Model: UnixCoder-MIL
Macro F1: 0.8782

              precision    recall  f1-score   support

       Human     0.9837    0.9704    0.9770    107885
      AI-Gen     0.8746    0.8434    0.8587     46770
      Hybrid     0.7943    0.8706    0.8307     19006
 Adversarial     0.8264    0.8673    0.8464     26339

    accuracy                         0.9176    200000
   macro avg     0.8698    0.8879    0.8782    200000
weighted avg     0.9195    0.9176    0.9182    200000