Multilingual Multi-hop Question Answering on 2WikiMultiHopQA (sampled 1k val)
47.4EM (Swedish)HippoRAG2
Evaluation results
| Method | Links | ||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| HippoRAG2Type=SA-Baselines2026.03 | 47.4 | 55.7 | 46.2 | 53.7 | 52.7 | 61.7 | 57 | 65.3 | 30 | 43.8 | 58.9 | 68.1 | 48.7 | 58.1 | |
| DaPT2026.03 | 46.5 | 53.1 | 45.4 | 51.6 | 55.1 | 59.1 | 57.5 | 62.8 | 39 | 51.1 | 56.8 | 67.2 | 50.1 | 57.5 | |
| Vanilla RAGType=SF-Baselines2026.03 | 23.6 | 37.9 | 37.5 | 41 | 38.1 | 43.1 | 38.4 | 43.5 | 36.7 | 36.8 | 38.5 | 43.7 | 37 | 41 | |
| Zero-shotType=SF-Baselines2026.03 | 22.5 | 26.1 | 27 | 28.5 | 28.4 | 32.5 | 27 | 31 | 21.5 | 21.6 | 26 | 31.4 | 25.4 | 28.5 | |
| GraphRAGType=SA-Baselines2026.03 | 22.3 | 25.2 | 24.8 | 27.1 | 29.3 | 32.9 | 27.4 | 31.4 | 24.2 | 32.7 | 35.4 | 40.1 | 27.2 | 31.6 | |
| CoTType=SF-Baselines2026.03 | 16.8 | 27 | 16.8 | 21.5 | 27.8 | 31.2 | 25.7 | 30.9 | 23.9 | 34 | 28.8 | 36.1 | 23.3 | 30.1 |