Question Answering on TrivialQA Normal
56.08F1OpenDecoder
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| OpenDecoderBackbone=Qwen-2.5-3B-Instruct2026.01 | 56.08 | 54.87 | |
| RbFTBackbone=Qwen-2.5-3B-Instruct2026.01 | 53.49 | 52.3 | |
| AstuteRAGBackbone=Qwen-2.5-3B-Instruct2026.01 | 52.28 | 51.8 | |
| Vanilla SFTBackbone=Qwen-2.5-3B-Instruct2026.01 | 50.31 | 50.53 | |
| InstructRAGBackbone=Qwen-2.5-3B-Instruct2026.01 | 45.81 | 50.73 | |
| RobustRAGBackbone=Qwen-2.5-3B-Instruct2026.01 | 45.25 | 47.4 | |
| Vanilla RAGBackbone=Qwen-2.5-3B-Instruct2026.01 | 31.09 | 48.4 | |
| No RAGBackbone=Qwen-2.5-3B-Instruct2026.01 | 30.04 | — |