List your company

Multi-hop Question Answering on HotpotQA offline Wiki-18 (test val)

44.7EMESearch-instruct
25.9830.8435.740.56Oct 1, 2025

Evaluation results

MethodLinks
ESearch-instructModel Scale=3B2025.1044.758.7
ESearch-instructModel Scale=7B2025.1044.257.6
ESearch-baseModel Scale=7B2025.1043.456.4
Search-R1-baseModel Scale=7B2025.1043.254.7
ESearch-baseModel Scale=3B2025.1041.554.8
Search-R1-instructModel Scale=7B2025.1039.450.2
R-Search-instruct-GRPOModel Scale=7B2025.1039.150
ZeroSearch-instructModel Scale=7B2025.1038.849.7
StepSearch-instructModel Scale=7B2025.1038.650.2
SSRL-instructModel Scale=7B2025.103848.9
StepSearch-baseModel Scale=7B2025.103849.3
ZeroSearch-baseModel Scale=7B2025.1037.548.1
Research-instructModel Scale=7B2025.1036.247.1
StepSearch-instructModel Scale=3B2025.1034.545.2
R-Search-instruct-PPOModel Scale=7B2025.1033.843.9
R-Search-instruct-GRPOModel Scale=3B2025.1032.942.7
StepSearch-baseModel Scale=3B2025.1032.943.4
SSRL-instructModel Scale=3B2025.1031.440.8
Search-R1-instructModel Scale=3B2025.1030.440.1
Research-baseModel Scale=7B2025.1029.438.8
R-Search-instruct-PPOModel Scale=3B2025.1028.938.1
ZeroSearch-baseModel Scale=3B2025.1028.137.7
Search-R1-baseModel Scale=3B2025.1027.236.1
ZeroSearch-instructModel Scale=3B2025.1026.735.3