Safety Evaluation on BeaverTails & DirectRefusal Average
32.84Average ASRSafeGene
Evaluation Results
| Method | Links | |
|---|---|---|
| SafeGeneBackbone=Qwen2.5-7B, Downstream Task Context=MNLI, Safety Judge=Beaver-dam-7B2026.06 | 32.84 | |
| SafeGeneBackbone=Qwen2.5-7B, Downstream Task Context=BoolQ, Safety Judge=Beaver-dam-7B2026.06 | 33 | |
| SafeGeneBackbone=Qwen2.5-7B, Downstream Task Context=AG News, Safety Judge=Beaver-dam-7B2026.06 | 33.45 | |
| SafeGeneBackbone=Qwen2.5-7B, Downstream Task Context=Average, Safety Judge=Beaver-dam-7B2026.06 | 33.84 | |
| SaLoRABackbone=Qwen2.5-7B, Downstream Task Context=BoolQ, Safety Judge=Beaver-dam-7B2026.06 | 35.54 | |
| SafeGeneBackbone=Qwen2.5-7B, Downstream Task Context=SST2, Safety Judge=Beaver-dam-7B2026.06 | 36.05 | |
| SaLoRABackbone=Qwen2.5-7B, Downstream Task Context=AG News, Safety Judge=Beaver-dam-7B2026.06 | 36.64 | |
| SaLoRABackbone=Qwen2.5-7B, Downstream Task Context=Average, Safety Judge=Beaver-dam-7B2026.06 | 36.84 | |
| SafeMERGEBackbone=Qwen2.5-7B, Downstream Task Context=MNLI, Safety Judge=Beaver-dam-7B2026.06 | 36.89 | |
| SaLoRABackbone=Qwen2.5-7B, Downstream Task Context=SST2, Safety Judge=Beaver-dam-7B2026.06 | 36.95 | |
| SafeMERGEBackbone=Qwen2.5-7B, Downstream Task Context=BoolQ, Safety Judge=Beaver-dam-7B2026.06 | 37.12 | |
| SafeMERGEBackbone=Qwen2.5-7B, Downstream Task Context=Average, Safety Judge=Beaver-dam-7B2026.06 | 37.75 | |
| SafeMERGEBackbone=Qwen2.5-7B, Downstream Task Context=SST2, Safety Judge=Beaver-dam-7B2026.06 | 38.02 | |
| SaLoRABackbone=Qwen2.5-7B, Downstream Task Context=MNLI, Safety Judge=Beaver-dam-7B2026.06 | 38.23 | |
| SafeMERGEBackbone=Qwen2.5-7B, Downstream Task Context=AG News, Safety Judge=Beaver-dam-7B2026.06 | 38.97 | |
| SafeLoRABackbone=Qwen2.5-7B, Downstream Task Context=BoolQ, Safety Judge=Beaver-dam-7B2026.06 | 39.64 | |
| SafeLoRABackbone=Qwen2.5-7B, Downstream Task Context=Average, Safety Judge=Beaver-dam-7B2026.06 | 40.63 | |
| Fine-tunedBackbone=Qwen2.5-7B, Downstream Task Context=MNLI, Safety Judge=Beaver-dam-7B2026.06 | 40.71 | |
| SafeLoRABackbone=Qwen2.5-7B, Downstream Task Context=AG News, Safety Judge=Beaver-dam-7B2026.06 | 40.87 | |
| SafeLoRABackbone=Qwen2.5-7B, Downstream Task Context=MNLI, Safety Judge=Beaver-dam-7B2026.06 | 40.89 | |
| SafeLoRABackbone=Qwen2.5-7B, Downstream Task Context=SST2, Safety Judge=Beaver-dam-7B2026.06 | 41.14 | |
| Fine-tunedBackbone=Qwen2.5-7B, Downstream Task Context=BoolQ, Safety Judge=Beaver-dam-7B2026.06 | 46.38 | |
| Fine-tunedBackbone=Qwen2.5-7B, Downstream Task Context=Average, Safety Judge=Beaver-dam-7B2026.06 | 47.46 | |
| Fine-tunedBackbone=Qwen2.5-7B, Downstream Task Context=AG News, Safety Judge=Beaver-dam-7B2026.06 | 48 | |
| Fine-tunedBackbone=Qwen2.5-7B, Downstream Task Context=SST2, Safety Judge=Beaver-dam-7B2026.06 | 54.73 |