Cross-lingual Question Answering on MKQA English
72.07fEMCroSearch-R1
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CroSearch-R1Backbone=Qwen2.5-7B-Instruct, Paradigm=Post-training2026.04 | 72.07 | 82.68 | |
| Search-R1Backbone=Qwen2.5-7B-Instruct, Paradigm=Post-training2026.04 | 69.43 | 79.82 | |
| CroSearch-R1Backbone=Qwen2.5-3B-Instruct, Paradigm=Post-training2026.04 | 67.68 | 83.22 | |
| Search-R1Backbone=Qwen2.5-3B-Instruct, Paradigm=Post-training2026.04 | 63.38 | 77.31 | |
| SFTBackbone=Qwen2.5-7B-Instruct, Paradigm=Post-training2026.04 | 59.91 | 66.86 | |
| SFTBackbone=Qwen2.5-3B-Instruct, Paradigm=Post-training2026.04 | 56.85 | 63.6 | |
| RAGBackbone=Qwen2.5-7B-Instruct, Paradigm=Training-free2026.04 | 55.05 | 64.12 | |
| RAGBackbone=Qwen2.5-3B-Instruct, Paradigm=Training-free2026.04 | 52.08 | 60.73 | |
| Direct InferenceBackbone=Qwen2.5-7B-Instruct, Paradigm=Training-free2026.04 | 49.96 | 60 | |
| Search-o1Backbone=Qwen2.5-7B-Instruct, Paradigm=Training-free2026.04 | 44.24 | 56.06 | |
| Search-o1Backbone=Qwen2.5-3B-Instruct, Paradigm=Training-free2026.04 | 41.66 | 53.23 | |
| Direct InferenceBackbone=Qwen2.5-3B-Instruct, Paradigm=Training-free2026.04 | 41.11 | 51.8 | |
| IRCoTBackbone=Qwen2.5-7B-Instruct, Paradigm=Training-free2026.04 | 40.88 | 53.67 | |
| IRCoTBackbone=Qwen2.5-3B-Instruct, Paradigm=Training-free2026.04 | 38.83 | 51.49 |