List your company

Single-Hop Question Answering on NQ (Natural Questions) 2018 Wikipedia dump (dev)

50.2AccuracyMR-Search
9.01619.70830.441.092Mar 11, 2026

Evaluation results

MethodLinks
MR-SearchBackbone=Qwen2.5-7b2026.0350.2
MR-SearchBackbone=Qwen2.5-3b2026.0347.7
StepResearchBackbone=Qwen2.5-7b2026.0347.3
Search-R1Backbone=Qwen2.5-3b2026.0346.2
Search-R1Backbone=Qwen2.5-7b2026.0345.9
PPRMBackbone=Qwen2.5-7b2026.0345.8
StepResearchBackbone=Qwen2.5-3b2026.0344.6
ReSearchBackbone=Qwen2.5-3b2026.0342.7
PPRMBackbone=Qwen2.5-3b2026.0342.3
ReSearchBackbone=Qwen2.5-7b2026.0336.6
Search-o1Backbone=Qwen2.5-3b2026.0323.8
Search-o1Backbone=Qwen2.5-7b2026.0315.1
Direct InferenceBackbone=Qwen2.5-7b2026.0313.4
Direct InferenceBackbone=Qwen2.5-3b2026.0310.6