Reasoning on 15 reasoning benchmarks weighted mean (test)
81.35AccuracyCoT
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CoTModel=Qwen3-4B-T, Routing Strategy=CoT2026.05 | 81.35 | 642.5 | |
| EDRM-MLPModel=Qwen3-4B-T, Routing Strategy=EDRM-MLP2026.05 | 81.2 | 424.8 | |
| StandardModel=Qwen3-4B-T, Routing Strategy=Standard2026.05 | 80.96 | 518.4 | |
| EDRM-Inst-EModel=Qwen3-4B-T, Routing Strategy=Inst-E2026.05 | 80.29 | 401.1 | |
| EDRM-Inst-CModel=Qwen3-4B-T, Routing Strategy=Inst-C2026.05 | 79.03 | 467.5 | |
| EDRM-Global-EModel=Qwen3-4B-T, Routing Strategy=Global-E2026.05 | 78.23 | 335.1 | |
| EDRM-Global-CModel=Qwen3-4B-T, Routing Strategy=Global-C2026.05 | 78.23 | 335.1 | |
| EDRM-Inst-EModel=Qwen2.5-7B, Routing Strategy=Inst-E2026.05 | 78.11 | 242.7 | |
| EDRM-MLPModel=Qwen2.5-7B, Routing Strategy=EDRM-MLP2026.05 | 77.93 | 240.4 | |
| Token-SignModel=Qwen3-4B-T, Routing Strategy=Token-Sign2026.05 | 77.65 | 400.6 | |
| EDRM-Inst-CModel=Qwen2.5-7B, Routing Strategy=Inst-C2026.05 | 76.91 | 255.3 | |
| EDRM-Global-EModel=Qwen2.5-7B, Routing Strategy=Global-E2026.05 | 74.01 | 191.8 | |
| EDRM-Global-CModel=Qwen2.5-7B, Routing Strategy=Global-C2026.05 | 74.01 | 190.6 | |
| CoTModel=Qwen2.5-7B, Routing Strategy=CoT2026.05 | 73.38 | 330.3 | |
| StandardModel=Qwen2.5-7B, Routing Strategy=Standard2026.05 | 72.89 | 240.6 | |
| Token-SignModel=Qwen2.5-7B, Routing Strategy=Token-Sign2026.05 | 72.67 | 247.4 | |
| EDRM-MLPModel=Llama3.1-8B, Routing Strategy=EDRM-MLP2026.05 | 72.27 | 149.9 | |
| EDRM-Inst-EModel=Llama3.1-8B, Routing Strategy=Inst-E2026.05 | 71.97 | 153.9 | |
| EDRM-Inst-CModel=Llama3.1-8B, Routing Strategy=Inst-C2026.05 | 71.01 | 170.16 | |
| EDRM-Global-EModel=Llama3.1-8B, Routing Strategy=Global-E2026.05 | 68.48 | 164.4 | |
| EDRM-Global-CModel=Llama3.1-8B, Routing Strategy=Global-C2026.05 | 68.46 | 167.4 | |
| StandardModel=Llama3.1-8B, Routing Strategy=Standard2026.05 | 68.21 | 127.1 | |
| CoTModel=Llama3.1-8B, Routing Strategy=CoT2026.05 | 68.07 | 277.8 | |
| EDRM-MLPModel=Llama3.2-3B, Routing Strategy=EDRM-MLP2026.05 | 66.64 | 178.9 | |
| Token-SignModel=Llama3.1-8B, Routing Strategy=Token-Sign2026.05 | 65.71 | 170.8 | |
| DirectModel=Qwen3-4B-T, Routing Strategy=Direct2026.05 | 65.63 | 6.2 | |
| EDRM-Inst-EModel=Llama3.2-3B, Routing Strategy=Inst-E2026.05 | 65.24 | 179.4 | |
| EDRM-Inst-CModel=Llama3.2-3B, Routing Strategy=Inst-C2026.05 | 64.7 | 186.9 | |
| DirectModel=Qwen2.5-7B, Routing Strategy=Direct2026.05 | 64.63 | 3.4 | |
| EDRM-Global-EModel=Llama3.2-3B, Routing Strategy=Global-E2026.05 | 61.89 | 113.2 | |
| EDRM-Global-CModel=Llama3.2-3B, Routing Strategy=Global-C2026.05 | 61.89 | 113.2 | |
| CoTModel=Llama3.2-3B, Routing Strategy=CoT2026.05 | 60.84 | 251.7 | |
| StandardModel=Llama3.2-3B, Routing Strategy=Standard2026.05 | 59.92 | 194.4 | |
| Token-SignModel=Llama3.2-3B, Routing Strategy=Token-Sign2026.05 | 59.63 | 184.9 | |
| DirectModel=Llama3.1-8B, Routing Strategy=Direct2026.05 | 59 | 5 | |
| DirectModel=Llama3.2-3B, Routing Strategy=Direct2026.05 | 52.75 | 4.4 |