Question Answering on Hebrew judicial decisions synthetic-organic instruction pairs 4B (test)
11.997BLEUVanilla-Gemma
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Vanilla-GemmaCategory=Non-personalized baselines2026.04 | 11.997 | 0.113 | 0.192 | -0.017 | |
| Vanilla-DictaLM3Category=Non-personalized baselines2026.04 | 11.845 | 0.089 | 0.188 | -0.005 | |
| Pers-CLMCategory=Alternative personalization models2026.04 | 11.748 | 0.079 | 0.181 | -0.001 | |
| Gemma-3Category=Retrieval-augmented baselines2026.04 | 10.421 | 0.023 | 0.148 | 0.011 | |
| Mixed-GemmaCategory=Non-personalized baselines2026.04 | 9.205 | 0.043 | 0.083 | -0.012 | |
| Joint-GemmaCategory=Non-personalized baselines2026.04 | 8.912 | 0.037 | 0.078 | -0.009 | |
| Gemini-3-ProCategory=Retrieval-augmented baselines2026.04 | 8.773 | 0.021 | 0.077 | -0.001 | |
| Pers-ITCategory=Alternative personalization models2026.04 | 4.973 | 0.02 | 0.043 | -0.001 |