Cross-domain Evaluation on Multi-domain Evaluation Suite (mid-training)
60.4Math ScoreMergeMix
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| MergeMix2026.01 | 60.4 | 51.6 | 55.6 | 78.9 | 57.3 | 58.4 | |
| Adapted RegMix2026.01 | 59.7 | 51.5 | 55.5 | 78.3 | 57 | 58.1 | |
| Manual2026.01 | 59.3 | 50.5 | 55.5 | 78.9 | 57.1 | 57.9 | |
| Uniform2026.01 | 57.1 | 44.6 | 50 | 76 | 56.7 | 54.2 |