Evidence Retrieval on 406-paper benchmark (full pool)
44.3Evidence Retrieval (eR)SPIRE (full)
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| SPIRE (full)k=10, agent_configuration=seven-agent2026.05 | 44.3 | 24 | 42.4 | 5.6 | 15.3 | 33.5 | 28.9 | |
| Text RAGk=102026.05 | 22.4 | 14.5 | 12.5 | 2.7 | 3.5 | 6.5 | 4 | |
| Naive LLMk=102026.05 | 14.3 | 14.3 | 17.4 | 3.6 | 4.4 | 15.7 | 15 | |
| GraphRAGk=102026.05 | 12.4 | 12.5 | 13.2 | 2.3 | 3.6 | 6.2 | 8.9 |