Bias-Penalized Accuracy Evaluation on Low-Resource Languages (LRLs)
76.96BPASFT
Evaluation results
| Method | Links | |
|---|---|---|
| SFTModel=Gemma-2-9B2026.02 | 76.96 | |
| SteeringModel=Gemma-2-9B2026.02 | 76.62 | |
| SFTModel=Qwen-2.5-7B2026.02 | 70.12 | |
| SteeringModel=Qwen-2.5-7B2026.02 | 65.3 | |
| SteeringModel=Mistral-7B2026.02 | 56.49 | |
| CoTModel=Qwen-2.5-7B2026.02 | 55 | |
| SFTModel=Mistral-7B2026.02 | 53.68 | |
| CoTModel=Gemma-2-9B2026.02 | 53.58 | |
| CoTModel=Mistral-7B2026.02 | 49.97 |