List your company

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

73.8AccuracySKILLRL
12.02428.06244.160.138Aug 11, 2025Sep 10, 2025Oct 10, 2025Nov 10, 2025Dec 10, 2025Jan 9, 2026Feb 9, 2026

Evaluation results

MethodLinks
SKILLRLBackbone=Qwen2.5-7B-Instruct2026.0273.8—
WebFilterEnvironment=Web Search, Backbone=Qwen2.5-7B-Instruct2025.0873.174.3
DeepResearcherEnvironment=Web Search, Backbone=Qwen2.5-7B-Instruct2025.087172.8
R1-SearcherEnvironment=Web Search, Backbone=Qwen2.5-7B-Instruct2025.0864.865.6
Search-r1-baseEnvironment=Local RAG2025.0856.657.6
EvolveRBackbone=Qwen2.5-7B-Instruct2026.0254.4—
GPT-4oEnvironment=Direct Reasoning2025.0852.643.2
Search-o1Environment=Local RAG2025.0846.653.6
Search-r1-instructEnvironment=Local RAG2025.084547.2
StepSearchBackbone=Qwen2.5-7B-Instruct2026.0240—
Search-R1Backbone=Qwen2.5-7B-Instruct2026.0236.8—
Gemini-2.0-FlashEnvironment=Direct Reasoning2025.0836.528
Search-o1Backbone=Qwen2.5-7B-Instruct2026.0230.4—
ZeroSearchBackbone=Qwen2.5-7B-Instruct2026.0227.8—
Qwen3-32BEnvironment=Direct Reasoning2025.0824.718.4
CoTBackbone=Qwen2.5-7B-Instruct2026.0224—
R1-InstructBackbone=Qwen2.5-7B-Instruct2026.0219.2—
RAGBackbone=Qwen2.5-7B-Instruct2026.0216.8—
Qwen2.5Backbone=Qwen2.5-7B-Instruct2026.0214.4—