Multi-Hop QA on FictionalHot
6.1Exact Match (EM)ReSeek
Evaluation Results
| Method | Links | |
|---|---|---|
| ReSeekBackbone=Qwen2.5-7b-Instruct2025.10 | 6.1 | |
| ReSeekBackbone=Qwen2.5-3b-Instruct2025.10 | 5.9 | |
| Search-R1Backbone=Qwen2.5-3b-Instruct2025.10 | 3.7 | |
| Search-R1Backbone=Qwen2.5-7b-Instruct2025.10 | 3.4 | |
| ZeroSearchBackbone=Qwen2.5-7b-Instruct2025.10 | 3.1 | |
| ZeroSearchBackbone=Qwen2.5-3b-Instruct2025.10 | 3 | |
| Search-o1Backbone=Qwen2.5-7b-Instruct2025.10 | 2 | |
| RAGBackbone=Qwen2.5-7b-Instruct2025.10 | 1.2 | |
| Search-o1Backbone=Qwen2.5-3b-Instruct2025.10 | 1 | |
| RAGBackbone=Qwen2.5-3b-Instruct2025.10 | 0.8 | |
| R1Backbone=Qwen2.5-7b-Instruct2025.10 | 0.3 | |
| R1Backbone=Qwen2.5-3b-Instruct2025.10 | 0.3 | |
| Direct InferenceBackbone=Qwen2.5-7b-Instruct2025.10 | 0.1 | |
| CoTBackbone=Qwen2.5-7b-Instruct2025.10 | 0.1 | |
| Direct InferenceBackbone=Qwen2.5-3b-Instruct2025.10 | 0.1 | |
| CoTBackbone=Qwen2.5-3b-Instruct2025.10 | 0.1 |