Safety Alignment on CoSApien Arab Publisher
91.6SafetyLlama-3.1-8B-Inst
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Llama-3.1-8B-InstOptimization Method=Reinforce++2026.05 | 91.6 | 3.8 | 3.475 | |
| CSRMOptimization Method=Reinforce++2026.05 | 76.6 | 4.75 | 3.603 | |
| Llama-3.1-8B-InstOptimization Method=Reward Distillation2026.05 | 70 | 4.7 | 3.1 | |
| CSRMOptimization Method=Reward Distillation2026.05 | 69.5 | 4.645 | 3.145 | |
| LlamaGuard-3-8BOptimization Method=Reinforce++2026.05 | 62.2 | 4.456 | 2.756 | |
| LlamaGuard-3-8BOptimization Method=Reward Distillation2026.05 | 57.5 | 4.975 | 2.825 |