Odds Ratio Preference Optimization (ORPO) introduced in paper: ORPO: Monolithic Preference Optimization without Reference Model (Hong et al., 2024).
Subject
Odds Ratio Preference Optimization (ORPO)
Predicate
introduced_in_paper
Object
ORPO: Monolithic Preference Optimization without Reference Model (Hong et al., 2024)
Primary source · preprint · 2024-03-12
ORPO: Monolithic Preference Optimization without Reference Model — arXiv (Hong, Lee, Thorne — KAIST AI)