Question Answering on HotpotQA (EM, F1, AVG)
54.7Exact Match (EM)DMQR-RAG
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| DMQR-RAGRAG Framework=Multi-path RAG2026.06 | 54.7 | 68.56 | 61.63 | |
| CQC-RAGRAG Framework=Multi-path RAG2026.06 | 52.5 | 67.6 | 60.05 | |
| Self-CertaintyRAG Framework=Multi-path RAG2026.06 | 52.2 | 65 | 58.6 | |
| SPARC-RAGRAG Framework=Multi-path RAG2026.06 | 42.2 | 58.99 | 50.6 | |
| Speculative RAGRAG Framework=Multi-path RAG2026.06 | 33 | 42.05 | 37.53 | |
| Llama-3.1-8B-InstructRAG Framework=Standard RAG2026.06 | 16.5 | 23.19 | 19.85 | |
| Qwen3-8BRAG Framework=Standard RAG2026.06 | 12.6 | 25.63 | 19.12 | |
| Mistral-7B-InstructRAG Framework=Standard RAG2026.06 | 6.6 | 16.62 | 11.61 |