Jailbreak Robustness on JB-R1
98.1Evaluation Score (avg@4)Self-ReSET
Evaluation Results
| Method | Links | |
|---|---|---|
| Self-ReSETBase Model=DeepSeek-R1-Distill-Llama-8B2026.05 | 98.1 | |
| Self-ReSETBase Model=Qwen3-8B2026.05 | 97.4 | |
| RECAPBase Model=DeepSeek-R1-Distill-Llama-8B2026.05 | 96.3 | |
| DAPOBase Model=DeepSeek-R1-Distill-Llama-8B2026.05 | 94.9 | |
| Self-ReSETBase Model=DeepSeek-R1-Distill-Qwen-7B2026.05 | 93.2 | |
| STAR-1Base Model=DeepSeek-R1-Distill-Llama-8B2026.05 | 92 | |
| STAR-1Base Model=Qwen3-8B2026.05 | 91.9 | |
| DAPOBase Model=Qwen3-8B2026.05 | 91.1 | |
| RECAPBase Model=Qwen3-8B2026.05 | 90.3 | |
| STAR-1Base Model=DeepSeek-R1-Distill-Qwen-7B2026.05 | 85.8 | |
| RECAPBase Model=DeepSeek-R1-Distill-Qwen-7B2026.05 | 83.7 | |
| DAPOBase Model=DeepSeek-R1-Distill-Qwen-7B2026.05 | 81.4 | |
| SafechainBase Model=Qwen3-8B2026.05 | 73.3 | |
| BaseBase Model=Qwen3-8B2026.05 | 64.9 | |
| SafechainBase Model=DeepSeek-R1-Distill-Llama-8B2026.05 | 63.7 | |
| SafechainBase Model=DeepSeek-R1-Distill-Qwen-7B2026.05 | 58 | |
| BaseBase Model=DeepSeek-R1-Distill-Llama-8B2026.05 | 50.6 | |
| BaseBase Model=DeepSeek-R1-Distill-Qwen-7B2026.05 | 44.8 |