Question Answering on HotpotQA Extreme
23.59F1 ScoreOpenDecoder
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| OpenDecoderBackbone=Qwen-2.5-3B-Instruct2026.01 | 23.59 | 17.2 | |
| RbFTBackbone=Qwen-2.5-3B-Instruct2026.01 | 22.11 | 15.5 | |
| Vanilla SFTBackbone=Qwen-2.5-3B-Instruct2026.01 | 18.26 | 15.07 | |
| AstuteRAGBackbone=Qwen-2.5-3B-Instruct2026.01 | 14.38 | 10.6 | |
| InstructRAGBackbone=Qwen-2.5-3B-Instruct2026.01 | 9.14 | 5.8 | |
| Vanilla RAGBackbone=Qwen-2.5-3B-Instruct2026.01 | 4.2 | 9.67 | |
| RobustRAGBackbone=Qwen-2.5-3B-Instruct2026.01 | 1.6 | 4.67 |