Code on LCB (AVG@8)
64.3AVG@8DARE + TIES
Evaluation Results
| Method | Links | |
|---|---|---|
| DARE + TIESScale=14B, Category=Cross-Domain Merging2026.07 | 64.3 | |
| ProjectedScale=14B, Category=Cross-Domain Merging2026.07 | 63.4 | |
| TIES MergingScale=14B, Category=Cross-Domain Merging2026.07 | 63.3 | |
| DeepCoderScale=14B, Category=Code Expert2026.07 | 61 | |
| Best Single-Domain ExpertScale=14B2026.07 | 61 | |
| OR-ZeroScale=14B, Category=Math Expert2026.07 | 56.7 | |
| Task ArithmeticScale=14B, Category=Cross-Domain Merging2026.07 | 54.6 | |
| Distill-Qwen-14BScale=14B, Category=Single-Domain Reference2026.07 | 54.2 | |
| ProjectedScale=1.5B, Category=Cross-Domain Merging2026.07 | 32.25 | |
| Archer-CodeScale=1.5B, Category=Code Expert2026.07 | 31.8 | |
| Best Single-Domain ExpertScale=1.5B2026.07 | 31.8 | |
| Task ArithmeticScale=1.5B, Category=Cross-Domain Merging2026.07 | 31.8 | |
| DARE + TIESScale=1.5B, Category=Cross-Domain Merging2026.07 | 31.7 | |
| TIES MergingScale=1.5B, Category=Cross-Domain Merging2026.07 | 31.4 | |
| DeepScaleRScale=1.5B, Category=Math Expert2026.07 | 27.2 | |
| Distill-Qwen-1.5BScale=1.5B, Category=Single-Domain Reference2026.07 | 23 |