List your company

Multi-Hop Question Answering on MuSiQue (Accuracy)

22.6AccuracyStepSearch
5.3369.81814.318.782Mar 7, 2026

Evaluation results

MethodLinks
StepSearchType=RL Training, Base Model=Qwen2.5-7B-Instruct2026.0322.6
GiGPOType=RL Training, Base Model=Qwen2.5-7B-Instruct2026.0318.9
ZeroSearchType=RL Training, Base Model=Qwen2.5-7B-Instruct2026.0318.4
HCAPO (Ours)Type=RL Training, Base Model=Qwen2.5-7B-Instruct2026.0317.7
StepSearchType=RL Training, Base Model=Qwen2.5-3B-Instruct2026.0317.4
HCAPO (Ours)Type=RL Training, Base Model=Qwen2.5-3B-Instruct2026.0314.8
Search-R1Type=RL Training, Base Model=Qwen2.5-7B-Instruct2026.0314.6
GiGPOType=RL Training, Base Model=Qwen2.5-3B-Instruct2026.0312.6
Search-R1Type=RL Training, Base Model=Qwen2.5-3B-Instruct2026.0310.3
ZeroSearchType=RL Training, Base Model=Qwen2.5-3B-Instruct2026.039.8
R1-InstructType=RL Training, Base Model=Qwen2.5-3B-Instruct2026.037.2
R1-InstructType=RL Training, Base Model=Qwen2.5-7B-Instruct2026.036