Multi-hop Question Answering on HotpotQA (top-5 accuracy)
82Top-5 AccuracyMoG-IRCoT
Evaluation results
| Method | Links | |
|---|---|---|
| MoG-IRCoTBase LLM=GPT-4o-mini, Evaluation protocol=top-5 accuracy, Reasoning Strategy=IRCoT2026.05 | 82 | |
| MoGBase LLM=GPT-4o-mini, Evaluation protocol=top-5 accuracy2026.05 | 79.7 | |
| LinearRAGBase LLM=GPT-4o-mini, Evaluation protocol=top-5 accuracy2026.05 | 66.5 | |
| Youtu-GraphRAG-IRCoTBase LLM=GPT-4o-mini, Evaluation protocol=top-5 accuracy, Reasoning Strategy=IRCoT2026.05 | 66.3 | |
| GFM-RAGBase LLM=GPT-4o-mini, Evaluation protocol=top-5 accuracy2026.05 | 65.6 | |
| Youtu-GraphRAGBase LLM=GPT-4o-mini, Evaluation protocol=top-5 accuracy2026.05 | 65.3 | |
| HippoRAG2Base LLM=GPT-4o-mini, Evaluation protocol=top-5 accuracy2026.05 | 64.3 | |
| E2GraphRAGBase LLM=GPT-4o-mini, Evaluation protocol=top-5 accuracy2026.05 | 63.9 | |
| HippoRAGBase LLM=GPT-4o-mini, Evaluation protocol=top-5 accuracy2026.05 | 59.6 | |
| RAPTORBase LLM=GPT-4o-mini, Evaluation protocol=top-5 accuracy2026.05 | 55.3 | |
| LightRAGBase LLM=GPT-4o-mini, Evaluation protocol=top-5 accuracy2026.05 | 52.7 | |
| G-retrieverBase LLM=GPT-4o-mini, Evaluation protocol=top-5 accuracy2026.05 | 41.9 |