Question Answering on 2WikiMultihopQA (LLM-Acc/Cont-Acc)
89.7LLM-AccA-RAG (Full)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| A-RAG (Full)Backbone LLM=GPT-5-mini2026.02 | 89.7 | 88.9 | |
| LinearRAGBackbone LLM=GPT-5-mini2026.02 | 87.2 | 84.8 | |
| HippoRAG2Backbone LLM=GPT-5-mini2026.02 | 82 | 79.7 | |
| A-RAG (Naive)Backbone LLM=GPT-5-mini2026.02 | 70.6 | 76.9 | |
| GraphRAGBackbone LLM=GPT-5-mini2026.02 | 66.5 | 70.7 | |
| HippoRAG2Backbone LLM=GPT-4o-mini2026.02 | 64.7 | 68.5 | |
| LinearRAGBackbone LLM=GPT-4o-mini2026.02 | 62.9 | 62.3 | |
| A-RAG (Full)Backbone LLM=GPT-4o-mini2026.02 | 60.2 | 63.7 | |
| MA-RAGBackbone LLM=GPT-5-mini2026.02 | 54.7 | 54.3 | |
| A-RAG (Naive)Backbone LLM=GPT-4o-mini2026.02 | 52.3 | 62.4 | |
| FaithfulRAGBackbone LLM=GPT-5-mini2026.02 | 51.8 | 56.6 | |
| Direct AnswerBackbone LLM=GPT-5-mini2026.02 | 51.3 | 54 | |
| MA-RAGBackbone LLM=GPT-4o-mini2026.02 | 51 | 53.4 | |
| Naive RAGBackbone LLM=GPT-5-mini2026.02 | 50.2 | 66.5 | |
| Naive RAGBackbone LLM=GPT-4o-mini2026.02 | 42.6 | 59 | |
| FaithfulRAGBackbone LLM=GPT-4o-mini2026.02 | 38.8 | 38.1 | |
| Direct AnswerBackbone LLM=GPT-4o-mini2026.02 | 30.3 | 49.7 | |
| RAGentABackbone LLM=GPT-4o-mini2026.02 | 27.7 | 50.3 | |
| RAGentABackbone LLM=GPT-5-mini2026.02 | 24 | 53.5 | |
| GraphRAGBackbone LLM=GPT-4o-mini2026.02 | 18.4 | 47.2 |