Multi-hop Question Answering on MuSiQue (F1, ACC_L)
53.03F1 ScoreBubbleRAG
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| BubbleRAGModel Scale=30B2026.03 | 53.03 | 53.1 | |
| BubbleRAGModel Scale=8B2026.03 | 45.15 | 44 | |
| HippoRAG2Model Scale=30B2026.03 | 45.04 | 47.2 | |
| Naive RAGModel Scale=30B2026.03 | 43.68 | 42.2 | |
| HippoRAG2Model Scale=8B2026.03 | 42.92 | 41.6 | |
| LightRAG (local)Model Scale=30B2026.03 | 41.72 | 43.8 | |
| LightRAG (hybrid)Model Scale=30B2026.03 | 39.57 | 42.2 | |
| LightRAG (local)Model Scale=8B2026.03 | 36.57 | 37.7 | |
| Naive RAGModel Scale=8B2026.03 | 34.53 | 35 | |
| LightRAG (hybrid)Model Scale=8B2026.03 | 34.21 | 35.1 | |
| RAPTORModel Scale=30B2026.03 | 33.1 | 32.2 | |
| Clue-RAGModel Scale=30B2026.03 | 33.01 | 33.9 | |
| RAPTORModel Scale=8B2026.03 | 29.35 | 28.9 | |
| Clue-RAGModel Scale=8B2026.03 | 27.39 | 27.8 | |
| LightRAG (global)Model Scale=30B2026.03 | 20.56 | 22.5 | |
| LightRAG (global)Model Scale=8B2026.03 | 20.11 | 20.3 | |
| ToGModel Scale=30B2026.03 | 16.76 | 21.6 | |
| CoT + LLMModel Scale=30B2026.03 | 8.42 | 13.2 | |
| ToGModel Scale=8B2026.03 | 8.39 | 13.9 | |
| CoT + LLMModel Scale=8B2026.03 | 7.11 | 12 | |
| Vanilla LLMModel Scale=30B2026.03 | 5.8 | 10.4 | |
| Vanilla LLMModel Scale=8B2026.03 | 5.18 | 12.9 |