LLM Routing on LogicGrid, MGSM, and CommonGen (Phase-2 test)
78Average AccuracyGraphPlanner
Evaluation Results
| Method | Links | |
|---|---|---|
| GraphPlannerEvaluation Mode=Zero-shot2026.04 | 78 | |
| RouterDCEvaluation Mode=Zero-shot2026.04 | 58 | |
| GraphRouterEvaluation Mode=Zero-shot2026.04 | 46 | |
| Router-R1Evaluation Mode=Zero-shot2026.04 | 38 |