Question Answering on NarrativeQA (EM, F1)
40.28F1 ScoreDocTrace
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DocTraceCategory=Ours2026.06 | 40.28 | 23 | |
| DocTrace (w/o DM)Category=Ours2026.06 | 37.95 | 21.8 | |
| ComoRAGCategory=Hybrid Structured RAG2026.06 | 31.43 | 18.6 | |
| RAPTOR+IRCoTCategory=Summary-based RAG2026.06 | 31.35 | 16 | |
| HippoRAG2+IRCoTCategory=Graph-based RAG2026.06 | 28.98 | 13 | |
| RAPTORCategory=Summary-based RAG2026.06 | 27.84 | 17.8 | |
| GPT-4o-miniCategory=LLM-only2026.06 | 27.29 | 7 | |
| Naive RAGCategory=Flat RAG2026.06 | 27.18 | 17.8 | |
| HippoRAG2Category=Graph-based RAG2026.06 | 23.12 | 15.2 | |
| Self-RAGCategory=Flat RAG2026.06 | 19.6 | 6.4 |