List your company

Multi-Hop QA on Bamboogle (Exact Match (EM))

57.8Exact Match (EM)SDGA-Phase
-2.31213.29428.944.506Oct 1, 2025Nov 7, 2025Dec 14, 2025Jan 20, 2026Feb 26, 2026Apr 4, 2026May 12, 2026

Evaluation results

MethodLinks
SDGA-PhaseFramework=ZeroSearch, Model=Qwen2.5-7B, Search Depth (Smax)=52026.0557.8
SDGA-PhaseFramework=ZeroSearch, Model=Qwen2.5-3B, Search Depth (Smax)=52026.0551.4
SDGA-AutoFramework=ZeroSearch, Model=Qwen2.5-7B, Search Depth (Smax)=52026.0547.4
GRPO-FullFramework=ZeroSearch, Model=Qwen2.5-7B2026.0546.8
GRPO-FullFramework=ZeroSearch, Model=Qwen2.5-3B2026.0539.8
ReSeekBackbone=Qwen2.5-7b-Instruct2025.1039.2
SDGA-AutoFramework=ZeroSearch, Model=Qwen2.5-3B, Search Depth (Smax)=52026.0539.1
SDGA-PhaseFramework=Search-R1, Model=Qwen2.5-7B, Search Depth (Smax)=52026.0538.4
Search-R1Backbone=Qwen2.5-7b-Instruct2025.1036.8
SDGA-PhaseFramework=Search-R1, Model=Qwen2.5-3B, Search Depth (Smax)=52026.0532.3
ReSeekBackbone=Qwen2.5-3b-Instruct2025.1030.4
Search-o1Backbone=Qwen2.5-7b-Instruct2025.1029.6
GRPO-HalfFramework=ZeroSearch, Model=Qwen2.5-7B2026.0529.6
R1Backbone=Qwen2.5-7b-Instruct2025.1029.3
ZeroSearchBackbone=Qwen2.5-7b-Instruct2025.1027.8
Search-R1Backbone=Qwen2.5-3b-Instruct2025.1026.4
SDGA-AutoFramework=Search-R1, Model=Qwen2.5-7B, Search Depth (Smax)=52026.0524.6
GRPO-HalfFramework=ZeroSearch, Model=Qwen2.5-3B2026.0524.1
GRPO-FullFramework=Search-R1, Model=Qwen2.5-7B2026.0523.6
SDGA-AutoFramework=Search-R1, Model=LLaMA-3.2-3B-Base, Search Depth (Smax)=52026.0523.4
CoTBackbone=Qwen2.5-7b-Instruct2025.1023.2
SDGA-PhaseFramework=Search-R1, Model=LLaMA-3.2-3B-Base, Search Depth (Smax)=52026.0522.9
SDGA-AutoFramework=ZeroSearch, Model=LLaMA-3.2-3B-Base, Search Depth (Smax)=52026.0522.1
SDGA-PhaseFramework=ZeroSearch, Model=LLaMA-3.2-3B-Base, Search Depth (Smax)=52026.0521.8
SDGA-AntiFramework=ZeroSearch, Model=Qwen2.5-7B, Search Depth (Smax)=52026.0521.4
SDGA-AutoFramework=Search-R1, Model=Qwen2.5-3B, Search Depth (Smax)=52026.0521.4
SDGA-AntiFramework=ZeroSearch, Model=Qwen2.5-3B, Search Depth (Smax)=52026.0520.9
GRPO-HalfFramework=Search-R1, Model=Qwen2.5-3B2026.0520.9
RAGBackbone=Qwen2.5-7b-Instruct2025.1020.8
GRPO-FullFramework=Search-R1, Model=Qwen2.5-3B2026.0520.1
R1Backbone=Qwen2.5-3b-Instruct2025.1019.2
SDGA-AntiFramework=Search-R1, Model=Qwen2.5-7B, Search Depth (Smax)=52026.0516.8
GRPO-FullFramework=ZeroSearch, Model=LLaMA-3.2-3B-Base2026.0516.4
GRPO-FullFramework=Search-R1, Model=LLaMA-3.2-3B-Base2026.0515.2
GRPO-HalfFramework=ZeroSearch, Model=LLaMA-3.2-3B-Base2026.0513.6
GRPO-HalfFramework=Search-R1, Model=LLaMA-3.2-3B-Base2026.0512.7
GRPO-HalfFramework=Search-R1, Model=Qwen2.5-7B2026.0512.4
Direct InferenceBackbone=Qwen2.5-7b-Instruct2025.1012
ZeroSearchBackbone=Qwen2.5-3b-Instruct2025.1011.1
SDGA-AntiFramework=Search-R1, Model=Qwen2.5-3B, Search Depth (Smax)=52026.0510.1
SDGA-AntiFramework=ZeroSearch, Model=LLaMA-3.2-3B-Base, Search Depth (Smax)=52026.058.5
SDGA-AntiFramework=Search-R1, Model=LLaMA-3.2-3B-Base, Search Depth (Smax)=52026.058.3
RAGBackbone=Qwen2.5-3b-Instruct2025.108
Direct InferenceBackbone=Qwen2.5-3b-Instruct2025.102.4
Search-o1Backbone=Qwen2.5-3b-Instruct2025.101.9
CoTBackbone=Qwen2.5-3b-Instruct2025.100