Question Answering on 2Wiki Extreme
26.99F1 ScoreOpenDecoder
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| OpenDecoderBackbone=Qwen-2.5-3B-Instruct2026.01 | 26.99 | 24 | |
| RbFTBackbone=Qwen-2.5-3B-Instruct2026.01 | 24.28 | 22.6 | |
| Vanilla SFTBackbone=Qwen-2.5-3B-Instruct2026.01 | 21.76 | 19.93 | |
| AstuteRAGBackbone=Qwen-2.5-3B-Instruct2026.01 | 17.36 | 15.1 | |
| InstructRAGBackbone=Qwen-2.5-3B-Instruct2026.01 | 11.25 | 6.8 | |
| Vanilla RAGBackbone=Qwen-2.5-3B-Instruct2026.01 | 7.41 | 13.2 | |
| RobustRAGBackbone=Qwen-2.5-3B-Instruct2026.01 | 1.18 | 3.13 |