Cross-Corpus Ranking on Cross-Corpus Dataset
3.7Avg. RFRSciBERT
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| SciBERTEmbedding pooling strategy=CLS2025.12 | 3.7 | 40.69 | 60.78 | |
| BERTEmbedding pooling strategy=CLS2025.12 | 2.51 | 47.13 | 68.96 | |
| GTE-largeEmbedding pooling strategy=CLS2025.12 | 2.28 | 52.79 | 75.75 | |
| SciBERTEmbedding pooling strategy=Mean2025.12 | 1.73 | 67.88 | 86.32 | |
| Pub-PaECTEREmbedding pooling strategy=CLS, Fine-tuned=true2025.12 | 1.44 | 73.02 | 89.73 | |
| Pub-PaECTEREmbedding pooling strategy=Mean, Fine-tuned=true2025.12 | 1.37 | 76.56 | 91.65 | |
| BERTEmbedding pooling strategy=Mean2025.12 | 1.28 | 80.83 | 91.8 | |
| SPECTER2Embedding pooling strategy=Mean2025.12 | 1.26 | 80.17 | 93.31 | |
| SPECTEREmbedding pooling strategy=Mean2025.12 | 1.23 | 84.5 | 94.63 | |
| BERT for PatentsEmbedding pooling strategy=CLS2025.12 | 1.2 | 78.9 | 93.48 | |
| BERT for PatentsEmbedding pooling strategy=Mean2025.12 | 1.16 | 86.97 | 95.98 | |
| Pat-SPECTEREmbedding pooling strategy=Mean, Fine-tuned=true2025.12 | 1.12 | 88.85 | 96.51 | |
| SPECTER2Embedding pooling strategy=CLS2025.12 | 1.11 | 89.89 | 96.21 | |
| PatentSBERTaEmbedding pooling strategy=CLS2025.12 | 1.1 | 88.67 | 97.08 | |
| PatentSBERTaEmbedding pooling strategy=Mean2025.12 | 1.09 | 88.43 | 96.99 | |
| SPECTEREmbedding pooling strategy=CLS2025.12 | 1.09 | 91.64 | 97.37 | |
| PaECTEREmbedding pooling strategy=CLS2025.12 | 1.09 | 89.42 | 97.43 | |
| GTE-largeEmbedding pooling strategy=Mean2025.12 | 1.07 | 93.23 | 98.39 | |
| PaECTEREmbedding pooling strategy=Mean2025.12 | 1.07 | 91.89 | 98.03 | |
| Pat-SPECTEREmbedding pooling strategy=CLS, Fine-tuned=true2025.12 | 1.06 | 93.11 | 98.06 |