# ORPO: Odds Ratio Preference Optimization — Single-Stage SFT+Alignment

> Source: https://sukruyusufkaya.com/learn/fine-tuning-cookbook/ftc-orpo-single-stage-sft-alignment
> Updated: 2026-08-16T01:47:16.648Z
> Category: Fine-Tuning Cookbook (Model-by-Model)
> Module: Part XI — Alignment & Preference Optimization
**TLDR:** ORPO (Hong et al. 2024) — DPO'ya alternatif, SFT base gerektirmiyor. SFT loss + odds-ratio preference loss tek seferde. Ref model gerek yok → memory tasarrufu. Reference-free training, λ hyperparameter, RTX 4090 ORPO Lab.