Math on MATH 500
92.7AccuracyDense
Evaluation results
| Method | Links | ||
|---|---|---|---|
| DenseModel=DeepSeek R1 Distill Qwen 32B2025.07 | 92.7 | 0 | |
| ReasonCacheModel=DeepSeek R1 Distill Qwen 32B2025.07 | 91.8 | 39.1 | |
| ReasonCacheModel=Phi 4 reasoning plus2025.07 | 90.3 | 21.41 | |
| DenseModel=QwQ 32B2025.07 | 89.4 | 0 | |
| DenseModel=Phi 4 reasoning plus2025.07 | 89.3 | 0 | |
| ReasonCacheModel=QwQ 32B2025.07 | 87.2 | 26.89 | |
| RAMGroup=Qwen3-4B-Base, Category=Merged Models2026.06 | 78.4 | — | |
| RESMERGEGroup=Qwen3-4B-Base, Category=Merged Models2026.06 | 78 | — | |
| TIESGroup=Qwen3-4B-Base, Category=Merged Models2026.06 | 77.8 | — | |
| GRPOGroup=Qwen3-4B-Base, Category=Base and Expert Models2026.06 | 77.4 | — | |
| StreamingLLMModel=DeepSeek R1 Distill Qwen 32B2025.07 | 76.4 | 38.2 | |
| TAGroup=Qwen3-4B-Base, Category=Merged Models2026.06 | 76 | — | |
| TSV-MergeGroup=Qwen3-4B-Base, Category=Merged Models2026.06 | 76 | — | |
| DARE + TIESGroup=Qwen3-4B-Base, Category=Merged Models2026.06 | 74.4 | — | |
| SpiralGroup=Qwen3-4B-Base, Category=Base and Expert Models2026.06 | 73 | — | |
| StreamingLLMModel=QwQ 32B2025.07 | 67.2 | 29.7 | |
| General-ReasonerGroup=Qwen3-4B-Base, Category=Base and Expert Models2026.06 | 66 | — | |
| Qwen3-4BGroup=Qwen3-4B-Base, Category=Base and Expert Models2026.06 | 52.4 | — | |
| StreamingLLMModel=Phi 4 reasoning plus2025.07 | 52.2 | 23 | |
| QuestModel=Phi 4 reasoning plus2025.07 | 51.8 | 21 | |
| QuestModel=QwQ 32B2025.07 | 50.6 | 28 | |
| QuestModel=DeepSeek R1 Distill Qwen 32B2025.07 | 50 | 39.9 | |
| SnapKVModel=DeepSeek R1 Distill Qwen 32B2025.07 | 44 | 2.79 | |
| SnapKVModel=QwQ 32B2025.07 | 17.6 | 1.55 | |
| SnapKVModel=Phi 4 reasoning plus2025.07 | 14.6 | 4.76 |