PerPO: Perceptual Preference Optimization via Discriminative Rewarding

Zining Zhu, Liang Zhao, Kangheng Lin, Jinze Yang, En Yu, Chenglong Liu, Haoran Wei, Jianjian Sun, Zheng Ge, Xiangyu Zhang

Keywords: MLLMs; RLHF; DPO

TL;DR: Perceptual Preference Optimization enhances MLLMs' visual comprehension ability through discriminative rewarding.

Abstract:

This paper presents Perceptual Preference Optimization (PerPO), a perception alignment method aimed at addressing the visual discrimination challenges in generative pre-trained multimodal large language models (MLLMs). PerPO employs discriminative rewarding and listwise preference optimization to align MLLMs with human visual perception processes. By utilizing the reward as a quantitative margin for ranking, our method effectively bridges generative preference optimization and discriminative empirical risk minimization. PerPO significantly enhances MLLMs’ visual discrimination capabilities while maintaining their generative strengths, mitigates image-unconditional reward hacking, and ensures consistent performance across visual tasks. This work marks a crucial step towards more perceptually aligned and versatile MLLMs. We also anticipate that PerPO will inspire the community to reconsider MLLM alignment strategies.

Primary Area: foundation or frontier models, including LLMs

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 1779