Case holding classification on CaseHOLD (test)
78.5Mean macro F1LexLM-Large
Evaluation results
| Method | Links | ||
|---|---|---|---|
| LexLM-LargeFine-tuning epochs=1, Classifier head=Standard N-way2023.05 | 78.5 | 78.5 | |
| CL-BERTFine-tuning epochs=1, Classifier head=Standard N-way2023.05 | 77.6 | 77.6 | |
| LegalBERTFine-tuning epochs=1, Classifier head=Standard N-way2023.05 | 76.1 | 76.1 | |
| ROBERTa-LargeFine-tuning epochs=1, Classifier head=Standard N-way2023.05 | 75.6 | 75.6 | |
| LexLM-BaseFine-tuning epochs=1, Classifier head=Standard N-way2023.05 | 74.8 | 74.8 | |
| POL-BERTFine-tuning epochs=1, Classifier head=Standard N-way2023.05 | 73.7 | 73.7 | |
| ROBERTa-BaseFine-tuning epochs=1, Classifier head=Standard N-way2023.05 | 72.7 | 72.7 | |
| Custom Legal-BERTNumber of Pretraining Steps=2M, Vocabulary Size (domain)=32,000 (legal)2021.04 | 69.5 | — | |
| Legal-BERTNumber of Pretraining Steps=2M, Vocabulary Size (domain)=30,522 (general)2021.04 | 68 | — | |
| BERT (double)Number of Pretraining Steps=2M, Vocabulary Size (domain)=30,522 (general)2021.04 | 62.3 | — | |
| BERTNumber of Pretraining Steps=1M, Vocabulary Size (domain)=30,522 (general)2021.04 | 61.3 | — | |
| Baseline2021.04 | 39.9 | — |