List your company

Multi-Hop Question Answering on Bamboogle out-of-domain (val test)

56.4Exact Match (EM)Search-R2
9.39221.59633.846.004Feb 3, 2026

Evaluation results

MethodLinks
Search-R2Backbone=Qwen2.5-32B2026.0256.4
Search-R2Backbone=Qwen3-8B2026.0251.2
Search-R2Backbone=Qwen2.5-7B2026.0246.2
Search-R1Backbone=Qwen2.5-32B2026.0245
Search-R1Backbone=Qwen3-8B2026.0243
Search-R1Backbone=Qwen2.5-7B2026.0236
Rejection SamplingBackbone=Qwen2.5-7B2026.0235.5
Search-o1Backbone=Qwen2.5-7B2026.0229.6
R1-baseBackbone=Qwen2.5-7B2026.0229.6
R1-instructBackbone=Qwen2.5-7B2026.0229.3
CoTBackbone=Qwen2.5-7B2026.0223.2
IRCoTBackbone=Qwen2.5-7B2026.0222.4
RAGBackbone=Qwen2.5-7B2026.0220.8
Direct InferenceBackbone=Qwen2.5-7B2026.0212
SFTBackbone=Qwen2.5-7B2026.0211.2