Question Answering on PopQA (Pattern-based, Model Averages)
71.26Pattern-based ScoreCoRAG
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| CoRAGbackbone=Llama (fine-tuned)2026.02 | 71.26 | 89.21 | 65.9 | 64.26 | 66.12 | 71.35 | |
| InstructRAGtraining_dataset=Pop2026.02 | 66.19 | 75.63 | 60.83 | 59.69 | 60.69 | 64.6 | |
| RetRobust2026.02 | 56.68 | 25.52 | 54.9 | 34.1 | 57.26 | 45.69 |