Orpo finetuning on arxiv papers

We'll show how to fine‑tune the Qwen 0.5B model on a curated arXiv dataset, creating a knowledge‑enhanced model for retrieval‑augmented generation.