Multi-Hop Knowledge Graph Question Answering on CWQ
81.4Hits@1PoG
Evaluation results
| Method | Links | ||
|---|---|---|---|
| PoGClass=ICL, LLM=GPT-4, External Knowledge=With external knowledge2024.10 | 81.4 | — | |
| Logits-to-LogicBackbone LLM=LLaMA3.1-8b2025.11 | 80.8 | — | |
| Chain-of-QuestionLLM=GPT-3.5-turbo, Question Decomposition Model=T52025.06 | 78.8 | — | |
| PoG-EClass=ICL, LLM=GPT-4, External Knowledge=With external knowledge2024.10 | 78.5 | — | |
| CoGBackbone=GPT-4, Evaluation Protocol=Prompting, KG-Augmented=true2026.01 | 77.8 | — | |
| GCRMethod Paradigm=Agentic Reasoning, Backbone LLM=LLaMA3.1-8b2025.11 | 75.8 | — | |
| GoGMethod Paradigm=Agentic Reasoning2025.11 | 75.2 | — | |
| PoGBackbone=GPT-4, Evaluation Protocol=Prompting, KG-Augmented=true2026.01 | 75 | — | |
| PoGClass=ICL, LLM=GPT-3.5-Turbo, External Knowledge=With external knowledge2024.10 | 74.7 | — | |
| FD-PORTCategory=Agentic Search, Backbone=LLaMA3-8B2026.07 | 74.5 | — | |
| iQUESTBackbone=GPT-4o2025.06 | 73.85 | — | |
| Question-Decomposion2025.06 | 72.8 | — | |
| KG-AgentEvaluation Protocol=Fine-Tuning, KG-Augmented=true2026.01 | 72.2 | — | |
| KG-AgentMethod Paradigm=Agentic Reasoning2025.11 | 72.2 | — | |
| PoG-EClass=ICL, LLM=GPT-3.5-Turbo, External Knowledge=With external knowledge2024.10 | 71.9 | — | |
| Prior FT SOTAClass=SL, External Knowledge=With external knowledge2024.10 | 70.4 | — | |
| DECAFEvaluation Protocol=Fine-Tuning, KG-Augmented=true2026.01 | 70.4 | — | |
| DECAFCategory=LLM + KG, Backbone=FiD-3B2026.07 | 70.4 | — | |
| ToG/ToG-RClass=ICL, LLM=GPT-4, External Knowledge=With external knowledge2024.10 | 69.5 | — | |
| ToGType=KG-centric RAG, LLM=GPT-42026.01 | 69.5 | — | |
| ToG2025.06 | 69.5 | — | |
| ToG-RMethod Paradigm=Agentic Reasoning, Backbone LLM=GPT42025.11 | 69.5 | — | |
| EffiQACategory=Agentic Search, Backbone=Llama3.1-8B2026.07 | 69.5 | — | |
| ToGBackbone=GPT-4, Evaluation Protocol=Prompting, KG-Augmented=true2026.01 | 67.6 | — | |
| ToGMethod Paradigm=Agentic Reasoning, Backbone LLM=GPT42025.11 | 67.6 | — | |
| RSF-GLLMCategory=Ours, Backbone=Qwen3-8B2026.07 | 67.39 | 61.87 | |
| CoGBackbone=GPT-3.5, Evaluation Protocol=Prompting, KG-Augmented=true2026.01 | 66.9 | — | |
| RSF-GLLMCategory=Ours, Backbone=LLaMA2-7B2026.07 | 66.44 | 60.8 | |
| PrivGemoType=Proposed, Brain LLM=GPT-3.5-Turbo, Hand LLM=Qwen3-32b2026.01 | 66.2 | — | |
| PrivGemoType=Proposed, Brain LLM=DeepSeek-V3, Hand LLM=Qwen3-32b2026.01 | 63.3 | — | |
| PoGBackbone=GPT-3.5, Evaluation Protocol=Prompting, KG-Augmented=true2026.01 | 63.2 | — | |
| PoGType=KG-centric RAG, LLM=GPT-3.5-Turbo2026.01 | 63.2 | — | |
| PoGMethod Paradigm=Agentic Reasoning, Backbone LLM=ChatGPT2025.11 | 63.2 | — | |
| GNN-RAG + RACategory=LLM + KG, Backbone=LLaMA2-7B2026.07 | 62.8 | 60.4 | |
| RoGEvaluation Protocol=Fine-Tuning, KG-Augmented=true2026.01 | 62.6 | — | |
| RoGMethod Paradigm=Agentic Reasoning2025.11 | 62.6 | — | |
| RoGCategory=LLM + KG, Backbone=LLaMA2-7B2026.07 | 62.6 | 56.2 | |
| KG-CoT2025.06 | 62.3 | — | |
| KG-CoTMethod Paradigm=Agentic Reasoning, Backbone LLM=ChatGPT2025.11 | 62.3 | — | |
| PrivGemoType=Proposed, Brain LLM=GPT-4o-mini, Hand LLM=Qwen3-32b2026.01 | 62.2 | — | |
| GNN-RAGCategory=LLM + KG, Backbone=LLaMA2-7B2026.07 | 61.7 | 59.4 | |
| ARoGType=Privacy-aware KGQA, LLM=GPT-4o-mini2026.01 | 60 | — | |
| ToG/ToG-RClass=ICL, LLM=GPT-3.5-Turbo, External Knowledge=With external knowledge2024.10 | 58.9 | — | |
| ToGType=KG-centric RAG, LLM=GPT-3.5-Turbo2026.01 | 58.9 | — | |
| ToG-RMethod Paradigm=Agentic Reasoning, Backbone LLM=ChatGPT2025.11 | 58.9 | — | |
| SymAgentMethod Paradigm=Agentic Reasoning2025.11 | 58.8 | — | |
| ToGBackbone=GPT-3.5, Evaluation Protocol=Prompting, KG-Augmented=true2026.01 | 57.1 | — | |
| ToGMethod Paradigm=Agentic Reasoning, Backbone LLM=ChatGPT2025.11 | 57.1 | — | |
| DoGMethod Paradigm=Agentic Reasoning2025.11 | 56 | — | |
| GoGType=KG-centric RAG, LLM=GPT-3.5-Turbo2026.01 | 55.7 | — | |
| KD-CoTCategory=LLM + KG2026.07 | 55.7 | — | |
| StructGPTBackbone=GPT-3.5, Evaluation Protocol=Prompting, KG-Augmented=true2026.01 | 54.3 | — | |
| StructGPTMethod Paradigm=Agentic Reasoning2025.11 | 54.3 | — | |
| CoGBackbone=Qwen2.5-7B, Evaluation Protocol=Prompting, KG-Augmented=true2026.01 | 54 | — | |
| ReaRevCategory=Graph Retrieval2026.07 | 52.9 | 47.8 | |
| UniKGQAEvaluation Protocol=Fine-Tuning, KG-Augmented=true2026.01 | 51.2 | — | |
| UniKGQACategory=Graph Retrieval2026.07 | 51.2 | 49.1 | |
| KD-CoTBackbone=GPT-3.5, Evaluation Protocol=Prompting, KG-Augmented=true2026.01 | 50.5 | — | |
| KD-CoTMethod Paradigm=Agentic Reasoning2025.11 | 50.5 | — | |
| RE-KBQAEvaluation Protocol=Fine-Tuning, KG-Augmented=true2026.01 | 50.3 | — | |
| SR+NSMCategory=Graph Retrieval2026.07 | 50.2 | 47.1 | |
| Interactive-KBQA2025.06 | 49.07 | — | |
| TransferNetCategory=Embedding-based2026.07 | 48.6 | — | |
| NSMCategory=Embedding-based2026.07 | 47.6 | 42.4 | |
| PoGBackbone=Qwen2.5-7B, Evaluation Protocol=Prompting, KG-Augmented=true2026.01 | 46 | — | |
| EmbedKGQACategory=Embedding-based2026.07 | 45.9 | — | |
| PullNetCategory=Graph Retrieval2026.07 | 45.9 | — | |
| SCLLM=GPT-3.5-Turbo, External Knowledge=Without external knowledge2024.10 | 45.4 | — | |
| SCBackbone=GPT-3.5, Evaluation Protocol=Prompting, KG-Augmented=false2026.01 | 45.4 | — | |
| SCType=LLM-only, LLM=GPT-3.5-Turbo2026.01 | 45.4 | — | |
| SCMethod Paradigm=LLMs Reasoning, Backbone LLM=ChatGPT2025.11 | 45.4 | — | |
| ToGBackbone=Qwen2.5-7B, Evaluation Protocol=Prompting, KG-Augmented=true2026.01 | 42.5 | — | |
| CoTLLM=GPT-3.5-Turbo, External Knowledge=Without external knowledge2024.10 | 38.8 | — | |
| CoTBackbone=GPT-3.5, Evaluation Protocol=Prompting, KG-Augmented=false2026.01 | 38.8 | — | |
| CoTType=LLM-only, LLM=GPT-3.5-Turbo2026.01 | 38.8 | — | |
| CoTMethod Paradigm=LLMs Reasoning, Backbone LLM=ChatGPT2025.11 | 38.8 | — | |
| IO promptLLM=GPT-3.5-Turbo, External Knowledge=Without external knowledge2024.10 | 37.6 | — | |
| IO PromptBackbone=GPT-3.5, Evaluation Protocol=Prompting, KG-Augmented=false2026.01 | 37.6 | — | |
| IO promptType=LLM-only, LLM=GPT-3.5-Turbo2026.01 | 37.6 | — | |
| IO promptMethod Paradigm=LLMs Reasoning, Backbone LLM=ChatGPT2025.11 | 37.6 | — | |
| GraftNetCategory=Graph Retrieval2026.07 | 36.8 | 32.7 | |
| LLaMA3.1-8BCategory=LLM Reasoning, Zero-shot=true2026.07 | 27.7 | 22.8 | |
| Qwen3-8BCategory=LLM Reasoning, Zero-shot=true2026.07 | 27.5 | 25.8 | |
| KV-MemCategory=Embedding-based2026.07 | 18.4 | 15.7 |