Open-domain Question Answering on PopQA (EM, RAG Improvement)
45.5Accuracy (EM)PyRAG
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| PyRAGBackbone=Qwen2.5-72B-Instruct2026.05 | 45.5 | 12.3 | |
| IRCoTBackbone=Qwen2.5-72B-Instruct2026.05 | 44.6 | — | |
| ITER-RETGENBackbone=Qwen2.5-72B-Instruct2026.05 | 43.4 | — | |
| Self-AskBackbone=Qwen2.5-72B-Instruct2026.05 | 41.4 | — | |
| PyRAGBackbone=Qwen2.5-7B-Instruct2026.05 | 33.5 | 6.8 | |
| Vanilla RAGBackbone=Qwen2.5-72B-Instruct2026.05 | 33.2 | — | |
| IRCoTBackbone=Qwen2.5-7B-Instruct2026.05 | 32.6 | — | |
| ITER-RETGENBackbone=Qwen2.5-7B-Instruct2026.05 | 31.4 | — | |
| Self-AskBackbone=Qwen2.5-7B-Instruct2026.05 | 29.4 | — | |
| Vanilla RAGBackbone=Qwen2.5-7B-Instruct2026.05 | 26.7 | — | |
| CoTBackbone=Qwen2.5-72B-Instruct2026.05 | 24.4 | — | |
| Direct InferenceBackbone=Qwen2.5-72B-Instruct2026.05 | 19.7 | — | |
| Direct InferenceBackbone=Qwen2.5-7B-Instruct2026.05 | 14 | — | |
| CoTBackbone=Qwen2.5-7B-Instruct2026.05 | 5.4 | — |