Question Answering on General-domain QA Benchmarks
36.982WQA ScoreReFilter
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| ReFilterBackbone=Qwen2.5-72B-Instruct2026.02 | 36.98 | 35.86 | 35.03 | 54.42 | 40.57 | |
| ReFilterBackbone=Qwen2.5-32B-Instruct2026.02 | 35.03 | 35.26 | 27.16 | 51.86 | 37.33 | |
| VanillaBackbone=Qwen2.5-72B-Instruct2026.02 | 34.03 | 35.54 | 27.14 | 53.05 | 37.44 | |
| S-RAGBackbone=Qwen2.5-32B-Instruct2026.02 | 33.87 | 32.98 | 27.46 | 51.2 | 36.38 | |
| VanillaBackbone=Qwen2.5-32B-Instruct2026.02 | 33.11 | 31.46 | 25.51 | 50.29 | 35.09 | |
| S-RAGBackbone=Qwen2.5-72B-Instruct2026.02 | 31.88 | 35.71 | 35.11 | 52.57 | 38.82 |