testing-auto-training-model

Fine-tuned multi-label codelist classifier based on svercoutere/robbert-2023-dutch-base-abb.

Labels (4)

• A1.1
• A1.2
• A1.7
• A2

Held-out evaluation

• macro_average_precision: 0.9032
• mean_brier_score: 0.0313
• macro_f1: 0.8580
• micro_f1: 0.8792
• weighted_f1: 0.8789
• macro_precision: 0.8667
• macro_recall: 0.8515
• micro_precision: 0.8841
• micro_recall: 0.8743
• subset_accuracy: 0.8827
• hamming_loss: 0.0336

Validation global threshold: 0.8500. Per-label thresholds are stored in threshold.json.