Math Word Problems on GSM8K (Expected utility)
89.2Expected UtilityFlowCompile
Evaluation Results
| Method | Links | |
|---|---|---|
| FlowCompileWorkflow Category=Main Method2026.05 | 89.2 | |
| MaAS (Qwen3-4B)Workflow Category=Workflow with Router2026.05 | 87.2 | |
| MaAS (Qwen3-8B)Workflow Category=Workflow with Router2026.05 | 82.2 | |
| KNN RouterWorkflow Category=Workflow with Router2026.05 | 81.9 | |
| Pref-Aware MaASWorkflow Category=Workflow with Router2026.05 | 80.7 | |
| Qwen3-32BWorkflow Category=Single Model2026.05 | 74.1 | |
| Qwen3-4BWorkflow Category=Fixed Workflow2026.05 | 69.9 | |
| Pref-Aware RouterWorkflow Category=Workflow with Router2026.05 | 68 | |
| QwQ-32BWorkflow Category=Single Model2026.05 | 61.6 | |
| Qwen3-8BWorkflow Category=Fixed Workflow2026.05 | 57.6 |