Question Answering on popQA Noisy
25.07F1 ScoreOpenDecoder
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| OpenDecoderBackbone=Qwen-2.5-3B-Instruct2026.01 | 25.07 | 22.02 | |
| RbFTBackbone=Qwen-2.5-3B-Instruct2026.01 | 23.71 | 20.2 | |
| Vanilla SFTBackbone=Qwen-2.5-3B-Instruct2026.01 | 21.06 | 18.16 | |
| AstuteRAGBackbone=Qwen-2.5-3B-Instruct2026.01 | 20.28 | 16.6 | |
| InstructRAGBackbone=Qwen-2.5-3B-Instruct2026.01 | 14.25 | 21.4 | |
| RobustRAGBackbone=Qwen-2.5-3B-Instruct2026.01 | 10.33 | 14.8 | |
| Vanilla RAGBackbone=Qwen-2.5-3B-Instruct2026.01 | 7.83 | 20.66 |