Knowledge Retrieval on STaRK-PRIME synthetic (test)
63.7Hit@1GraphRAFT
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| GraphRAFTTraining protocol=Supervised / Fine-tuned2025.05 | 63.7 | 75.4 | 69 | |
| AF-RetrieverShot-count=Zero/One-shot, Test Set Scale=Full2025.05 | 46.2 | 63.7 | 54 | |
| mFARTraining protocol=Supervised / Fine-tuned2025.05 | 40.9 | 62.8 | 51.2 | |
| 4StepFocusShot-count=Zero/One-shot, Test Set Scale=10%-version2025.05 | 39.3 | 53.2 | 45.8 | |
| MoRTraining protocol=Supervised / Fine-tuned2025.05 | 36.4 | 60 | 46.9 | |
| KARShot-count=Zero/One-shot, Test Set Scale=Full2025.05 | 30.4 | 49.3 | 39.2 | |
| PASemiQATraining protocol=Supervised / Fine-tuned2025.05 | 29.7 | — | 31 | |
| HybGRAGShot-count=Zero/One-shot, Test Set Scale=Full2025.05 | 28.6 | 41.4 | 34.5 | |
| AvaTaRShot-count=Zero/One-shot, Test Set Scale=Full2025.05 | 18.4 | 36.7 | 26.7 | |
| VSS+[pointw.]RerankerShot-count=Zero/One-shot, Test Set Scale=10%-version2025.05 | 18.3 | 37.3 | 26.6 | |
| ReActShot-count=Zero/One-shot, Test Set Scale=Full2025.05 | 15.3 | 32 | 22.8 | |
| VSS (Multi-Ada)Shot-count=Zero/One-shot, Test Set Scale=Full, Embedding=Multi-Ada2025.05 | 15.1 | 33.6 | 23.5 | |
| ReflexionShot-count=Zero/One-shot, Test Set Scale=Full2025.05 | 14.3 | 35 | 24.8 | |
| BM25Shot-count=Zero/One-shot, Test Set Scale=Full2025.05 | 12.8 | 27.9 | 19.8 | |
| VSS (Ada-002)Shot-count=Zero/One-shot, Test Set Scale=Full, Embedding=Ada-0022025.05 | 12.6 | 31.5 | 21.4 | |
| QA-GNNShot-count=Zero/One-shot, Test Set Scale=Full2025.05 | 8.8 | 21.4 | 14.7 | |
| ToGShot-count=Zero/One-shot, Test Set Scale=Full2025.05 | 6.1 | 15.7 | 10.2 | |
| DPRShot-count=Zero/One-shot, Test Set Scale=Full2025.05 | 4.5 | 21.8 | 12.4 |