Question Answering on PopQA Infrequent
42.9Exact Match AccuracyRPDR
Evaluation Results
| Method | Links | |
|---|---|---|
| RPDRGenerator=GPT-3.5, Retriever=RPDR2026.02 | 42.9 | |
| ContrieverGenerator=GPT-3.5, Retriever=Contriever2026.02 | 41.5 | |
| Parametric knowledgeGenerator=GPT-3.5, Retriever=Parametric2026.02 | 36.4 | |
| RPDRGenerator=GPT-neox 20B, Retriever=RPDR2026.02 | 35.5 | |
| ContrieverGenerator=LLaMA-3 8B, Retriever=Contriever2026.02 | 34.9 | |
| RPDRGenerator=LLaMA-3 8B, Retriever=RPDR2026.02 | 34.7 | |
| ContrieverGenerator=GPT-neox 20B, Retriever=Contriever2026.02 | 33.2 | |
| BM25Generator=GPT-3.5, Retriever=BM252026.02 | 31.8 | |
| RPDRGenerator=GPT-j-6B, Retriever=RPDR2026.02 | 29.7 | |
| BM25Generator=LLaMA-3 8B, Retriever=BM252026.02 | 29.2 | |
| BM25Generator=GPT-neox 20B, Retriever=BM252026.02 | 27.4 | |
| ContrieverGenerator=GPT-j-6B, Retriever=Contriever2026.02 | 24.7 | |
| Parametric knowledgeGenerator=LLaMA-3 8B, Retriever=Parametric2026.02 | 21.6 | |
| Parametric knowledgeGenerator=GPT-neox 20B, Retriever=Parametric2026.02 | 21.2 | |
| BM25Generator=GPT-j-6B, Retriever=BM252026.02 | 19.4 | |
| Parametric knowledgeGenerator=GPT-j-6B, Retriever=Parametric2026.02 | 14.4 |