Retrieval-Augmented Generation on Average All 6 Datasets
48.57Exact Match (EM)Graph-R1 (ours)
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Graph-R1 (ours)Backbone=Qwen2.5-7B-Instruct, Knowledge Representation=Graph-based, Learning Strategy=Training2025.07 | 48.57 | 57.82 | 60.4 | 76.23 | |
| StandardRAGBackbone=GPT-4o-mini, Knowledge Representation=Chunk-based, Learning Strategy=Prompting2025.07 | 18.1 | 32.05 | 52.68 | 78.07 | |
| HyperGraphRAGBackbone=GPT-4o-mini, Knowledge Representation=Graph-based, Learning Strategy=Prompting2025.07 | 13.15 | 29.4 | 61.82 | 78.92 | |
| NaiveGenerationBackbone=GPT-4o-mini, Knowledge Interaction=No, Learning Strategy=Prompting2025.07 | 11.36 | 25.92 | — | 78.45 |