Question Answering on PopQA (Frequent)
52.7Exact Match (EM)Contriever
Evaluation results
| Method | Links | |
|---|---|---|
| ContrieverGenerator=GPT-3.5, Retriever=Contriever2026.02 | 52.7 | |
| RPDRGenerator=GPT-3.5, Retriever=RPDR2026.02 | 52.1 | |
| Parametric knowledgeGenerator=GPT-3.5, Retriever=Parametric2026.02 | 50.3 | |
| BM25Generator=GPT-3.5, Retriever=BM252026.02 | 42.6 | |
| RPDRGenerator=LLaMA-3 8B, Retriever=RPDR2026.02 | 41.8 | |
| ContrieverGenerator=GPT-neox 20B, Retriever=Contriever2026.02 | 40.7 | |
| RPDRGenerator=GPT-j-6B, Retriever=RPDR2026.02 | 40 | |
| RPDRGenerator=GPT-neox 20B, Retriever=RPDR2026.02 | 40 | |
| ContrieverGenerator=GPT-j-6B, Retriever=Contriever2026.02 | 39.1 | |
| ContrieverGenerator=LLaMA-3 8B, Retriever=Contriever2026.02 | 39.1 | |
| BM25Generator=LLaMA-3 8B, Retriever=BM252026.02 | 30.1 | |
| Parametric knowledgeGenerator=LLaMA-3 8B, Retriever=Parametric2026.02 | 29.9 | |
| BM25Generator=GPT-neox 20B, Retriever=BM252026.02 | 29.2 | |
| Parametric knowledgeGenerator=GPT-neox 20B, Retriever=Parametric2026.02 | 28.9 | |
| BM25Generator=GPT-j-6B, Retriever=BM252026.02 | 24.5 | |
| Parametric knowledgeGenerator=GPT-j-6B, Retriever=Parametric2026.02 | 18.2 |