Question Answering on MedBrowseComp
23.2F1 ScoreHybridDeepSearcher
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| HybridDeepSearcherSearch Strategy=Hybrid Search (Ours), Backbone=Qwen2.5-7B-Instruct2025.08 | 23.2 | 32.7 | 3.3 | 0.25 | |
| HybridDeepSearcherSearch Strategy=Hybrid Search (Ours), Backbone=Qwen3-8B2025.08 | 19.8 | 30.4 | 3.4 | 0.26 | |
| RAG-R1Search Strategy=Parallel Search, Backbone=Qwen2.5-7B-Instruct2025.08 | 19.2 | 28.2 | 2.6 | 0.24 | |
| Search-R1Search Strategy=Sequential Search, Backbone=Qwen2.5-7B-Instruct2025.08 | 18.8 | 21.6 | 4 | 0.16 | |
| R1-SearcherSearch Strategy=Sequential Search, Backbone=Qwen2.5-7B-Instruct2025.08 | 15.8 | 24.4 | 3.1 | 0.2 | |
| DeepResearcherSearch Strategy=Parallel Search, Backbone=Qwen2.5-7B-Instruct2025.08 | 14.7 | 26.1 | 4.3 | 0.2 | |
| Search-o1Search Strategy=Sequential Search, Backbone=Qwen3-8B2025.08 | 12.9 | 21.6 | 4.7 | 0.16 | |
| Standard RAGSearch Strategy=Non-iterative, Backbone=Qwen3-8B2025.08 | 11.3 | 16.3 | — | — | |
| Naïve GenSearch Strategy=Non-iterative, Backbone=Qwen3-8B2025.08 | 8 | 11.9 | — | — |