Question Answering on TrivialQA Noisy
55.09F1 ScoreOpenDecoder
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| OpenDecoderBackbone=Qwen-2.5-3B-Instruct2026.01 | 55.09 | 53.33 | |
| RbFTBackbone=Qwen-2.5-3B-Instruct2026.01 | 52.62 | 51.7 | |
| Vanilla SFTBackbone=Qwen-2.5-3B-Instruct2026.01 | 48.54 | 48.07 | |
| AstuteRAGBackbone=Qwen-2.5-3B-Instruct2026.01 | 46.81 | 48.7 | |
| InstructRAGBackbone=Qwen-2.5-3B-Instruct2026.01 | 44.13 | 48.2 | |
| RobustRAGBackbone=Qwen-2.5-3B-Instruct2026.01 | 42.36 | 44.53 | |
| Vanilla RAGBackbone=Qwen-2.5-3B-Instruct2026.01 | 26.82 | 49.93 |