Multi-Hop Question Answering on MuSiQue (EM & F1)
33.5Exact Match (EM)QAFD-RAG
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| QAFD-RAGLLM=GPT-4o2026.04 | 33.5 | 47.99 | |
| DTKG (Full)Backbone=Llama 3.1:8B2025.10 | 33.5 | 45.2 | |
| CoRAGBackbone=Llama 3.1:8B2025.10 | 30.9 | 42.4 | |
| HippoRAGLLM=GPT-4o2026.04 | 27.55 | 38.33 | |
| TOGBackbone=Llama 3.1:8B2025.10 | 26.8 | 38.4 | |
| KGRBackbone=Llama 3.1:8B2025.10 | 25.4 | 36.2 | |
| CRITICBackbone=Llama 3.1:8B2025.10 | 23.5 | 33.8 | |
| Self-Correcting RAGCategory=Self-Correcting RAG2026.04 | 22.7 | 31.9 | |
| COTBackbone=Llama 3.1:8B2025.10 | 22.1 | 32.5 | |
| RECOMPCategory=Advanced Selection & Reranking2026.04 | 21.5 | 30.2 | |
| RRRCategory=Standard RAG Baselines2026.04 | 20.5 | 29.5 | |
| DRAGCategory=Iterative & Agentic RAG2026.04 | 20.4 | 28.1 | |
| Self-RAGCategory=Iterative & Agentic RAG2026.04 | 20.2 | 29.1 | |
| RAG + MMRCategory=Advanced Selection & Reranking2026.04 | 19.8 | 28.4 | |
| FilcoCategory=Advanced Selection & Reranking2026.04 | 19.5 | 28.1 | |
| IRCoTCategory=Iterative & Agentic RAG2026.04 | 18.5 | 27.8 | |
| HyDECategory=Standard RAG Baselines2026.04 | 18.2 | 27.5 | |
| CRAGCategory=Iterative & Agentic RAG2026.04 | 18.2 | 27.8 | |
| LongLLMLinguaCategory=Advanced Selection & Reranking2026.04 | 17.8 | 26.5 | |
| GraphRAGLLM=GPT-4o2026.04 | 17.6 | 39.4 | |
| NaiveCategory=Standard RAG Baselines2026.04 | 14.6 | 25.8 | |
| RAPTORLLM=GPT-4o2026.04 | 10.5 | 28.1 | |
| LightRAGLLM=GPT-4o2026.04 | 0.1 | 1.4 |