feat: upload actual trained LoRA adapter (Qwen2.5-1.5B ORPO, 3 epochs, 36 steps)

Upload model trained with Unsloth

Upload model trained with Unsloth

fix: corrected Colab notebook (judge-format) + fixed ablation HF loader

feat: add model card, inference example, and training scripts

(Trained with Unsloth)

(Trained with Unsloth)

(Trained with Unsloth)

Unsloth Model Card

initial commit