Perception, Reason, Think, and Plan: A Survey on Large Multimodal Reasoning Models
Yunxin Li, Zhenyu Liu, Zitao Li, Xuanyu Zhang, Zhenran Xu, Xinyu Chen, Haoyuan Shi, Shenyuan Jiang, Xintong Wang, Jifang Wang, Shouzheng Huang, Xinping Zhao, Borui Jiang, Lanqing Hong, Longyue Wang, Zhuotao Tian, Baoxing Huai, Wenhan Luo, Weihua Luo, Zheng Zhang, Baotian Hu, Min Zhang
2025-05-09
Summary
This paper talks about how AI models are getting better at handling and reasoning with different types of information, like images, text, and sounds, all at once. It reviews how these models have evolved from being good at just one thing to becoming more unified and flexible.
What's the problem?
The problem is that earlier AI systems could only work well with one type of data, like just pictures or just words, and they had trouble putting together information from different sources to solve complex problems. This made it hard for AI to understand the world in a way that's similar to how humans do.
What's the solution?
The researchers surveyed the progress in building multimodal reasoning models, showing how the field has moved from using separate modules for each task to creating unified frameworks that can handle many types of data together. They also discussed the new abilities these models are developing and the challenges that still need to be solved, like making sure the AI can reason well across all types of information.
Why it matters?
This matters because the more AI can understand and connect different types of information, the more helpful and intelligent it becomes. This progress opens up new possibilities for smarter digital assistants, better search engines, and more advanced technology that can interact with people and the world in a more natural way.
Abstract
This survey outlines the development of multimodal reasoning models, progressing from task-specific modules to unified frameworks, and discusses emerging capabilities and challenges in integrating reasoning across diverse data types.