Full accuracy+F1 metrics: laya zero-shot, laya fine-tuned, setfit (95 test rows)

Add full-metric evaluation script (accuracy + F1 for laya zero-shot, laya fine-tuned, setfit)

Add evaluation metrics

Laya fine-tuned on GIZ vulnerability data: macro-F1 0.654 (zero-shot 0.108)

Add evaluation metrics

Laya fine-tuned on GIZ vulnerability data: macro-F1 0.000 (zero-shot 0.000)

Add training script: RLCD fine-tune of laya on GIZ vulnerability data (17 binary questions per text)

initial commit