Information Retrieval on SCIDocs (nDCG@10)
23.11nDCG@10text-embedding-ada-002
Evaluation Results
| Method | Links | |
|---|---|---|
| text-embedding-ada-002Dim.=1,536, Type=Dense2025.05 | 23.11 | |
| AngIEDim.=1,024, Type=Dense2025.05 | 22.98 | |
| SBERTnewDim.=768, Type=Dense2025.05 | 21.82 | |
| ModernBERTembDim.=1,024, Type=Dense2025.05 | 21.4 | |
| LDIREncoder=SBERTnew, Dim.=500, Type=Dense2025.05 | 19.74 | |
| LDIREncoder=SBERTnew, Dim.=200, Type=Dense2025.05 | 18.34 | |
| LDIREncoder=AngIE, Dim.=500, Type=Dense2025.05 | 17.72 | |
| LSP/0Corpus Size=25K2026.02 | 16.2 | |
| SafeCorpus Size=25K2026.02 | 15.9 | |
| SPCorpus Size=25K2026.02 | 15.9 | |
| BMPCorpus Size=25K2026.02 | 15.8 | |
| SeismicWaveCorpus Size=25K2026.02 | 15.8 | |
| BM25Dim.=N/A, Type=Sparse2025.05 | 15.78 | |
| LDIREncoder=AngIE, Dim.=200, Type=Dense2025.05 | 15.44 | |
| LDIREncoder=ModernBERTemb, Dim.=500, Type=Dense2025.05 | 13.87 | |
| LDIREncoder=ModernBERTemb, Dim.=200, Type=Dense2025.05 | 11.27 | |
| LLM2VecDim.=4,096, Type=Dense2025.05 | 10.39 | |
| CQG-MBQADim.=9,614, Type=0/1 Emb.2025.05 | 8.67 | |
| GloVeDim.=300, Type=Dense2025.05 | 8.04 | |
| SimCSEsupDim.=768, Type=Dense2025.05 | 7.53 | |
| Bag-of-WordsVocabulary=BERT, Dim.=28,996, Type=Sparse2025.05 | 6.79 | |
| SimCSEunsupDim.=768, Type=Dense2025.05 | 5.5 | |
| QAEmb-MBQADim.=10,654, Type=0/1 Emb.2025.05 | 3.74 | |
| BERTbaseDim.=768, Type=Dense2025.05 | 2.82 |