Multilingual Multi-hop Question Answering on HotpotQA sampled 1,000 questions (val)
44.5Swedish EMDaPT
Evaluation Results
| Method | Links | ||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| DaPT2026.03 | 44.5 | 53.2 | 40.7 | 46.9 | 51.5 | 59.5 | 50.7 | 60.3 | 35.2 | 55.1 | 58.3 | 67.5 | 46.8 | 57.1 | |
| HippoRAG2Type=SA-Baselines2026.03 | 42.2 | 57.4 | 38.9 | 49.2 | 45.7 | 59.8 | 46.2 | 57 | 34.6 | 56.2 | 54 | 69 | 43.6 | 58.1 | |
| Vanilla RAGType=SF-Baselines2026.03 | 35.1 | 48.1 | 38.4 | 47.7 | 44 | 55.5 | 42.9 | 56.2 | 34.9 | 35.5 | 50.4 | 62.3 | 41 | 50.9 | |
| GraphRAGType=SA-Baselines2026.03 | 18.9 | 24.8 | 16.8 | 20.6 | 11.2 | 14.8 | 23.6 | 31.3 | 13.5 | 23.6 | 29 | 39.3 | 18.8 | 25.7 | |
| Zero-shotType=SF-Baselines2026.03 | 17.9 | 28.1 | 17.2 | 21.8 | 23.2 | 32.1 | 23.9 | 34.3 | 15.9 | 33.7 | 27.9 | 38.4 | 21 | 31.4 | |
| CoTType=SF-Baselines2026.03 | 16.8 | 27.3 | 16.7 | 21.2 | 24.6 | 34.6 | 24.3 | 35.1 | 15.3 | 33.5 | 29.8 | 40.5 | 21.3 | 32 |