Output Component=Final Answer2025.08 | 98 | 0 | 2 | |
| ParaphraseModel=R1-Llama-8B, Output Component=Final Answer2025.08 | 98 | 0 | 2 | |
| Self-ReminderModel=R1-Llama-8B, Output Component=Final Answer2025.08 | 98 | 0 | 2 | |
| ParaphraseModel=R1-Qwen-14B, Evaluation Component=Reasoning2025.08 | 98 | 0 | 2 | |
| ParaphraseModel=R1-Qwen-14B, Evaluation Component=Final Answer2025.08 | 98 | 0 | 2 | |
| ParaphraseModel=R1-Qwen-32B, Evaluation Component=Final Answer2025.08 | 98 | 0 | 2 | |
| No DefenseModel=R1-Qwen-14B, Evaluation Component=Final Answer2025.08 | 100 | 0 | 0 |