Multi-task Language Understanding on MMLU (pass@1 Accuracy)
88pass@1 AccuracySelf-Reminder
Evaluation Results
| Method | Links | |
|---|---|---|
| Self-ReminderBackbone=R1-Qwen-32B2025.08 | 88 | |
| No DefenseBackbone=R1-Qwen-32B2025.08 | 87 | |
| RealSafe-R1Backbone=R1-Qwen-32B2025.08 | 86 | |
| No DefenseBackbone=R1-Qwen-14B2025.08 | 85 | |
| ThinkingIBackbone=R1-Qwen-14B2025.08 | 85 | |
| SAFEPATH-ZSBackbone=R1-Qwen-32B2025.08 | 85 | |
| Self-ReminderBackbone=R1-Qwen-14B2025.08 | 84 | |
| SAFEPATH-ZSBackbone=R1-Qwen-14B2025.08 | 84 | |
| SafeKeyBackbone=R1-Qwen-14B2025.08 | 83 | |
| ReasoningGuardBackbone=R1-Qwen-32B2025.08 | 83 | |
| ThinkingIBackbone=R1-Qwen-32B2025.08 | 82 | |
| RealSafe-R1Backbone=R1-Qwen-14B2025.08 | 81 | |
| ReasoningGuardBackbone=R1-Qwen-14B2025.08 | 81 | |
| SmoothLLMBackbone=R1-Qwen-14B2025.08 | 72 | |
| SmoothLLMBackbone=R1-Qwen-32B2025.08 | 68 | |
| ParaphraseBackbone=R1-Qwen-32B2025.08 | 38 | |
| ParaphraseBackbone=R1-Qwen-14B2025.08 | 35 |