Question Answering on NQ, TriviaQA, PopQA, HotpotQA, 2wiki, Musique, Bamboogle, FictionalHot
37.7Exact Match (EM)ReSeek
Evaluation Results
| Method | Links | |
|---|---|---|
| ReSeekBackbone=Qwen2.5-7b-Instruct2025.10 | 37.7 | |
| ZeroSearchBackbone=Qwen2.5-7b-Instruct2025.10 | 34.6 | |
| Search-R1Backbone=Qwen2.5-7b-Instruct2025.10 | 34.2 | |
| ReSeekBackbone=Qwen2.5-3b-Instruct2025.10 | 31.2 | |
| Search-R1Backbone=Qwen2.5-3b-Instruct2025.10 | 28.8 | |
| ZeroSearchBackbone=Qwen2.5-3b-Instruct2025.10 | 28.1 | |
| RAGBackbone=Qwen2.5-7b-Instruct2025.10 | 26.7 | |
| R1Backbone=Qwen2.5-7b-Instruct2025.10 | 23.8 | |
| RAGBackbone=Qwen2.5-3b-Instruct2025.10 | 23.7 | |
| R1Backbone=Qwen2.5-3b-Instruct2025.10 | 19.6 | |
| Search-o1Backbone=Qwen2.5-3b-Instruct2025.10 | 18.7 | |
| Search-o1Backbone=Qwen2.5-7b-Instruct2025.10 | 18.3 | |
| Direct InferenceBackbone=Qwen2.5-7b-Instruct2025.10 | 15.8 | |
| Direct InferenceBackbone=Qwen2.5-3b-Instruct2025.10 | 11.8 | |
| CoTBackbone=Qwen2.5-7b-Instruct2025.10 | 9.3 | |
| CoTBackbone=Qwen2.5-3b-Instruct2025.10 | 1.3 |