Multi-hop QA on 2Wiki
62EMHELP
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| HELPBackbone=Qwen3-30B-A3B-Instruct-25072026.02 | 62 | 69.5 | |
| HippoRAG2Backbone=Qwen3-30B-A3B-Instruct-25072026.02 | 60.2 | 69.4 | |
| HyperGraphRAGBackbone=Qwen3-30B-A3B-Instruct-25072026.02 | 45.5 | 58.6 | |
| LinearRAGBackbone=Qwen3-30B-A3B-Instruct-25072026.02 | 42.4 | 53.2 | |
| Search-R1Backbone=Qwen2.5-7b-Instruct2025.10 | 41.4 | — | |
| ReSeekBackbone=Qwen2.5-7b-Instruct2025.10 | 38.2 | — | |
| ZeroSearchBackbone=Qwen2.5-7b-Instruct2025.10 | 35.2 | — | |
| Search-R1Backbone=Qwen2.5-3b-Instruct2025.10 | 31.9 | — | |
| ZeroSearchBackbone=Qwen2.5-3b-Instruct2025.10 | 30 | — | |
| ReSeekBackbone=Qwen2.5-3b-Instruct2025.10 | 29.8 | — | |
| R1Backbone=Qwen2.5-7b-Instruct2025.10 | 29.2 | — | |
| R1Backbone=Qwen2.5-3b-Instruct2025.10 | 27.5 | — | |
| Direct InferenceBackbone=Qwen2.5-7b-Instruct2025.10 | 25 | — | |
| Direct InferenceBackbone=Qwen2.5-3b-Instruct2025.10 | 24.4 | — | |
| RAGBackbone=Qwen2.5-7b-Instruct2025.10 | 23.5 | — | |
| RAGBackbone=Qwen2.5-3b-Instruct2025.10 | 22.6 | — | |
| Search-o1Backbone=Qwen2.5-3b-Instruct2025.10 | 21.8 | — | |
| Search-o1Backbone=Qwen2.5-7b-Instruct2025.10 | 17.6 | — | |
| CoTBackbone=Qwen2.5-7b-Instruct2025.10 | 11.1 | — | |
| CoTBackbone=Qwen2.5-3b-Instruct2025.10 | 2.1 | — | |
| CoopRAG (Gemma2-27B)Generator Backbone=GPT-4o-mini2025.12 | 0.717 | 0.79 | |
| CoopRAG (GPT-4o-mini)Generator Backbone=GPT-4o-mini2025.12 | 0.717 | 0.792 | |
| CoopRAG (Llama3.3-70B)Generator Backbone=GPT-4o-mini2025.12 | 0.712 | 0.788 | |
| CoopRAG (Gemma2-9B)Generator Backbone=GPT-4o-mini2025.12 | 0.7 | 0.781 | |
| HippoRAGGenerator Backbone=Llama3.3-70B2025.12 | 0.65 | 0.718 | |
| HippoRAG2Generator Backbone=Llama3.3-70B2025.12 | 0.65 | 0.71 | |
| HopRAGGenerator Backbone=GPT-4o-mini2025.12 | 0.611 | 0.683 | |
| HippoRAG2Generator Backbone=GPT-4o-mini2025.12 | 0.605 | 0.697 | |
| SiReRAGGenerator Backbone=GPT-4o-mini2025.12 | 0.596 | 0.679 | |
| HippoRAGGenerator Backbone=GPT-4o-mini2025.12 | 0.594 | 0.673 | |
| NV-Embed-v2 (7B)Generator Backbone=Llama3.3-70B2025.12 | 0.575 | 0.615 | |
| GritLM-7BGenerator Backbone=Llama3.3-70B2025.12 | 0.558 | 0.606 | |
| GTE-Qwen2-7B-InstructGenerator Backbone=Llama3.3-70B2025.12 | 0.551 | 0.6 | |
| GraphRAGGenerator Backbone=Llama3.3-70B2025.12 | 0.514 | 0.586 | |
| GTR (T5-base)Generator Backbone=Llama3.3-70B2025.12 | 0.492 | 0.528 | |
| BM25Generator Backbone=Llama3.3-70B2025.12 | 0.479 | 0.512 | |
| RAPTORGenerator Backbone=Llama3.3-70B2025.12 | 0.473 | 0.521 | |
| GraphRAGGenerator Backbone=GPT-4o-mini2025.12 | 0.457 | 0.61 | |
| RAPTORGenerator Backbone=GPT-4o-mini2025.12 | 0.397 | 0.484 | |
| ContrieverGenerator Backbone=Llama3.3-70B2025.12 | 0.381 | 0.419 | |
| LightRAGGenerator Backbone=Llama3.3-70B2025.12 | 0.094 | 0.116 | |
| LightRAGGenerator Backbone=GPT-4o-mini2025.12 | 0.025 | 0.121 | |
| IRCoT2026.05 | — | 60.5 | |
| Search-R12026.05 | — | 62.1 | |
| Static RAG2026.05 | — | 57.8 | |
| Step-Levelretrieval_strategy=step-level adaptive retrieval2026.05 | — | 64.7 |