Multi-hop Question Answering on 2WikiMultiHopQA N=200
77Judge EMSEAL-RAG
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| SEAL-RAGModel=GPT-4o, Retrieval Depth (k)=32025.12 | 77 | 97 | 77 | 84 | 17 | |
| SEAL-RAGModel=GPT-4o, Retrieval Depth (k)=12025.12 | 76 | 95 | 75 | 82 | 40 | |
| SEAL-RAGModel=GPT-4o, Retrieval Depth (k)=52025.12 | 74 | 96 | 77 | 84 | 10 | |
| SEAL-RAGModel=GPT-4o-mini, Retrieval Depth (k)=52025.12 | 68 | 89 | 45 | 59 | 11 | |
| SEAL-RAGModel=GPT-4o-mini, Retrieval Depth (k)=32025.12 | 64 | 91 | 46 | 60 | 8 | |
| CRAGModel=GPT-4o, Retrieval Depth (k)=52025.12 | 64 | 22 | 23 | 22 | — | |
| Basic RAGModel=GPT-4o, Retrieval Depth (k)=52025.12 | 62 | 34 | 75 | 46 | — | |
| SEAL-RAGModel=GPT-4o-mini, Retrieval Depth (k)=12025.12 | 61 | 92 | 45 | 59 | 31 | |
| CRAGModel=GPT-4o, Retrieval Depth (k)=32025.12 | 60 | 18 | 19 | 18 | — | |
| Self-RAGModel=GPT-4o, Retrieval Depth (k)=52025.12 | 60 | 63 | 38 | 45 | — | |
| Basic RAGModel=GPT-4o-mini, Retrieval Depth (k)=52025.12 | 57 | 34 | 75 | 46 | — | |
| CRAGModel=GPT-4o-mini, Retrieval Depth (k)=32025.12 | 56 | 9 | 9 | 9 | — | |
| Self-RAGModel=GPT-4o, Retrieval Depth (k)=32025.12 | 56 | 60 | 35 | 42 | — | |
| Self-RAGModel=GPT-4o-mini, Retrieval Depth (k)=52025.12 | 56 | 45 | 20 | 26 | — | |
| CRAGModel=GPT-4o-mini, Retrieval Depth (k)=52025.12 | 55 | 11 | 10 | 11 | — | |
| Basic RAGModel=GPT-4o, Retrieval Depth (k)=32025.12 | 54 | 53 | 69 | 59 | — | |
| Self-RAGModel=GPT-4o-mini, Retrieval Depth (k)=32025.12 | 53 | 41 | 18 | 25 | — | |
| Basic RAGModel=GPT-4o-mini, Retrieval Depth (k)=32025.12 | 49 | 53 | 69 | 59 | — | |
| Self-RAGModel=GPT-4o, Retrieval Depth (k)=12025.12 | 36 | 41 | 19 | 26 | — | |
| CRAGModel=GPT-4o-mini, Retrieval Depth (k)=12025.12 | 30 | 12 | 5 | 7 | — | |
| CRAGModel=GPT-4o, Retrieval Depth (k)=12025.12 | 28 | 21 | 10 | 14 | — | |
| Self-RAGModel=GPT-4o-mini, Retrieval Depth (k)=12025.12 | 26 | 27 | 10 | 15 | — | |
| Basic RAGModel=GPT-4o-mini, Retrieval Depth (k)=12025.12 | 18 | 93 | 41 | 56 | — | |
| Basic RAGModel=GPT-4o, Retrieval Depth (k)=12025.12 | 14 | 93 | 41 | 56 | — |