Question Answering on 2WikiMQA (test)
44.3Unreliability RateKbSD
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| KbSDBackbone=Qwen2.5-7B, Paradigm=RL-based Agentic RAG2026.06 | 44.3 | — | |
| KbSDBackbone=Qwen2.5-3B, Paradigm=RL-based Agentic RAG2026.06 | 45.9 | — | |
| IRCOTBackbone=Qwen2.5-7B, Paradigm=Agentic RAG2026.06 | 63.55 | — | |
| Search-R1Backbone=Qwen2.5-7B, Paradigm=RL-based Agentic RAG2026.06 | 64.97 | — | |
| KbPOBackbone=Qwen2.5-7B, Paradigm=RL-based Agentic RAG2026.06 | 65.2 | — | |
| ARPOBackbone=Qwen2.5-7B, Paradigm=RL-based Agentic RAG2026.06 | 69.29 | — | |
| ReSearchBackbone=Qwen2.5-7B, Paradigm=RL-based Agentic RAG2026.06 | 69.97 | — | |
| Search-R1Backbone=Qwen2.5-3B, Paradigm=RL-based Agentic RAG2026.06 | 70.1 | — | |
| TCRAGBackbone=Qwen2.5-7B, Paradigm=Agentic RAG2026.06 | 70.3 | — | |
| ARPOBackbone=Qwen2.5-3B, Paradigm=RL-based Agentic RAG2026.06 | 70.45 | — | |
| KbPOBackbone=Qwen2.5-3B, Paradigm=RL-based Agentic RAG2026.06 | 70.5 | — | |
| TCRAGBackbone=Qwen2.5-3B, Paradigm=Agentic RAG2026.06 | 71.53 | — | |
| ReACTBackbone=Qwen2.5-7B, Paradigm=Agentic RAG2026.06 | 72.49 | — | |
| ReSearchBackbone=Qwen2.5-3B, Paradigm=RL-based Agentic RAG2026.06 | 72.77 | — | |
| BaseBackbone=Qwen2.5-7B, Paradigm=No RAG2026.06 | 74.59 | — | |
| Mem1Backbone=Qwen2.5-7B, Paradigm=RL-based Agentic RAG2026.06 | 74.71 | — | |
| ReACTBackbone=Qwen2.5-3B, Paradigm=Agentic RAG2026.06 | 74.91 | — | |
| BaseBackbone=Qwen2.5-3B, Paradigm=No RAG2026.06 | 76.02 | — | |
| COTBackbone=Qwen2.5-7B, Paradigm=No RAG2026.06 | 76.45 | — | |
| AEPOBackbone=Qwen2.5-3B, Paradigm=RL-based Agentic RAG2026.06 | 76.99 | — | |
| AEPOBackbone=Qwen2.5-7B, Paradigm=RL-based Agentic RAG2026.06 | 80.12 | — | |
| FL-RAGBackbone=Qwen2.5-7B, Paradigm=Naive RAG2026.06 | 80.22 | — | |
| COTBackbone=Qwen2.5-3B, Paradigm=No RAG2026.06 | 81.1 | — | |
| Mem1Backbone=Qwen2.5-3B, Paradigm=RL-based Agentic RAG2026.06 | 81.94 | — | |
| FS-RAGBackbone=Qwen2.5-7B, Paradigm=Naive RAG2026.06 | 82.29 | — | |
| FL-RAGBackbone=Qwen2.5-3B, Paradigm=Naive RAG2026.06 | 83.2 | — | |
| IRCOTBackbone=Qwen2.5-3B, Paradigm=Agentic RAG2026.06 | 84.11 | — | |
| FS-RAGBackbone=Qwen2.5-3B, Paradigm=Naive RAG2026.06 | 84.53 | — | |
| DirectBackbone=Qwen-2.5-3B, Category=Model Inference2025.12 | — | 24.4 | |
| DirectBackbone=Qwen-2.5-7B, Category=Model Inference2025.12 | — | 25 | |
| Dr. Zero2026.05 | — | 7.7 | |
| EVE-Agent2026.05 | — | 17.6 | |
| GRPOBackbone=Qwen-2.5-3B, Category=Model Inference2025.12 | — | 26.8 | |
| GRPOBackbone=Qwen-2.5-7B, Category=Model Inference2025.12 | — | 27.3 | |
| Initialsearch=disabled2026.05 | — | 3.4 | |
| Initialsearch=enabled2026.05 | — | 1.3 | |
| INSPOBackbone=Qwen-2.5-3B, Category=Tool-integrated Methods2025.12 | — | 35.9 | |
| INSPOBackbone=Qwen-2.5-7B, Category=Tool-integrated Methods2025.12 | — | 34.9 | |
| IRCoTBackbone=Qwen-2.5-3B, Category=Tool-integrated Methods2025.12 | — | 17.1 | |
| IRCoTBackbone=Qwen-2.5-7B, Category=Tool-integrated Methods2025.12 | — | 14.9 | |
| RAGBackbone=Qwen-2.5-3B, Category=Tool-integrated Methods2025.12 | — | 22.6 | |
| RAGBackbone=Qwen-2.5-7B, Category=Tool-integrated Methods2025.12 | — | 23.5 | |
| Search-o1Backbone=Qwen-2.5-3B, Category=Tool-integrated Methods2025.12 | — | 21.8 | |
| Search-o1Backbone=Qwen-2.5-7B, Category=Tool-integrated Methods2025.12 | — | 17.6 | |
| Search-R1Backbone=Qwen-2.5-3B, Category=Tool-integrated Methods2025.12 | — | 28.1 | |
| Search-R1Backbone=Qwen-2.5-7B, Category=Tool-integrated Methods2025.12 | — | 30.8 | |
| SFTBackbone=Qwen-2.5-3B, Category=Model Inference2025.12 | — | 24.8 | |
| SFTBackbone=Qwen-2.5-3B, Category=Tool-integrated Methods2025.12 | — | 23.3 | |
| SFTBackbone=Qwen-2.5-7B, Category=Model Inference2025.12 | — | 25.9 | |
| SFTBackbone=Qwen-2.5-7B, Category=Tool-integrated Methods2025.12 | — | 29.6 |