Work

Transferring Textual Preferences to Vision-Language Understanding through Model Merging

Year: 2025

Type: conference-paper

Abstract: Large vision-language models (LVLMs) perform outstandingly across various multimodal tasks.However, their ability to evaluate generated content remains limited, and training vision-language reward models (VLRMs) with preference data is computationally expensive.This paper explores a training-free alternative by merging text-based reward models (RMs) with LVLMs to create VLRMs.Our approach shows that integrating these models leads to improved performance over LVLMs' scoring and text-based RMs, offering an efficient method for incorporating textual preferences into LVLMs.The code and data are publicly available at https://github.com/ lca0503/MergeToVLRM.

Source: Annual Meeting of the Association for Computational Linguistics (ACL)

DOI: 10.18653/v1/2025.acl-short.72

Authors: Chen-An Li, Tzu-Han Lin, Yun-Nung Chen, Hung-yi Lee

Institution: National Taiwan University

Language: English

Topic: Multimodal Machine Learning Applications

Subfield: Computer Vision and Pattern Recognition

Field: Computer Science

Domain: Physical Sciences

Open Access status: hybrid

Funders: National Applied Research Laboratories, National Taiwan University, National Science and Technology Council

OpenAlex ID: W4412889482