---
title: "Multimodal Chain-of-Thought Reasoning: A Comprehensive Survey"
canonical_url: "https://www.modelscope.cn/papers/127180"
md_url: "https://www.modelscope.cn/papers/127180.md"
arxiv_id: 2503.12605
published: 2025-03-23
last_updated: 2025-03-23
authors:
  - "Yaoting Wang"
  - "Shengqiong Wu"
  - "Yuecheng Zhang"
  - "Shuicheng Yan"
  - "Ziwei Liu"
  - "Jiebo Luo"
  - "Hao Fei"
model_name: MCoT
model_developer: "NUS、CUHK、NTU、UR"
domain:
  - "计算机视觉"
  - "多模态学习"
  - "思维链推理"
  - "大语言模型"
  - "人工智能综述"
type:
  - "计算机视觉"
  - "多模态学习"
  - "思维链推理"
  - "大语言模型"
  - "人工智能综述"
  - "Computer Vision and Pattern Recognition"
arxiv_url: "https://arxiv.org/abs/2503.12605"
pdf_url: "https://arxiv.org/pdf/2503.12605"
code_link: "https://github.com/yaotingwangofficial/Awesome-MCoT"
---

# Multimodal Chain-of-Thought Reasoning: A Comprehensive Survey

> By extending the advantage of chain-of-thought (CoT) reasoning in human-like step-by-step processes to multimodal contexts, multimodal CoT (MCoT) reasoning has recently garnered significant research attention, especially in the integration with multimodal…

「Multimodal Chain-of-Thought Reasoning: A Comprehensive Survey」是 ModelScope 魔搭社区收录的论文，arXiv 2503.12605，作者为 Yaoting Wang, Shengqiong Wu, Yuecheng Zhang et al.，发表于 2025-03-23，属于 计算机视觉、多模态学习、思维链推理 领域。

- **ArXiv**: 2503.12605
- **Published**: 2025-03-23
- **Authors**: Yaoting Wang, Shengqiong Wu, Yuecheng Zhang, Shuicheng Yan, Ziwei Liu, Jiebo Luo, Hao Fei
- **Model**: MCoT
- **Developer**: NUS、CUHK、NTU、UR
- **Domain**: 计算机视觉, 多模态学习, 思维链推理, 大语言模型, 人工智能综述
- **ArXiv URL**: https://arxiv.org/abs/2503.12605
- **PDF**: https://arxiv.org/pdf/2503.12605
- **Code**: https://github.com/yaotingwangofficial/Awesome-MCoT

Source: https://www.modelscope.cn/papers/127180

---

> 多模态思维链推理：一项全面综述

## 摘要

本文是首篇系统性综述多模态思维链（MCoT）推理的论文。文章将思维链（CoT）逐步推理的优势从纯文本扩展至图像、视频、音频、3D等多模态场景，提出了涵盖提示构建、结构化推理、信息增强、目标粒度、多模态理由及测试时计算扩展等六个维度的详细分类体系。论文全面梳理了MCoT在具身智能、自动驾驶、医疗AI、多模态生成等领域的应用，汇总了相关数据集与基准测试，并探讨了当前面临的挑战与未来研究方向。

## Abstract

By extending the advantage of chain-of-thought (CoT) reasoning in human-like step-by-step processes to multimodal contexts, multimodal CoT (MCoT) reasoning has recently garnered significant research attention, especially in the integration with multimodal large language models (MLLMs). Existing MCoT studies design various methodologies and innovative reasoning paradigms to address the unique challenges of image, video, speech, audio, 3D, and structured data across different modalities, achieving extensive success in applications such as robotics, healthcare, autonomous driving, and multimodal generation. However, MCoT still presents distinct challenges and opportunities that require further focus to ensure consistent thriving in this field, where, unfortunately, an up-to-date review of this domain is lacking. To bridge this gap, we present the first systematic survey of MCoT reasoning, elucidating the relevant foundational concepts and definitions. We offer a comprehensive taxonomy and an in-depth analysis of current methodologies from diverse perspectives across various application scenarios. Furthermore, we provide insights into existing challenges and future research directions, aiming to foster innovation toward multimodal AGI.