Safety Alignment on HEx-PHI (Accuracy)
99.06AccuracyHELLoRA
Evaluation Results
| Method | Links | |
|---|---|---|
| HELLoRAModel=OlMoE-1B-7B, Params=0.678%2026.05 | 99.06 | |
| LoRAModel=OlMoE-1B-7B, Params=4.300%2026.05 | 98.75 | |
| HELLoRAModel=DeepSeekMoE, Params=0.823%2026.05 | 98.49 | |
| LoRAModel=DeepSeekMoE, Params=3.545%2026.05 | 97.87 | |
| HELLoRAModel=Mixtral-8×7B, Params=0.310%2026.05 | 96.67 | |
| LoRAModel=Mixtral-8×7B, Params=1.030%2026.05 | 93.94 |