False Refusal Evaluation on ORB-H
97.2CRELS
Evaluation Results
| Method | Links | |
|---|---|---|
| ELSModel=QWEN 3 1.7B2025.10 | 97.2 | |
| CASTModel=QWEN 3 1.7B2025.10 | 96.2 | |
| AlphaSteerModel=QWEN 3 1.7B2025.10 | 96 | |
| SurgicalModel=QWEN 3 1.7B2025.10 | 95.8 | |
| AdaSteerModel=QWEN 3 1.7B2025.10 | 95.8 | |
| QWEN 3 1.7BModel=QWEN 3 1.7B2025.10 | 95.5 | |
| ELSModel=QWEN 3 14B2025.10 | 84.8 | |
| AlphaSteerModel=QWEN 3 14B2025.10 | 84.1 | |
| AdaSteerModel=QWEN 3 14B2025.10 | 83.7 | |
| CASTModel=QWEN 3 14B2025.10 | 83 | |
| ELSModel=LLAMA3.1-8B-INST2025.10 | 82.6 | |
| SurgicalModel=QWEN 3 14B2025.10 | 82.6 | |
| AdaSteerModel=LLAMA3.1-8B-INST2025.10 | 81.1 | |
| QWEN 3 14BModel=QWEN 3 14B2025.10 | 81.1 | |
| ELSModel=QWEN 3 8B2025.10 | 80.6 | |
| AlphaSteerModel=QWEN 3 8B2025.10 | 80.3 | |
| CASTModel=QWEN 3 8B2025.10 | 79.5 | |
| ELSModel=LLAMA2-7B-CHAT2025.10 | 78.4 | |
| AdaSteerModel=QWEN 3 8B2025.10 | 78 | |
| SurgicalModel=QWEN 3 8B2025.10 | 77.6 | |
| AlphaSteerModel=LLAMA3.1-8B-INST2025.10 | 77.3 | |
| SurgicalModel=LLAMA3.1-8B-INST2025.10 | 76.6 | |
| AdaSteerModel=LLAMA2-7B-CHAT2025.10 | 75.7 | |
| AlphaSteerModel=LLAMA2-7B-CHAT2025.10 | 75 | |
| QWEN 3 8BModel=QWEN 3 8B2025.10 | 75 | |
| CASTModel=LLAMA3.1-8B-INST2025.10 | 70.3 | |
| CASTModel=LLAMA2-7B-CHAT2025.10 | 66.7 | |
| SurgicalModel=LLAMA2-7B-CHAT2025.10 | 65.5 | |
| LLAMA3.1-8B-INSTModel=LLAMA3.1-8B-INST2025.10 | 57.3 | |
| system promptModel=QWEN 3 1.7B2025.10 | 54.2 | |
| system promptModel=QWEN 3 14B2025.10 | 50.8 | |
| system promptModel=QWEN 3 8B2025.10 | 43.2 | |
| system promptModel=LLAMA3.1-8B-INST2025.10 | 41 | |
| LLAMA2-7B-CHATModel=LLAMA2-7B-CHAT2025.10 | 14.8 | |
| system promptModel=LLAMA2-7B-CHAT2025.10 | 8.6 |