This model is based on svenbl80/deberta-v3-Base-finetuned-chatdoc-V5's model but further finetuned a synthetic dataset.



It performs poorly on a different benchmark from the same document:
                precision    recall  f1-score   support
  
             0       0.19      0.22      0.20        23
             1       0.62      0.44      0.52        75
             2       0.00      0.00      0.00        19
  
      accuracy                           0.32       117
     macro avg       0.27      0.22      0.24       117
  weighted avg       0.44      0.32      0.37       117