Single-hop Question Answering on NQ 2019 (test)
61.4F1 ScoreCIRAG
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CIRAGBackbone=Qwen2.5-7B-Instruct2026.01 | 61.4 | 46 | |
| MetaRAGBackbone=Qwen2.5-7B-Instruct2026.01 | 61.1 | 47.8 | |
| Dualrag-FTBackbone=Qwen2.5-7B-Instruct, Fine-tuned=true2026.01 | 60.2 | 45.7 | |
| NativeBackbone=Qwen2.5-7B-Instruct2026.01 | 59.6 | 43.7 | |
| DualragBackbone=Qwen2.5-7B-Instruct2026.01 | 58.2 | 44.2 | |
| KiragBackbone=Qwen2.5-7B-Instruct2026.01 | 57.1 | 41.7 | |
| IRCOTBackbone=Qwen2.5-7B-Instruct2026.01 | 56.7 | 41 | |
| FLAREBackbone=Qwen2.5-7B-Instruct2026.01 | 56.3 | 43 |