Multi-Hop Question Answering on 2wiki (F1 score)
61.3F1 ScoreReCal
Evaluation Results
| Method | Links | |
|---|---|---|
| ReCalRouting=RL-based Routing2026.06 | 61.3 | |
| ReCalRouting=RL-based Routing, Variant=+2026.06 | 61.3 | |
| Router-R1Routing=RL-based Routing2026.06 | 52.9 | |
| Largest LLMRouting=Heuristic & Discriminative Routing2026.06 | 39.7 | |
| Prompt LLMRouting=Heuristic & Discriminative Routing2026.06 | 39.3 | |
| Prompt LLM*Routing=Heuristic & Discriminative Routing2026.06 | 35.5 | |
| Search-R1Routing=No Routing2026.06 | 31.7 | |
| VanillaRouting=No Routing2026.06 | 30.4 | |
| KNN Router*Routing=Heuristic & Discriminative Routing2026.06 | 29.9 | |
| RouterDCRouting=Heuristic & Discriminative Routing2026.06 | 29.3 | |
| SFTRouting=No Routing2026.06 | 29.1 | |
| KNN RouterRouting=Heuristic & Discriminative Routing2026.06 | 28.9 | |
| RouteLLMRouting=Heuristic & Discriminative Routing2026.06 | 28.4 | |
| MLP RouterRouting=Heuristic & Discriminative Routing2026.06 | 27.7 | |
| GraphRouterRouting=Heuristic & Discriminative Routing2026.06 | 25.8 | |
| CoTRouting=No Routing2026.06 | 25.1 | |
| RAGRouting=No Routing2026.06 | 18.7 |