Safety Evaluation on MultiJail-KO (test)
90.053Safety RateAWQ-trust
Evaluation Results
| Method | Links | |
|---|---|---|
| AWQ-trustBackbone=Qwen-2.5-7B-Instruct, Quantization=AWQ-trust2026.01 | 90.053 | |
| Full PrecisionBackbone=Qwen-2.5-7B-Instruct, Quantization=Full Precision2026.01 | 81.27 | |
| AWQBackbone=Qwen-2.5-7B-Instruct, Quantization=AWQ2026.01 | 80.212 | |
| AWQ-trustBackbone=Llama-3.1-8B-Instruct, Quantization=AWQ-trust2026.01 | 66.138 | |
| AWQ-trustBackbone=Gemma-7B-Instruct, Quantization=AWQ-trust2026.01 | 54.497 | |
| Full PrecisionBackbone=Gemma-7B-Instruct, Quantization=Full Precision2026.01 | 50.899 | |
| AWQBackbone=Gemma-7B-Instruct, Quantization=AWQ2026.01 | 45.291 | |
| Full PrecisionBackbone=Llama-3.1-8B-Instruct, Quantization=Full Precision2026.01 | 15.238 | |
| AWQBackbone=Llama-3.1-8B-Instruct, Quantization=AWQ2026.01 | 9.63 |