Single-hop QA on Natural Questions (NQ)
72EMCoopRAG (GPT-4o-mini)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CoopRAG (GPT-4o-mini)Generator Backbone=GPT-4o-mini2025.12 | 72 | 82.3 | |
| CoopRAG (Llama3.3-70B)Generator Backbone=GPT-4o-mini2025.12 | 70.9 | 80.3 | |
| CoopRAG (Gemma2-27B)Generator Backbone=GPT-4o-mini2025.12 | 67.3 | 75.5 | |
| CoopRAG (Gemma2-9B)Generator Backbone=GPT-4o-mini2025.12 | 63.8 | 72.7 | |
| SDGA-PhaseFramework=ZeroSearch, Model=Qwen2.5-7B, Search Depth (Smax)=52026.05 | 60.8 | — | |
| SDGA-AutoFramework=ZeroSearch, Model=Qwen2.5-7B, Search Depth (Smax)=52026.05 | 58.2 | — | |
| SDGA-PhaseFramework=ZeroSearch, Model=Qwen2.5-3B, Search Depth (Smax)=52026.05 | 55.1 | — | |
| SDGA-AutoFramework=ZeroSearch, Model=Qwen2.5-3B, Search Depth (Smax)=52026.05 | 53.4 | — | |
| SDGA-PhaseFramework=Search-R1, Model=Qwen2.5-7B, Search Depth (Smax)=52026.05 | 51.4 | — | |
| GRPO-FullFramework=ZeroSearch, Model=Qwen2.5-7B2026.05 | 50.4 | — | |
| GRPO-HalfFramework=ZeroSearch, Model=Qwen2.5-7B2026.05 | 49.8 | — | |
| HippoRAG2Generator Backbone=Llama3.3-70B2025.12 | 48.6 | 63.3 | |
| SDGA-AutoFramework=Search-R1, Model=Qwen2.5-7B, Search Depth (Smax)=52026.05 | 48.2 | — | |
| NV-Embed-v2 (7B)Generator Backbone=Llama3.3-70B2025.12 | 47.3 | 61.9 | |
| GritLM-7BGenerator Backbone=Llama3.3-70B2025.12 | 46.8 | 61.3 | |
| GTE-Qwen2-7B-InstructGenerator Backbone=Llama3.3-70B2025.12 | 46.6 | 62 | |
| GTR (T5-base)Generator Backbone=Llama3.3-70B2025.12 | 45.5 | 59.9 | |
| ContrieverGenerator Backbone=Llama3.3-70B2025.12 | 45 | 58.9 | |
| SDGA-PhaseFramework=Search-R1, Model=Qwen2.5-3B, Search Depth (Smax)=52026.05 | 44.8 | — | |
| GRPO-FullFramework=Search-R1, Model=Qwen2.5-7B2026.05 | 44.8 | — | |
| BM25Generator Backbone=Llama3.3-70B2025.12 | 44.7 | 59 | |
| GRPO-FullFramework=ZeroSearch, Model=Qwen2.5-3B2026.05 | 44.3 | — | |
| GRPO-HalfFramework=ZeroSearch, Model=Qwen2.5-3B2026.05 | 44.1 | — | |
| HippoRAG2Generator Backbone=GPT-4o-mini2025.12 | 43.4 | 60 | |
| SDGA-PhaseFramework=ZeroSearch, Model=LLaMA-3.2-3B-Base, Search Depth (Smax)=52026.05 | 43.2 | — | |
| SDGA-PhaseFramework=Search-R1, Model=LLaMA-3.2-3B-Base, Search Depth (Smax)=52026.05 | 43.2 | — | |
| HippoRAGGenerator Backbone=Llama3.3-70B2025.12 | 43 | 55.3 | |
| HopRAGGenerator Backbone=GPT-4o-mini2025.12 | 42.9 | 59.2 | |
| SDGA-AutoFramework=Search-R1, Model=LLaMA-3.2-3B-Base, Search Depth (Smax)=52026.05 | 42.9 | — | |
| SDGA-AutoFramework=Search-R1, Model=Qwen2.5-3B, Search Depth (Smax)=52026.05 | 42.9 | — | |
| GRPO-HalfFramework=Search-R1, Model=Qwen2.5-7B2026.05 | 42.6 | — | |
| SDGA-AutoFramework=ZeroSearch, Model=LLaMA-3.2-3B-Base, Search Depth (Smax)=52026.05 | 42.5 | — | |
| SiReRAGGenerator Backbone=GPT-4o-mini2025.12 | 42.4 | 58.7 | |
| GRPO-FullFramework=Search-R1, Model=Qwen2.5-3B2026.05 | 41.6 | — | |
| GraphRAGGenerator Backbone=GPT-4o-mini2025.12 | 38 | 55.5 | |
| GRPO-FullFramework=ZeroSearch, Model=LLaMA-3.2-3B-Base2026.05 | 37.9 | — | |
| RAPTORGenerator Backbone=GPT-4o-mini2025.12 | 37.8 | 54.5 | |
| HippoRAGGenerator Backbone=GPT-4o-mini2025.12 | 37.2 | 55.2 | |
| RAPTORGenerator Backbone=Llama3.3-70B2025.12 | 36.9 | 50.7 | |
| GRPO-HalfFramework=Search-R1, Model=Qwen2.5-3B2026.05 | 36.7 | — | |
| GRPO-FullFramework=Search-R1, Model=LLaMA-3.2-3B-Base2026.05 | 35.8 | — | |
| GRPO-HalfFramework=ZeroSearch, Model=LLaMA-3.2-3B-Base2026.05 | 34.6 | — | |
| GRPO-HalfFramework=Search-R1, Model=LLaMA-3.2-3B-Base2026.05 | 33.1 | — | |
| GraphRAGGenerator Backbone=Llama3.3-70B2025.12 | 30.8 | 46.9 | |
| SDGA-AntiFramework=Search-R1, Model=Qwen2.5-3B, Search Depth (Smax)=52026.05 | 22.9 | — | |
| SDGA-AntiFramework=ZeroSearch, Model=Qwen2.5-7B, Search Depth (Smax)=52026.05 | 22.6 | — | |
| SDGA-AntiFramework=Search-R1, Model=Qwen2.5-7B, Search Depth (Smax)=52026.05 | 20.8 | — | |
| SDGA-AntiFramework=ZeroSearch, Model=Qwen2.5-3B, Search Depth (Smax)=52026.05 | 20.2 | — | |
| SDGA-AntiFramework=ZeroSearch, Model=LLaMA-3.2-3B-Base, Search Depth (Smax)=52026.05 | 19.8 | — | |
| SDGA-AntiFramework=Search-R1, Model=LLaMA-3.2-3B-Base, Search Depth (Smax)=52026.05 | 18.7 | — | |
| LightRAGGenerator Backbone=Llama3.3-70B2025.12 | 8.6 | 16.6 | |
| LightRAGGenerator Backbone=GPT-4o-mini2025.12 | 2.8 | 15.4 |