MathFlow-ZJU/MathFlow

Tasks

• image-to-text

Citation text from the card

@misc{chen2025mathflowenhancingperceptualflow,
      title={MathFlow: Enhancing the Perceptual Flow of MLLMs for Visual Mathematical Problems}, 
      author={Felix Chen and Hangjie Yuan and Yunqiu Xu and Tao Feng and Jun Cen and Pengwei Liu and Zeying Huang and Yi Yang},
      year={2025},
      eprint={2503.16549},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2503.16549}, 
}

This dataset is from the paper [MathFlow: Enhancing the Perceptual Flow of MLLMs for Visual Mathematical Problems](https://arxiv.org/abs/2503.16549).

Despite impressive performance across diverse tasks, Multimodal Large Language Models (MLLMs) have yet to fully demonstrate their potential in visual mathematical problem-solving, particularly in accurately perceiving and interpreting diagrams.

Inspired by typical processes of humans, the authors hypothesize that the perception capabilities to extract meaningful information from diagrams is crucial, as it directly impacts subsequent inference processes. To validate this hypothesis, they developed FlowVerse, a comprehensive benchmark that categorizes all information used during problem-solving into four components, which are then combined into six problem versions for evaluation.

All datasets marked Use restrictions | JSON