Question Answering on HotpotQA (EM, ACC)
35.6EMDRAGIN
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DRAGINBackbone=Gemma2-9B2026.04 | 35.6 | 37.5 | |
| FLAREBackbone=Gemma2-9B2026.04 | 29.5 | 33.8 | |
| Adaptive-RAGBackbone=Gemma2-9B2026.04 | 29.1 | 32.3 | |
| Single-step ApproachBackbone=Gemma2-9B2026.04 | 27.8 | 42.6 | |
| No RetrievalBackbone=Gemma2-9B2026.04 | 24.6 | 37.1 | |
| Skill-RAGBackbone=Gemma2-9B2026.04 | 24.2 | 46.1 | |
| Probing-RAGBackbone=Gemma2-9B2026.04 | 22.8 | 44.7 |