Question Answering on PopQA (EM, F1, AVG.)
46.86EMCQC-RAG
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| CQC-RAGRAG Framework=Multi-path RAG2026.06 | 46.86 | 53.19 | 50.03 | |
| DMQR-RAGRAG Framework=Multi-path RAG2026.06 | 46.46 | 53.31 | 49.89 | |
| Llama-3.1-8B-InstructRAG Framework=Standard RAG2026.06 | 42.96 | 50.35 | 46.66 | |
| Self-CertaintyRAG Framework=Multi-path RAG2026.06 | 42 | 48 | 45 | |
| Qwen3-8BRAG Framework=Standard RAG2026.06 | 40.67 | 49.59 | 45.13 | |
| Speculative RAGRAG Framework=Multi-path RAG2026.06 | 37.88 | 42.29 | 40.09 | |
| SPARC-RAGRAG Framework=Multi-path RAG2026.06 | 30.38 | 40.59 | 35.49 | |
| Mistral-7B-InstructRAG Framework=Standard RAG2026.06 | 25.73 | 36.79 | 31.26 |