List your company

Multi-Hop Question Answering on Bamboogle 2018 Wikipedia dump (dev)

45.2AccuracyMR-Search
0.68812.24423.835.356Mar 11, 2026

Evaluation results

MethodLinks
MR-SearchBackbone=Qwen2.5-7b2026.0345.2
StepResearchBackbone=Qwen2.5-7b2026.0343.5
Search-R1Backbone=Qwen2.5-7b2026.0340
ReSearchBackbone=Qwen2.5-7b2026.0337.6
PPRMBackbone=Qwen2.5-7b2026.0335.5
MR-SearchBackbone=Qwen2.5-3b2026.0334.4
StepResearchBackbone=Qwen2.5-3b2026.0332.5
Search-o1Backbone=Qwen2.5-3b2026.0332
Search-o1Backbone=Qwen2.5-7b2026.0329.6
PPRMBackbone=Qwen2.5-3b2026.0328
Search-R1Backbone=Qwen2.5-3b2026.0317.6
Direct InferenceBackbone=Qwen2.5-7b2026.0312
ReSearchBackbone=Qwen2.5-3b2026.0311.5
Direct InferenceBackbone=Qwen2.5-3b2026.032.4