Multi-hop Question Answering on 2WikiMultiHopQA (Latency and Throughput)
0.58Average Inference Time (s)LinearRAG
Evaluation results
| Method | Links | ||
|---|---|---|---|
| LinearRAGBase Model=GPT-4o-mini2026.05 | 0.58 | 609.4 | |
| Vanilla RAGBase Model=GPT-4o-mini2026.05 | 4.28 | 489.7 | |
| ConRAGBase Model=GPT-4o-mini2026.05 | 5.64 | 2,517.2 | |
| RAPTORBase Model=GPT-4o-mini2026.05 | 5.79 | 2,568 | |
| Zero-shotBase Model=GPT-4o-mini2026.05 | 5.88 | 216.2 | |
| HippoRAG2Base Model=GPT-4o-mini2026.05 | 5.89 | 2,809.2 | |
| HippoRAGBase Model=GPT-4o-mini2026.05 | 6.3 | 2,608.8 | |
| LogicRAGBase Model=GPT-4o-mini2026.05 | 9.83 | 1,777.9 | |
| G-RetrieverBase Model=GPT-4o-mini2026.05 | 12.5 | 1,000 | |
| GraphRAGBase Model=GPT-4o-mini2026.05 | 13.05 | 4,699.8 | |
| LightRAGBase Model=GPT-4o-mini2026.05 | 35.14 | 5,730.6 | |
| KGPBase Model=GPT-4o-mini2026.05 | 70.72 | 11,097.8 | |
| Youtu-GraphRAGBase Model=GPT-4o-mini2026.05 | 71.46 | 12,855.1 |