Code Generation on LiveCodeBench (Expected utility)
80.1Expected UtilityFlowCompile
Evaluation Results
| Method | Links | |
|---|---|---|
| FlowCompileWorkflow Category=Main Method2026.05 | 80.1 | |
| MaAS (Qwen3-4B)Workflow Category=Workflow with Router2026.05 | 66 | |
| Pref-Aware MaASWorkflow Category=Workflow with Router2026.05 | 65.5 | |
| Qwen3-4BWorkflow Category=Fixed Workflow2026.05 | 61 | |
| Pref-Aware RouterWorkflow Category=Workflow with Router2026.05 | 59.7 | |
| Qwen3-32BWorkflow Category=Single Model2026.05 | 55.9 | |
| MaAS (Qwen3-8B)Workflow Category=Workflow with Router2026.05 | 55.7 | |
| QwQ-32BWorkflow Category=Single Model2026.05 | 54.6 | |
| KNN RouterWorkflow Category=Workflow with Router2026.05 | 53.6 | |
| Qwen3-8BWorkflow Category=Fixed Workflow2026.05 | 49.1 |