Question Answering on HotpotQA Normal
29.71F1 ScoreRbFT
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| RbFTBackbone=Qwen-2.5-3B-Instruct2026.01 | 29.71 | 24.5 | |
| AstuteRAGBackbone=Qwen-2.5-3B-Instruct2026.01 | 29.44 | 23.3 | |
| OpenDecoderBackbone=Qwen-2.5-3B-Instruct2026.01 | 29.44 | 24 | |
| Vanilla SFTBackbone=Qwen-2.5-3B-Instruct2026.01 | 24.02 | 20.17 | |
| InstructRAGBackbone=Qwen-2.5-3B-Instruct2026.01 | 20.26 | 22.93 | |
| No RAGBackbone=Qwen-2.5-3B-Instruct2026.01 | 16.86 | — | |
| RobustRAGBackbone=Qwen-2.5-3B-Instruct2026.01 | 13.58 | 15.73 | |
| Vanilla RAGBackbone=Qwen-2.5-3B-Instruct2026.01 | 12.06 | 20.73 |