Question Answering on HotpotQA (Accuracy and Recall)
65.4AccuracyArchRAG
Evaluation results
| Method | Links | ||
|---|---|---|---|
| ArchRAGBaseline Type=Our proposed2025.02 | 65.4 | 69.2 | |
| HippoRAGBaseline Type=Graph-based RAG2025.02 | 51.3 | 56.8 | |
| Vanilla RAGBaseline Type=Retrieval-only2025.02 | 50.6 | 56.1 | |
| BM25Baseline Type=Retrieval-only2025.02 | 49.7 | 53.6 | |
| HyLightRAGBaseline Type=Graph-based RAG2025.02 | 35.6 | 43.3 | |
| LLightRAGBaseline Type=Graph-based RAG2025.02 | 34.1 | 41.8 | |
| GGraphRAGBaseline Type=Graph-based RAG2025.02 | 33.5 | 42.6 | |
| CoTBaseline Type=Inference-only2025.02 | 32.5 | 39.7 | |
| LGraphRAGBaseline Type=Graph-based RAG2025.02 | 29.7 | 35.5 | |
| Zero-shotBaseline Type=Inference-only2025.02 | 28 | 31.8 | |
| HLightRAGBaseline Type=Graph-based RAG2025.02 | 25.6 | 33.3 |