Math on AMC 23 (accuracy)
72.5Accuracy (AMC 23)GRPO
Evaluation Results
| Method | Links | |
|---|---|---|
| GRPOGroup=Qwen3-4B-Base, Category=Base and Expert Models2026.06 | 72.5 | |
| DARE + TIESGroup=Qwen3-4B-Base, Category=Merged Models2026.06 | 70 | |
| RAMGroup=Qwen3-4B-Base, Category=Merged Models2026.06 | 70 | |
| TIESGroup=Qwen3-4B-Base, Category=Merged Models2026.06 | 65 | |
| TSV-MergeGroup=Qwen3-4B-Base, Category=Merged Models2026.06 | 65 | |
| RESMERGEGroup=Qwen3-4B-Base, Category=Merged Models2026.06 | 65 | |
| TAGroup=Qwen3-4B-Base, Category=Merged Models2026.06 | 55 | |
| General-ReasonerGroup=Qwen3-4B-Base, Category=Base and Expert Models2026.06 | 45 | |
| SpiralGroup=Qwen3-4B-Base, Category=Base and Expert Models2026.06 | 42.5 | |
| Qwen3-4BGroup=Qwen3-4B-Base, Category=Base and Expert Models2026.06 | 32.5 |