Question Answering (EM) on TriviaQA (test)
73.3EMWorkflow-R1-Search
Evaluation Results
| Method | Links | |
|---|---|---|
| Workflow-R1-SearchMulti-turn=true, Backbone=Qwen2.5-32B-Instruct2026.02 | 73.3 | |
| ReCalRouting Strategy=RL-based Routing, Task Decomposition=false2026.06 | 71.2 | |
| ReCal+Routing Strategy=RL-based Routing, Task Decomposition=false2026.06 | 69.8 | |
| Router-R1Routing Strategy=RL-based Routing, Task Decomposition=false2026.06 | 66.8 | |
| Search-R1 (GRPO)Multi-turn=true, Backbone=Qwen2.5-32B-Instruct2026.02 | 62.3 | |
| Workflow-R1Multi-turn=true, Backbone=Qwen2.5-32B-Instruct2026.02 | 61.2 | |
| Search-R1 (PPO)Multi-turn=true, Backbone=Qwen2.5-32B-Instruct2026.02 | 61 | |
| MedPromptMulti-turn=false, Backbone=Qwen2.5-32B-Instruct2026.02 | 60.3 | |
| AFlowMulti-turn=false, Backbone=Qwen2.5-32B-Instruct2026.02 | 60.3 | |
| SC (CoT×5)Multi-turn=false, Backbone=Qwen2.5-32B-Instruct2026.02 | 59.5 | |
| RouterDCRouting Strategy=Heuristic & Discriminative Routing, Task Decomposition=false2026.06 | 59.2 | |
| GraphRouterRouting Strategy=Heuristic & Discriminative Routing, Task Decomposition=false2026.06 | 58.6 | |
| Prompt LLMRouting Strategy=Heuristic & Discriminative Routing, Task Decomposition=false2026.06 | 58 | |
| CoTMulti-turn=false, Backbone=Qwen2.5-32B-Instruct2026.02 | 57.9 | |
| Largest LLMRouting Strategy=Heuristic & Discriminative Routing, Task Decomposition=false2026.06 | 57.8 | |
| MaASMulti-turn=false, Backbone=Qwen2.5-32B-Instruct2026.02 | 57.1 | |
| RAGRouting Strategy=No Routing, Task Decomposition=false2026.06 | 54 | |
| KNN RouterRouting Strategy=Heuristic & Discriminative Routing, Task Decomposition=false2026.06 | 52.8 | |
| RouteLLMRouting Strategy=Heuristic & Discriminative Routing, Task Decomposition=false2026.06 | 51.6 | |
| Search-R1Routing Strategy=No Routing, Task Decomposition=false2026.06 | 51 | |
| Prompt LLMRouting Strategy=Heuristic & Discriminative Routing, Task Decomposition=true2026.06 | 50 | |
| KNN RouterRouting Strategy=Heuristic & Discriminative Routing, Task Decomposition=true2026.06 | 47.8 | |
| DirectMulti-turn=false, Backbone=Qwen2.5-32B-Instruct2026.02 | 47.1 | |
| MLP RouterRouting Strategy=Heuristic & Discriminative Routing, Task Decomposition=false2026.06 | 46 | |
| SFTRouting Strategy=No Routing, Task Decomposition=false2026.06 | 40 | |
| CoTRouting Strategy=No Routing, Task Decomposition=false2026.06 | 35.8 | |
| VanillaRouting Strategy=No Routing, Task Decomposition=false2026.06 | 26 |