Refusal Evaluation on XSTest Seemingly Toxic Subsets
98XSDCR
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| DCRBase Model=Qwen2.5-1.5B2026.02 | 98 | 98 | 83 | 86 | 86 | |
| SurgicalBase Model=Qwen2.5-7B2026.02 | 93 | 96 | 71 | 96 | 89 | |
| DCRBase Model=Qwen2.5-7B2026.02 | 93 | 96 | 71 | 94 | 91 | |
| DCRBase Model=LLaMA-3-8B2026.02 | 93 | 99 | 85 | 92 | 90 | |
| SCANSBase Model=Qwen2.5-7B2026.02 | 84 | 92 | 50 | 97 | 91 | |
| STL-augBase Model=LLaMA-3-8B2026.02 | 84 | 96 | 59 | 87 | 85 | |
| SCANSBase Model=LLaMA-3-8B2026.02 | 84 | 97 | 86 | 80 | 90 | |
| SCANSBase Model=Qwen2.5-1.5B2026.02 | 83 | 92 | 87 | 84 | 87 | |
| SurgicalBase Model=Qwen2.5-1.5B2026.02 | 81 | 84 | 54 | 78 | 54 | |
| STLBase Model=LLaMA-3-8B2026.02 | 79 | 94 | 59 | 89 | 85 | |
| STL-augBase Model=Qwen2.5-1.5B2026.02 | 75 | 90 | 69 | 76 | 75 | |
| STL-augBase Model=Qwen2.5-7B2026.02 | 74 | 89 | 53 | 85 | 83 | |
| STLBase Model=Qwen2.5-1.5B2026.02 | 73 | 88 | 72 | 75 | 75 | |
| SurgicalBase Model=LLaMA-3-8B2026.02 | 72 | 90 | 53 | 89 | 85 | |
| STLBase Model=Qwen2.5-7B2026.02 | 66 | 87 | 34 | 87 | 80 |