Saved in:
Bibliographic Details
Main Authors: Chen, Xi, Zhu, Mingkang, Liu, Shaoteng, Wu, Xiaoyang, Xu, Xiaogang, Liu, Yu, Bai, Xiang, Zhao, Hengshuang
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2506.22434
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913915578875904
author Chen, Xi
Zhu, Mingkang
Liu, Shaoteng
Wu, Xiaoyang
Xu, Xiaogang
Liu, Yu
Bai, Xiang
Zhao, Hengshuang
author_facet Chen, Xi
Zhu, Mingkang
Liu, Shaoteng
Wu, Xiaoyang
Xu, Xiaogang
Liu, Yu
Bai, Xiang
Zhao, Hengshuang
contents This work explores enabling Chain-of-Thought (CoT) reasoning to link visual cues across multiple images. A straightforward solution is to adapt rule-based reinforcement learning for Vision-Language Models (VLMs). However, such methods typically rely on manually curated question-answer pairs, which can be particularly challenging when dealing with fine grained visual details and complex logic across images. Inspired by self-supervised visual representation learning, we observe that images contain inherent constraints that can serve as supervision. Based on this insight, we construct image triplets comprising two augmented views of the same image and a third, similar but distinct image. During training, the model is prompted to generate a reasoning process to compare these images (i.e., determine same or different). Then we optimize the model with rule-based reinforcement learning. Due to the high visual similarity and the presence of augmentations, the model must attend to subtle visual changes and perform logical reasoning to succeed. Experiments show that, although trained solely on visual comparison tasks, the learned reasoning ability generalizes effectively to a wide range of questions. Without relying on any human-annotated question-answer pairs, our method achieves significant improvements on multi-image reasoning benchmarks and shows strong performance on general vision tasks.
format Preprint
id arxiv_https___arxiv_org_abs_2506_22434
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle MiCo: Multi-image Contrast for Reinforcement Visual Reasoning
Chen, Xi
Zhu, Mingkang
Liu, Shaoteng
Wu, Xiaoyang
Xu, Xiaogang
Liu, Yu
Bai, Xiang
Zhao, Hengshuang
Computer Vision and Pattern Recognition
This work explores enabling Chain-of-Thought (CoT) reasoning to link visual cues across multiple images. A straightforward solution is to adapt rule-based reinforcement learning for Vision-Language Models (VLMs). However, such methods typically rely on manually curated question-answer pairs, which can be particularly challenging when dealing with fine grained visual details and complex logic across images. Inspired by self-supervised visual representation learning, we observe that images contain inherent constraints that can serve as supervision. Based on this insight, we construct image triplets comprising two augmented views of the same image and a third, similar but distinct image. During training, the model is prompted to generate a reasoning process to compare these images (i.e., determine same or different). Then we optimize the model with rule-based reinforcement learning. Due to the high visual similarity and the presence of augmentations, the model must attend to subtle visual changes and perform logical reasoning to succeed. Experiments show that, although trained solely on visual comparison tasks, the learned reasoning ability generalizes effectively to a wide range of questions. Without relying on any human-annotated question-answer pairs, our method achieves significant improvements on multi-image reasoning benchmarks and shows strong performance on general vision tasks.
title MiCo: Multi-image Contrast for Reinforcement Visual Reasoning
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2506.22434