Authors
Wenyi Xiao, Zechuan Wang, Leilei Gan, Shuai Zhao, Zongyue Li, Ruirui Lei, Wanggui He, Luu Anh Tuan, Long Chan, Hao Jiang, Zhou Zhao, Fei Wu
Published in
IEEE transactions on pattern analysis and machine intelligence. Volume PP. Jun 16, 2026. Epub Jun 16, 2026.
Abstract
With the rapid advancement of large language models (LLMs), aligning policy models with human preferences has become increasingly critical. Direct Preference Optimization (DPO) has emerged as a promising approach for alignment, acting as an RL-free alternative to Reinforcement Learning from Human Feedback (RLHF). Despite DPO's various advancements and inherent limitations, an in-depth review of these aspects is currently lacking in the literature. In this work, we present a review of the challenges and opportunities in DPO, covering theoretical analyses, variants, relevant preference datasets, and applications. Specifically, we categorize recent studies on DPO based on key research questions to provide a thorough understanding of DPO's current landscape. Additionally, we propose several future research directions to offer insights on model alignment for the research community.
PMID:
42301828
Bibliographic data and abstract were imported from PubMed on 17 Jun 2026.
Read full publication at:
Please sign in
to see all details.
Advertisement
Stats
- Recommendations n/a n/a positive of 0 vote(s)
- Views 9
- Comments 0