List your company

Single-Hop Question Answering on PopQA out-of-domain

52.4AccuracyGEPO
-0.84812.97626.840.624Oct 30, 2025Dec 5, 2025Jan 10, 2026Feb 15, 2026Mar 23, 2026Apr 28, 2026Jun 4, 2026

Evaluation results

MethodLinks
GEPOType=RL Training, Model=Qwen2.5-7B-Instruct2025.1052.4
ZeroSearchBackbone=Qwen2.5-7B-Instruct2026.0251.5
ZeroSearchType=RL Training, Model=Qwen2.5-7B-Instruct2025.1051.5
GEPOType=RL Training, Model=Qwen2.5-3B-Instruct2025.1047.2
GiGPOType=RL Training, Model=Qwen2.5-7B-Instruct2025.1046.1
SKILLRLBackbone=Qwen2.5-7B-Instruct2026.0245.9
ZeroSearchType=RL Training, Model=Qwen2.5-3B-Instruct2025.1044.8
ZeroSearch2026.0644.8
EvolveRBackbone=Qwen2.5-7B-Instruct2026.0244.6
ECPO2026.0644.1
GiGPOType=RL Training, Model=Qwen2.5-3B-Instruct2025.1042.4
GiGPO2026.0641.5
Search-R1Backbone=Qwen2.5-7B-Instruct2026.0239.7
Search-R1Type=RL Training, Model=Qwen2.5-7B-Instruct2025.1039.7
Search-R1Type=RL Training, Model=Qwen2.5-3B-Instruct2025.1037.8
Search-R12026.0637.8
R1-InstructType=RL Training, Model=Qwen2.5-3B-Instruct2025.1019.9
R1-Instruct2026.0619.9
RAGBackbone=Qwen2.5-7B-Instruct2026.0217.8
R1-InstructBackbone=Qwen2.5-7B-Instruct2026.0217.1
R1-InstructType=RL Training, Model=Qwen2.5-7B-Instruct2025.1017.1
Search-o1Backbone=Qwen2.5-7B-Instruct2026.0211.4
CoTBackbone=Qwen2.5-7B-Instruct2026.023.8
Qwen2.5Backbone=Qwen2.5-7B-Instruct2026.021.2