Multi-hop QA Retrieval on MuSiQue (R@2, R@5, Avg. Rank)
47.7R@2PropRAG
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| PropRAG2026.05 | 47.7 | 57.9 | 1.9 | |
| HippoRAG 22026.05 | 47 | 56.7 | 2.4 | |
| HippoRAG + IRCoTreasoning_strategy=IRCoT2026.05 | 45.3 | 57.6 | 3.6 | |
| FLARE2026.05 | 44.3 | 55.1 | 6.5 | |
| GFM-RAG2026.05 | 43.5 | 57.6 | 2.9 | |
| SAGE2026.05 | 43.2 | 53.1 | 7 | |
| SubgraphRAG2026.05 | 42.1 | 49.3 | 9.7 | |
| ColBERTv2 + IRCoTreasoning_strategy=IRCoT2026.05 | 41.7 | 53.7 | 7.2 | |
| HippoRAG2026.05 | 41.2 | 53.2 | 8.7 | |
| Contriever + IRCoTreasoning_strategy=IRCoT2026.05 | 39.1 | 52.2 | 10.7 | |
| G-Retriever2026.05 | 38.8 | 45.1 | 15.7 | |
| ColBERTv22026.05 | 37.9 | 49.2 | 11.5 | |
| Proposition2026.05 | 37.6 | 49.3 | 14.8 | |
| GTR2026.05 | 37.4 | 49.1 | 13.8 | |
| RAPTOR2026.05 | 35.7 | 45.3 | 17.7 | |
| GraphRAG2026.05 | 35.4 | 49.3 | 12.2 | |
| Adaptive-RAG2026.05 | 35.1 | 44.7 | 16.6 | |
| Contriever2026.05 | 34.8 | 46.6 | 17.2 | |
| BM25 + IRCoTreasoning_strategy=IRCoT2026.05 | 34.2 | 44.7 | 12.4 | |
| BM252026.05 | 32.3 | 41.2 | 18.2 | |
| LightRAG2026.05 | 24.8 | 34.7 | 20.5 |