Safety Evaluation on SORRY-Bench (expected HFR)
0.9Expected HFRG
Evaluation Results
| Method | Links | |
|---|---|---|
| GBase Model=GPT-OSS-20B2025.10 | 0.9 | |
| MIBase Model=GPT-OSS-20B2025.10 | 0.9 | |
| SGBase Model=LLaMA-2-7B2025.10 | 5.1 | |
| SCBase Model=GPT-OSS-20B2025.10 | 11.3 | |
| SGBase Model=LLaMA-2-13B2025.10 | 14.2 | |
| SGBase Model=Llama-3.1-8B2025.10 | 16.7 | |
| ER-GBase Model=GPT-OSS-20B2025.10 | 20.7 | |
| ER-GBase Model=Llama-3.2-1B2025.10 | 22 | |
| SGBase Model=GPT-OSS-20B2025.10 | 22.3 | |
| ER-DBase Model=Llama-3.2-1B2025.10 | 24.5 | |
| DBase Model=Llama-3.2-1B2025.10 | 26.3 | |
| SCBase Model=LLaMA-2-7B2025.10 | 27.2 | |
| MIBase Model=Llama-3.2-1B2025.10 | 27.3 | |
| GBase Model=Llama-3.2-1B2025.10 | 27.6 | |
| SCBase Model=LLaMA-2-13B2025.10 | 29.5 | |
| ER-DBase Model=GPT-OSS-20B2025.10 | 30.1 | |
| ER-GBase Model=Llama-3.1-8B2025.10 | 30.4 | |
| MIBase Model=LLaMA-2-7B2025.10 | 30.7 | |
| GBase Model=LLaMA-2-7B2025.10 | 30.8 | |
| ER-DBase Model=Llama-3.1-8B2025.10 | 30.9 | |
| DBase Model=GPT-OSS-20B2025.10 | 31 | |
| ER-GBase Model=LLaMA-2-7B2025.10 | 31.2 | |
| SCBase Model=Llama-3.2-1B2025.10 | 31.2 | |
| DBase Model=LLaMA-2-7B2025.10 | 31.6 | |
| ER-DBase Model=LLaMA-2-7B2025.10 | 31.6 | |
| SCBase Model=Llama-3.1-8B2025.10 | 32.3 | |
| DBase Model=LLaMA-2-13B2025.10 | 32.8 | |
| GBase Model=Llama-3.1-8B2025.10 | 32.9 | |
| MIBase Model=Llama-3.1-8B2025.10 | 32.9 | |
| DBase Model=Llama-3.1-8B2025.10 | 33.2 | |
| GBase Model=LLaMA-2-13B2025.10 | 33.5 | |
| MIBase Model=LLaMA-2-13B2025.10 | 33.5 | |
| ER-DBase Model=LLaMA-2-13B2025.10 | 33.5 | |
| ER-GBase Model=LLaMA-2-13B2025.10 | 35.6 | |
| SGBase Model=Llama-3.2-1B2025.10 | 41.5 |