Bias-Penalized Accuracy Evaluation on High-Resource Languages (HRLs)
93.39Bias-Penalized Accuracy (BPA)SFT
Evaluation results
| Method | Links | |
|---|---|---|
| SFTModel=Gemma-2-9B2026.02 | 93.39 | |
| SteeringModel=Gemma-2-9B2026.02 | 89.65 | |
| SteeringModel=Qwen-2.5-7B2026.02 | 85.43 | |
| SFTModel=Qwen-2.5-7B2026.02 | 84.63 | |
| SFTModel=Mistral-7B2026.02 | 78.17 | |
| SteeringModel=Mistral-7B2026.02 | 77.71 | |
| CoTModel=Gemma-2-9B2026.02 | 58.54 | |
| CoTModel=Qwen-2.5-7B2026.02 | 57.73 | |
| CoTModel=Mistral-7B2026.02 | 34.2 |