Question Answering on HotpotQA Noisy
28.76F1 ScoreOpenDecoder
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| OpenDecoderBackbone=Qwen-2.5-3B-Instruct2026.01 | 28.76 | 22.77 | |
| RbFTBackbone=Qwen-2.5-3B-Instruct2026.01 | 25.28 | 19 | |
| AstuteRAGBackbone=Qwen-2.5-3B-Instruct2026.01 | 23.63 | 17 | |
| Vanilla SFTBackbone=Qwen-2.5-3B-Instruct2026.01 | 23.55 | 20.8 | |
| InstructRAGBackbone=Qwen-2.5-3B-Instruct2026.01 | 18.16 | 11.6 | |
| RobustRAGBackbone=Qwen-2.5-3B-Instruct2026.01 | 12.04 | 14.13 | |
| Vanilla RAGBackbone=Qwen-2.5-3B-Instruct2026.01 | 11.05 | 19 |