Safety Evaluation on Do-Not-Answer (test)
3.195ASRFull Precision
Evaluation Results
| Method | Links | |
|---|---|---|
| Full PrecisionBackbone=Qwen-2.5-7B-Instruct, Quantization=Full Precision2026.01 | 3.195 | |
| AWQ-trustBackbone=Qwen-2.5-7B-Instruct, Quantization=AWQ-trust2026.01 | 3.301 | |
| AWQBackbone=Qwen-2.5-7B-Instruct, Quantization=AWQ2026.01 | 3.585 | |
| AWQ-trustBackbone=Llama-3.1-8B-Instruct, Quantization=AWQ-trust2026.01 | 4.26 | |
| Full PrecisionBackbone=Gemma-7B-Instruct, Quantization=Full Precision2026.01 | 5.431 | |
| AWQBackbone=Llama-3.1-8B-Instruct, Quantization=AWQ2026.01 | 5.964 | |
| Full PrecisionBackbone=Llama-3.1-8B-Instruct, Quantization=Full Precision2026.01 | 6.07 | |
| AWQ-trustBackbone=Gemma-7B-Instruct, Quantization=AWQ-trust2026.01 | 6.567 | |
| AWQBackbone=Gemma-7B-Instruct, Quantization=AWQ2026.01 | 7.987 |