List your company

Multi-Hop Question Answering on HotpotQA (in-domain) (Accuracy)

45.1AccuracyGEPO
15.04422.84730.6538.453Oct 30, 2025Nov 16, 2025Dec 3, 2025Dec 20, 2025Jan 6, 2026Jan 23, 2026Feb 9, 2026

Evaluation results

MethodLinks
GEPOType=RL Training, Model=Qwen2.5-7B-Instruct2025.1045.1
SKILLRLBackbone=Qwen2.5-7B-Instruct2026.0243.2
GiGPOType=RL Training, Model=Qwen2.5-7B-Instruct2025.1041.6
GEPOType=RL Training, Model=Qwen2.5-3B-Instruct2025.1040.6
StepSearchBackbone=Qwen2.5-7B-Instruct2026.0238.6
StepSearchType=RL Training, Model=Qwen2.5-7B-Instruct2025.1038.6
EvolveRBackbone=Qwen2.5-7B-Instruct2026.0238.2
Search-R1Backbone=Qwen2.5-7B-Instruct2026.0237
Search-R1Type=RL Training, Model=Qwen2.5-7B-Instruct2025.1037
GiGPOType=RL Training, Model=Qwen2.5-3B-Instruct2025.1036.9
ZeroSearchBackbone=Qwen2.5-7B-Instruct2026.0234.6
ZeroSearchType=RL Training, Model=Qwen2.5-7B-Instruct2025.1034.6
StepSearchType=RL Training, Model=Qwen2.5-3B-Instruct2025.1034.5
Search-R1Type=RL Training, Model=Qwen2.5-3B-Instruct2025.1032.4
ZeroSearchType=RL Training, Model=Qwen2.5-3B-Instruct2025.1027.4
RAGBackbone=Qwen2.5-7B-Instruct2026.0225.8
R1-InstructType=RL Training, Model=Qwen2.5-3B-Instruct2025.1023.7
R1-InstructBackbone=Qwen2.5-7B-Instruct2026.0220.8
R1-InstructType=RL Training, Model=Qwen2.5-7B-Instruct2025.1020.8
Search-o1Backbone=Qwen2.5-7B-Instruct2026.0217
Qwen2.5Backbone=Qwen2.5-7B-Instruct2026.0216.4
CoTBackbone=Qwen2.5-7B-Instruct2026.0216.2