Question Answering on Average (NQ, PopQA, MuSiQue, 2Wiki, HotpotQA, MultiHop-RAG)
37.1EMSelf-Correcting RAG
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Self-Correcting RAGCategory=Self-Correcting RAG2026.04 | 37.1 | 45.8 | |
| RECOMPCategory=Advanced Selection & Reranking2026.04 | 34.8 | 43.3 | |
| RAG + MMRCategory=Advanced Selection & Reranking2026.04 | 34.3 | 42.9 | |
| CRAGCategory=Iterative & Agentic RAG2026.04 | 34.3 | 43.3 | |
| FilcoCategory=Advanced Selection & Reranking2026.04 | 33.7 | 42.2 | |
| IRCoTCategory=Iterative & Agentic RAG2026.04 | 33.3 | 41.9 | |
| RRRCategory=Standard RAG Baselines2026.04 | 32.3 | 41 | |
| Self-RAGCategory=Iterative & Agentic RAG2026.04 | 32.2 | 41.2 | |
| LongLLMLinguaCategory=Advanced Selection & Reranking2026.04 | 31.6 | 40 | |
| DRAGCategory=Iterative & Agentic RAG2026.04 | 30.8 | 39 | |
| HyDECategory=Standard RAG Baselines2026.04 | 30.7 | 39.2 | |
| NaiveCategory=Standard RAG Baselines2026.04 | 25.8 | 36.1 |