General Question Answering on PopQA (F1 score)
50.5F1 ScoreReCal
Evaluation Results
| Method | Links | |
|---|---|---|
| ReCalRouting=RL-based Routing2026.06 | 50.5 | |
| ReCalRouting=RL-based Routing, Variant=+2026.06 | 50.2 | |
| Router-R1Routing=RL-based Routing2026.06 | 45.8 | |
| RAGRouting=No Routing2026.06 | 45.2 | |
| Largest LLMRouting=Heuristic & Discriminative Routing2026.06 | 42.3 | |
| Prompt LLMRouting=Heuristic & Discriminative Routing2026.06 | 40.9 | |
| Search-R1Routing=No Routing2026.06 | 38.3 | |
| RouterDCRouting=Heuristic & Discriminative Routing2026.06 | 32.8 | |
| GraphRouterRouting=Heuristic & Discriminative Routing2026.06 | 32.4 | |
| Prompt LLM*Routing=Heuristic & Discriminative Routing2026.06 | 31.5 | |
| KNN RouterRouting=Heuristic & Discriminative Routing2026.06 | 28.1 | |
| KNN Router*Routing=Heuristic & Discriminative Routing2026.06 | 27.2 | |
| MLP RouterRouting=Heuristic & Discriminative Routing2026.06 | 27.2 | |
| RouteLLMRouting=Heuristic & Discriminative Routing2026.06 | 25.1 | |
| SFTRouting=No Routing2026.06 | 20.7 | |
| CoTRouting=No Routing2026.06 | 18.5 | |
| VanillaRouting=No Routing2026.06 | 15.4 |