Gender bias evaluation on Pronoun Change
99.66Performance Score (P)PRISM-∆
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| PRISM-∆Model=QWEN3-14B2026.03 | 99.66 | 99.56 | |
| PRISM-∆VModel=QWEN3-14B2026.03 | 99.66 | 99.56 | |
| PRISM-∆VModel=QWEN3-8B2026.03 | 99.48 | 99.34 | |
| PRISM-∆Model=QWEN3-8B2026.03 | 99.4 | 99.3 | |
| PASTAModel=QWEN3-8B2026.03 | 98.86 | 98.72 | |
| **-markedModel=QWEN3-14B2026.03 | 98.86 | 98.68 | |
| SEKAModel=QWEN3-14B2026.03 | 98.66 | 98.54 | |
| SEKAModel=QWEN3-8B2026.03 | 98.56 | 98.26 | |
| OriginalModel=QWEN3-14B2026.03 | 98.42 | 98.22 | |
| **-markedModel=QWEN3-8B2026.03 | 98.1 | 97.84 | |
| OriginalModel=QWEN3-8B2026.03 | 98 | 97.84 | |
| PRISM-∆Model=GEMMA3-12B2026.03 | 97.94 | 97.46 | |
| SEKAModel=GEMMA3-12B2026.03 | 97.7 | 97.24 | |
| PRISM-∆Model=QWEN3-4B2026.03 | 96.18 | 94.6 | |
| PRISM-∆VModel=QWEN3-4B2026.03 | 96.06 | 94.52 | |
| PASTAModel=QWEN3-4B2026.03 | 95.82 | 94.64 | |
| **-markedModel=GEMMA3-12B2026.03 | 95.78 | 96.42 | |
| **-markedModel=QWEN3-4B2026.03 | 95.76 | 93.88 | |
| SEKAModel=QWEN3-4B2026.03 | 95.18 | 93.26 | |
| PRISM-∆VModel=GEMMA3-12B2026.03 | 94.54 | 94.42 | |
| OriginalModel=GEMMA3-12B2026.03 | 93.92 | 94.96 | |
| OriginalModel=QWEN3-4B2026.03 | 93.14 | 90.52 | |
| SPAModel=QWEN3-14B2026.03 | 91.6 | 92.2 | |
| PASTAModel=QWEN3-14B2026.03 | 90.98 | 90.94 | |
| PRISM-∆VModel=GEMMA3-4B2026.03 | 90.16 | 91.68 | |
| PRISM-∆Model=GEMMA3-4B2026.03 | 89.08 | 90.58 | |
| SPAModel=GEMMA3-12B2026.03 | 86.41 | 84.99 | |
| SEKAModel=GEMMA3-4B2026.03 | 81.53 | 81.11 | |
| SPAModel=QWEN3-4B2026.03 | 80.27 | 78.19 | |
| SPAModel=GEMMA3-4B2026.03 | 76.05 | 74.45 | |
| SPAModel=QWEN3-8B2026.03 | 72.61 | 74.95 | |
| PASTAModel=GEMMA3-12B2026.03 | 68.47 | 68.01 | |
| PASTAModel=GEMMA3-4B2026.03 | 67.39 | 66.43 | |
| OriginalModel=GEMMA3-4B2026.03 | 41.34 | 35.25 | |
| **-markedModel=GEMMA3-4B2026.03 | 38.86 | 32.45 |