sag-uniroma2/Climate_MODEL

Task Title

Climate Alignment

Task Description

Binary classification model for detecting climate-related sustainability disclosures.

Dataset Summary

• Total samples: 586
• Climate: 355
• Non-Climate: 231

Model

• Base model: ClimateBERT (distilroberta-base-climate-f)

Performance

• Accuracy: 93.3% ± 2.7
• Macro F1: 93.0% ± 2.9

Results are reported as Mean ± SD across 10-fold cross-validation, where each fold was used once as the test set.

External Benchmark Evaluation

• Combined training samples: 1,768
• External benchmark accuracy: 90.2%
• External benchmark macro F1: 86.5%
• Internal climate test accuracy: 89.0%
• Internal climate test macro F1: 88.7%