Question Answering on TriviaQA (EM, ACC)
59.3EMProbing-RAG
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Probing-RAGBackbone=Gemma2-9B2026.04 | 59.3 | 65.9 | |
| Skill-RAGBackbone=Gemma2-9B2026.04 | 59.3 | 65.9 | |
| DRAGINBackbone=Gemma2-9B2026.04 | 55.3 | 63.4 | |
| No RetrievalBackbone=Gemma2-9B2026.04 | 52.4 | 65.7 | |
| Single-step ApproachBackbone=Gemma2-9B2026.04 | 46.5 | 59.3 | |
| Adaptive-RAGBackbone=Gemma2-9B2026.04 | 46.2 | 58.6 | |
| FLAREBackbone=Gemma2-9B2026.04 | 44.8 | 57.2 |