Multi-Hop Question Answering on HotpotQA (Accuracy, Recall)
69.1AccuracyHyGRAG +Inference
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| HyGRAG +InferenceBackbone=Llama-3.1-8B-Instruct2026.06 | 69.1 | 71.08 | |
| HyGRAGBackbone=Llama-3.1-8B-Instruct2026.06 | 68.72 | 70.79 | |
| RAPTOR-KBackbone=Llama-3.1-8B-Instruct2026.06 | 61.24 | 65.19 | |
| EraRAGBackbone=Llama-3.1-8B-Instruct2026.06 | 61.03 | 65.22 | |
| VanillaRAGBackbone=Llama-3.1-8B-Instruct2026.06 | 60.97 | 64.24 | |
| HippoRAGBackbone=Llama-3.1-8B-Instruct2026.06 | 51.89 | 56.31 | |
| CoTBackbone=Llama-3.1-8B-Instruct2026.06 | 27.77 | 38.88 | |
| Zero-shotBackbone=Llama-3.1-8B-Instruct2026.06 | 27.36 | 32.97 |