Mathematical Reasoning on MATH 500 (Accuracy, Think ↓, Route%)
92.6AccuracyDART
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| DARTBackbone=DeepSeek-V3.2, Hosted API=true2026.06 | 92.6 | 56 | — | |
| SC-RouteBackbone=DeepSeek-V3.2, Hosted API=true2026.06 | 90.6 | — | — | |
| SC-RouteBackbone=Qwen3-32B2026.06 | 88.5 | — | — | |
| DARTBackbone=Qwen3-32B2026.06 | 88.5 | 69 | 80 | |
| ATBackbone=DeepSeek-V3.2, Hosted API=true2026.06 | 88.4 | — | — | |
| DARTBackbone=Qwen3-8B2026.06 | 88.2 | 67 | 78 | |
| SC-RouteBackbone=Qwen3-8B2026.06 | 87.6 | — | — | |
| ATBackbone=Qwen3-14B2026.06 | 87.6 | — | — | |
| SC-RouteBackbone=Qwen3-14B2026.06 | 87.6 | — | — | |
| DARTBackbone=Qwen3-14B2026.06 | 87.6 | 37 | 83.4 | |
| ATBackbone=Qwen3-32B2026.06 | 86.2 | — | — | |
| ATBackbone=Qwen3-8B2026.06 | 85.6 | — | — | |
| NTBackbone=DeepSeek-V3.2, Hosted API=true2026.06 | 84.8 | — | — | |
| NTBackbone=Qwen3-32B2026.06 | 82.2 | — | — | |
| NTBackbone=Qwen3-14B2026.06 | 81.2 | — | — | |
| NTBackbone=Qwen3-8B2026.06 | 76.6 | — | — |