English

PRIMA: Multi-Image Vision-Language Models for Reasoning Segmentation

Computer Vision and Pattern Recognition 2025-12-02 v2 Artificial Intelligence Machine Learning

Abstract

Despite significant advancements in Large Vision-Language Models (LVLMs)' capabilities, existing pixel-grounding models operate in single-image settings, limiting their ability to perform detailed, fine-grained comparisons across multiple images. Conversely, current multi-image understanding models lack pixel-level grounding. Our work addresses this gap by introducing the task of multi-image pixel-grounded reasoning alongside PRIMA, an LVLM that integrates pixel-level grounding with robust multi-image reasoning to produce contextually rich, pixel-grounded explanations. Central to PRIMA is SQuARE, a vision module that injects cross-image relational context into compact query-based visual tokens before fusing them with the language backbone. To support training and evaluation, we curate M4SEG, a new multi-image reasoning segmentation benchmark consisting of \sim744K question-answer pairs that require fine-grained visual understanding across multiple images. PRIMA outperforms state-of-the-art baselines with 7.83%7.83\% and 11.25%11.25\% improvements in Recall and S-IoU, respectively. Ablation studies further demonstrate the effectiveness of the proposed SQuARE module in capturing cross-image relationships.

Keywords

Cite

@article{arxiv.2412.15209,
  title  = {PRIMA: Multi-Image Vision-Language Models for Reasoning Segmentation},
  author = {Muntasir Wahed and Kiet A. Nguyen and Adheesh Sunil Juvekar and Xinzhuo Li and Xiaona Zhou and Vedant Shah and Tianjiao Yu and Pinar Yanardag and Ismini Lourentzou},
  journal= {arXiv preprint arXiv:2412.15209},
  year   = {2025}
}

Comments

Project page: https://plan-lab.github.io/prima

R2 v1 2026-06-28T20:42:48.603Z