Multi-hop Question Answering on HotpotQA N=1,000 (test)
83.5F1 ScoreMGRS
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| MGRSBackbone=GPT-4o-mini2025.11 | 83.5 | — | — | — | — | |
| AoTBackbone=GPT-4o-mini2025.11 | 80.6 | — | — | — | — | |
| AFlowBackbone=GPT-4o-mini2025.11 | 73.5 | — | — | — | — | |
| Self-refineBackbone=GPT-4o-mini2025.11 | 68.3 | — | — | — | — | |
| CoTBackbone=GPT-4o-mini2025.11 | 67.2 | — | — | — | — | |
| FoTBackbone=GPT-4o-mini, n=82025.11 | 66.7 | — | — | — | — | |
| CoT-SCBackbone=GPT-4o-mini, n=52025.11 | 66.2 | — | — | — | — | |
| SEAL-RAGModel=gpt-4.1, k=32025.12 | 0.79 | 0.76 | 0.91 | 0.73 | 3 | |
| SEAL-RAGModel=gpt-4o, k=32025.12 | 0.75 | 0.77 | 0.89 | 0.68 | 6 | |
| Self-RAGModel=gpt-4.1, k=32025.12 | 0.66 | 0.73 | 0.79 | 0.61 | — | |
| Self-RAGModel=gpt-4o, k=32025.12 | 0.61 | 0.71 | 0.76 | 0.55 | — | |
| Self-RAGModel=gpt-4.1-mini, k=32025.12 | 0.61 | 0.64 | 0.73 | 0.56 | — | |
| SEAL-RAGModel=gpt-4.1-mini, k=32025.12 | 0.61 | 0.77 | 0.86 | 0.49 | 10 | |
| Basic RAGModel=gpt-4o-mini, k=32025.12 | 0.59 | 0.63 | 0.49 | 0.72 | — | |
| Basic RAGModel=gpt-4o, k=32025.12 | 0.59 | 0.68 | 0.49 | 0.72 | — | |
| Basic RAGModel=gpt-4.1-mini, k=32025.12 | 0.59 | 0.64 | 0.49 | 0.72 | — | |
| Basic RAGModel=gpt-4.1, k=32025.12 | 0.59 | 0.68 | 0.49 | 0.72 | — | |
| SEAL-RAGModel=gpt-4o-mini, k=32025.12 | 0.57 | 0.69 | 0.84 | 0.44 | 6 | |
| Self-RAGModel=gpt-4o-mini, k=32025.12 | 0.53 | 0.6 | 0.66 | 0.47 | — | |
| CRAGModel=gpt-4.1, k=32025.12 | 0.44 | 0.72 | 0.41 | 0.5 | — | |
| CRAGModel=gpt-4.1-mini, k=32025.12 | 0.43 | 0.67 | 0.4 | 0.49 | — | |
| CRAGModel=gpt-4o, k=32025.12 | 0.4 | 0.69 | 0.37 | 0.44 | — | |
| CRAGModel=gpt-4o-mini, k=32025.12 | 0.33 | 0.62 | 0.3 | 0.36 | — |