aggregate_metrics.json

{
  "test_macro_f1_mean": 0.8794988613903062,
  "test_macro_f1_std": 0.013892906720731662,
  "test_macro_f1_text": "0.8795 ± 0.0139",
  "test_accuracy_mean": 0.8798342541436464,
  "test_accuracy_std": 0.013175955820676025,
  "test_accuracy_text": "0.8798 ± 0.0132",
  "test_macro_precision_mean": 0.8841078821315743,
  "test_macro_precision_std": 0.012595346716117863,
  "test_macro_precision_text": "0.8841 ± 0.0126",
  "test_macro_recall_mean": 0.8789883582683654,
  "test_macro_recall_std": 0.013142561557312037,
  "test_macro_recall_text": "0.8790 ± 0.0131",
  "dev_macro_f1_mean": 0.8809583207922672,
  "dev_macro_f1_std": 0.002994498567828775,
  "dev_macro_f1_text": "0.8810 ± 0.0030",
  "task": "ViFactCheck-gold-evidence",
  "dataset": "ViFactCheck",
  "model_key": "cafebert",
  "model_name": "CafeBERT",
  "base_model": "uitnlp/CafeBERT",
  "seeds": [
    22,
    42,
    202
  ],
  "representative_seed": 202,
  "selection_rule": "maximum development Macro-F1; seed ascending tie-break",
  "split_policy": "merged_stratified_80_10_10",
  "split_seed": 42,
  "max_length": 256,
  "epochs": 3,
  "effective_batch_size": 8
}