aggregate_metrics.json

{
  "test_macro_f1_mean": 0.6822236082683636,
  "test_macro_f1_std": 0.020221173156763098,
  "test_macro_f1_text": "0.6822 ± 0.0202",
  "test_accuracy_mean": 0.682780847145488,
  "test_accuracy_std": 0.019840203891726714,
  "test_accuracy_text": "0.6828 ± 0.0198",
  "test_macro_precision_mean": 0.6844905259101943,
  "test_macro_precision_std": 0.020775638611263955,
  "test_macro_precision_text": "0.6845 ± 0.0208",
  "test_macro_recall_mean": 0.6814768739986444,
  "test_macro_recall_std": 0.020071255760093012,
  "test_macro_recall_text": "0.6815 ± 0.0201",
  "dev_macro_f1_mean": 0.7188106084932735,
  "dev_macro_f1_std": 0.0063273577854815344,
  "dev_macro_f1_text": "0.7188 ± 0.0063",
  "task": "ViFactCheck-full-context",
  "dataset": "ViFactCheck",
  "model_key": "bn_newsbert",
  "model_name": "BN-NewsBERT",
  "base_model": "BaoNhan/BN-NewsBERT",
  "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
}