Linguistic Competence Classification on BLiMP
83AccuracyFine-Tuned
Evaluation results
| Method | Links | |
|---|---|---|
| Fine-TunedBackbone=RoBERTa-large2025.12 | 83 | |
| Task MatrixBackbone=RoBERTa-large, best layer=(4,5,6)2025.12 | 75.3 | |
| Linear ProbeBackbone=RoBERTa-large2025.12 | 66.3 |