Question Answering on NQ Normal
40.17F1 ScoreRbFT
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| RbFTBackbone=Qwen-2.5-3B-Instruct2026.01 | 40.17 | 36.6 | |
| OpenDecoderBackbone=Qwen-2.5-3B-Instruct2026.01 | 39.26 | 35.9 | |
| AstuteRAGBackbone=Qwen-2.5-3B-Instruct2026.01 | 37.84 | 34.1 | |
| Vanilla SFTBackbone=Qwen-2.5-3B-Instruct2026.01 | 33.63 | 32.63 | |
| InstructRAGBackbone=Qwen-2.5-3B-Instruct2026.01 | 30.39 | 31 | |
| RobustRAGBackbone=Qwen-2.5-3B-Instruct2026.01 | 26.58 | 30.8 | |
| Vanilla RAGBackbone=Qwen-2.5-3B-Instruct2026.01 | 25.46 | 34.12 | |
| No RAGBackbone=Qwen-2.5-3B-Instruct2026.01 | 12.11 | — |