Multi-turn Conversational Question Answering on multi-turn conversational dataset Tier 4
76CorrectnessAgentic GraphRAG
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Agentic GraphRAG2026.04 | 76 | 84.7 | 74.2 | 75 | 68.3 | 90 | |
| Vector RAG Baseline2026.04 | 34.4 | 64.4 | 31.7 | 33.3 | 43.3 | — |