List your company

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

46.8AccuracyMR-Search
13.62422.23730.8539.463Mar 11, 2026

Evaluation results

MethodLinks
MR-SearchBackbone=Qwen2.5-7b2026.0346.8
Search-R1Backbone=Qwen2.5-7b2026.0343.9
StepResearchBackbone=Qwen2.5-7b2026.0343.9
MR-SearchBackbone=Qwen2.5-3b2026.0341.9
PPRMBackbone=Qwen2.5-7b2026.0338.6
ReSearchBackbone=Qwen2.5-7b2026.0337.8
StepResearchBackbone=Qwen2.5-3b2026.0337.3
PPRMBackbone=Qwen2.5-3b2026.0335.3
Search-R1Backbone=Qwen2.5-3b2026.0332.6
ReSearchBackbone=Qwen2.5-3b2026.0330.5
Search-o1Backbone=Qwen2.5-3b2026.0322.1
Search-o1Backbone=Qwen2.5-7b2026.0318.7
Direct InferenceBackbone=Qwen2.5-7b2026.0318.3
Direct InferenceBackbone=Qwen2.5-3b2026.0314.9