MathLLMs/ImgCode-8.6M

---
license: apache-2.0
language:
- en
metrics:
- accuracy
pipeline_tag: image-text-to-text
dataset_info:
- config_name: Img2Python-00
features:
- name: id
dtype: int64
- name: image
dtype: image
- name: text
dtype: string
- name: source
dtype: string
- config_name: Img2TikZ-00
features:
- name: id
dtype: int64
- name: image
dtype: image
- name: text
dtype: string
- name: source
dtype: string
- config_name: Img2Python-01
features:
- name: id
dtype: int64
- name: image
dtype: image
- name: text
dtype: string
- name: source
dtype: string
- config_name: Img2TikZ-01
features:
- name: id
dtype: int64
- name: image
dtype: image
- name: text
dtype: string
- name: source
dtype: string
- config_name: Img2Python-02
features:
- name: id
dtype: int64
- name: image
dtype: image
- name: text
dtype: string
- name: source
dtype: string
- config_name: Img2TikZ-02
features:
- name: id
dtype: int64
- name: image
dtype: image
- name: text
dtype: string
- name: source
dtype: string
- config_name: Img2Python-03
features:
- name: id
dtype: int64
- name: image
dtype: image
- name: text
dtype: string
- name: source
dtype: string
- config_name: Img2TikZ-03
features:
- name: id
dtype: int64
- name: image
dtype: image
- name: text
dtype: string
- name: source
dtype: string
- config_name: Img2Python-04
features:
- name: id
dtype: int64
- name: image
dtype: image
- name: text
dtype: string
- name: source
dtype: string
- config_name: Img2TikZ-04
features:
- name: id
dtype: int64
- name: image
dtype: image
- name: text
dtype: string
- name: source
dtype: string
- config_name: Img2Python-05
features:
- name: id
dtype: int64
- name: image
dtype: image
- name: text
dtype: string
- name: source
dtype: string
- config_name: Img2TikZ-05
features:
- name: id
dtype: int64
- name: image
dtype: image
- name: text
dtype: string
- name: source
dtype: string
- config_name: Img2Python-06
features:
- name: id
dtype: int64
- name: image
dtype: image
- name: text
dtype: string
- name: source
dtype: string
- config_name: Img2TikZ-06
features:
- name: id
dtype: int64
- name: image
dtype: image
- name: text
dtype: string
- name: source
dtype: string
- config_name: Img2Python-07
features:
- name: id
dtype: int64
- name: image
dtype: image
- name: text
dtype: string
- name: source
dtype: string
- config_name: Img2TikZ-07
features:
- name: id
dtype: int64
- name: image
dtype: image
- name: text
dtype: string
- name: source
dtype: string
- config_name: Img2Python-08
features:
- name: id
dtype: int64
- name: image
dtype: image
- name: text
dtype: string
- name: source
dtype: string
- config_name: Img2TikZ-08
features:
- name: id
dtype: int64
- name: image
dtype: image
- name: text
dtype: string
- name: source
dtype: string
- config_name: Img2Python-09
features:
- name: id
dtype: int64
- name: image
dtype: image
- name: text
dtype: string
- name: source
dtype: string
- config_name: Img2TikZ-09
features:
- name: id
dtype: int64
- name: image
dtype: image
- name: text
dtype: string
- name: source
dtype: string
- config_name: Img2Python-10
features:
- name: id
dtype: int64
- name: image
dtype: image
- name: text
dtype: string
- name: source
dtype: string
- config_name: Img2TikZ-10
features:
- name: id
dtype: int64
- name: image
dtype: image
- name: text
dtype: string
- name: source
dtype: string
- config_name: Img2Python-11
features:
- name: id
dtype: int64
- name: image
dtype: image
- name: text
dtype: string
- name: source
dtype: string
- config_name: Img2TikZ-11
features:
- name: id
dtype: int64
- name: image
dtype: image
- name: text
dtype: string
- name: source
dtype: string
- config_name: Img2Python-12
features:
- name: id
dtype: int64
- name: image
dtype: image
- name: text
dtype: string
- name: source
dtype: string
- config_name: Img2TikZ-12
features:
- name: id
dtype: int64
- name: image
dtype: image
- name: text
dtype: string
- name: source
dtype: string
- config_name: Img2Python-13
features:
- name: id
dtype: int64
- name: image
dtype: image
- name: text
dtype: string
- name: source
dtype: string
- config_name: Img2TikZ-13
features:
- name: id
dtype: int64
- name: image
dtype: image
- name: text
dtype: string
- name: source
dtype: string
- config_name: Img2Python-14
features:
- name: id
dtype: int64
- name: image
dtype: image
- name: text
dtype: string
- name: source
dtype: string
- config_name: Img2TikZ-14
features:
- name: id
dtype: int64
- name: image
dtype: image
- name: text
dtype: string
- name: source
dtype: string
- config_name: Img2Python-15
features:
- name: id
dtype: int64
- name: image
dtype: image
- name: text
dtype: string
- name: source
dtype: string
- config_name: Img2TikZ-15
features:
- name: id
dtype: int64
- name: image
dtype: image
- name: text
dtype: string
- name: source
dtype: string
- config_name: Img2Python-16
features:
- name: id
dtype: int64
- name: image
dtype: image
- name: text
dtype: string
- name: source
dtype: string
- config_name: Img2TikZ-16
features:
- name: id
dtype: int64
- name: image
dtype: image
- name: text
dtype: string
- name: source
dtype: string
- config_name: Img2Python-17
features:
- name: id
dtype: int64
- name: image
dtype: image
- name: text
dtype: string
- name: source
dtype: string
- config_name: Img2TikZ-17
features:
- name: id
dtype: int64
- name: image
dtype: image
- name: text
dtype: string
- name: source
dtype: string
- config_name: Img2Python-18
features:
- name: id
dtype: int64
- name: image
dtype: image
- name: text
dtype: string
- name: source
dtype: string
- config_name: Img2TikZ-18
features:
- name: id
dtype: int64
- name: image
dtype: image
- name: text
dtype: string
- name: source
dtype: string
- config_name: Img2Python-19
features:
- name: id
dtype: int64
- name: image
dtype: image
- name: text
dtype: string
- name: source
dtype: string
- config_name: Img2TikZ-19
features:
- name: id
dtype: int64
- name: image
dtype: image
- name: text
dtype: string
- name: source
dtype: string
configs:
- config_name: Img2Python-00
data_files:
- split: train
path: Python/train-00000-of-00020.parquet
- config_name: Img2TikZ-00
data_files:
- split: train
path: TikZ/train-00000-of-00020.parquet
- config_name: Img2Python-01
data_files:
- split: train
path: Python/train-00001-of-00020.parquet
- config_name: Img2TikZ-01
data_files:
- split: train
path: TikZ/train-00001-of-00020.parquet
- config_name: Img2Python-02
data_files:
- split: train
path: Python/train-00002-of-00020.parquet
- config_name: Img2TikZ-02
data_files:
- split: train
path: TikZ/train-00002-of-00020.parquet
- config_name: Img2Python-03
data_files:
- split: train
path: Python/train-00003-of-00020.parquet
- config_name: Img2TikZ-03
data_files:
- split: train
path: TikZ/train-00003-of-00020.parquet
- config_name: Img2Python-04
data_files:
- split: train
path: Python/train-00004-of-00020.parquet
- config_name: Img2TikZ-04
data_files:
- split: train
path: TikZ/train-00004-of-00020.parquet
- config_name: Img2Python-05
data_files:
- split: train
path: Python/train-00005-of-00020.parquet
- config_name: Img2TikZ-05
data_files:
- split: train
path: TikZ/train-00005-of-00020.parquet
- config_name: Img2Python-06
data_files:
- split: train
path: Python/train-00006-of-00020.parquet
- config_name: Img2TikZ-06
data_files:
- split: train
path: TikZ/train-00006-of-00020.parquet
- config_name: Img2Python-07
data_files:
- split: train
path: Python/train-00007-of-00020.parquet
- config_name: Img2TikZ-07
data_files:
- split: train
path: TikZ/train-00007-of-00020.parquet
- config_name: Img2Python-08
data_files:
- split: train
path: Python/train-00008-of-00020.parquet
- config_name: Img2TikZ-08
data_files:
- split: train
path: TikZ/train-00008-of-00020.parquet
- config_name: Img2Python-09
data_files:
- split: train
path: Python/train-00009-of-00020.parquet
- config_name: Img2TikZ-09
data_files:
- split: train
path: TikZ/train-00009-of-00020.parquet
- config_name: Img2Python-10
data_files:
- split: train
path: Python/train-00010-of-00020.parquet
- config_name: Img2TikZ-10
data_files:
- split: train
path: TikZ/train-00010-of-00020.parquet
- config_name: Img2Python-11
data_files:
- split: train
path: Python/train-00011-of-00020.parquet
- config_name: Img2TikZ-11
data_files:
- split: train
path: TikZ/train-00011-of-00020.parquet
- config_name: Img2Python-12
data_files:
- split: train
path: Python/train-00012-of-00020.parquet
- config_name: Img2TikZ-12
data_files:
- split: train
path: TikZ/train-00012-of-00020.parquet
- config_name: Img2Python-13
data_files:
- split: train
path: Python/train-00013-of-00020.parquet
- config_name: Img2TikZ-13
data_files:
- split: train
path: TikZ/train-00013-of-00020.parquet
- config_name: Img2Python-14
data_files:
- split: train
path: Python/train-00014-of-00020.parquet
- config_name: Img2TikZ-14
data_files:
- split: train
path: TikZ/train-00014-of-00020.parquet
- config_name: Img2Python-15
data_files:
- split: train
path: Python/train-00015-of-00020.parquet
- config_name: Img2TikZ-15
data_files:
- split: train
path: TikZ/train-00015-of-00020.parquet
- config_name: Img2Python-16
data_files:
- split: train
path: Python/train-00016-of-00020.parquet
- config_name: Img2TikZ-16
data_files:
- split: train
path: TikZ/train-00016-of-00020.parquet
- config_name: Img2Python-17
data_files:
- split: train
path: Python/train-00017-of-00020.parquet
- config_name: Img2TikZ-17
data_files:
- split: train
path: TikZ/train-00017-of-00020.parquet
- config_name: Img2Python-18
data_files:
- split: train
path: Python/train-00018-of-00020.parquet
- config_name: Img2TikZ-18
data_files:
- split: train
path: TikZ/train-00018-of-00020.parquet
- config_name: Img2Python-19
data_files:
- split: train
path: Python/train-00019-of-00020.parquet
- config_name: Img2TikZ-19
data_files:
- split: train
path: TikZ/train-00019-of-00020.parquet
task_categories:
- image-to-text
- text-generation
- image-text-to-text
- visual-question-answering
tags:
- Geometry
- Diagrams
- Charts
- Tables
- Graphs
- Figures
- Plots
- Math
size_categories:
- 1M<n<10M
---
# MathCoder-VL: Bridging Vision and Code for Enhanced Multimodal Mathematical Reasoning

Repo: [https://github.com/mathllm/MathCoder](https://github.com/mathllm/MathCoder)

Paper: [https://huggingface.co/papers/2505.10557](https://huggingface.co/papers/2505.10557)

## Introduction
We introduce MathCoder-VL, a series of open-source large multimodal models (LMMs) specifically tailored for general math problem-solving. We also introduce [FigCodifier-8B](https://huggingface.co/MathLLMs/FigCodifier), an image-to-code model.

| Base Model |Ours |
|-------------------------------------------------------------------|-----------------------------------------------------------------------|
| [Mini-InternVL-Chat-2B-V1-5](https://huggingface.co/OpenGVLab/Mini-InternVL-Chat-2B-V1-5) | [MathCoder-VL-2B](https://huggingface.co/MathLLMs/MathCoder-VL-2B) |
| [InternVL2-8B](https://huggingface.co/OpenGVLab/InternVL2-8B) | [MathCoder-VL-8B](https://huggingface.co/MathLLMs/MathCoder-VL-8B)|
| [InternVL2-8B](https://huggingface.co/OpenGVLab/InternVL2-8B) | [FigCodifier-8B](https://huggingface.co/MathLLMs/FigCodifier)|

## Usage

```
from PIL import Image
from io import BytesIO
from datasets import load_dataset

dataset = load_dataset("parquet", data_files="MathLLMs/Img2Code/TikZ/train-00000-of-00020.parquet", split="train")
print(dataset)
print(dataset[0])
img = Image.open(BytesIO(dataset[0]['image']))
img.show()
```

## Distribution

We randomly extracted 1% samples from our Img2Code data and manually classified them. As shown in the table below, we observe that plane geometry images constitute the largest portion (approximately 48.6%), with triangle-related images making up 14.19%. Various chart types (13.0%) are the second-largest category.

| Category | Subcategory | Percentage |
| :------------------------------ | :-------------------- | :--------- |
| Plane Geometry (48.6%) | Triangle-related | 14.19% |
| | Quadrilateral-related | 11.74% |
| | Circle-related | 9.77% |
| | Line-related | 5.12% |
| | Other | 7.79% |
| Solid Geometry (4.5%) | — | 4.53% |
| Analytical Geometry (7.9%) | — | 7.91% |
| Charts & Plots (13.0%) | Bar Chart | 5.12% |
| | Line Graph | 5.35% |
| | Pie Chart | 1.05% |
| | Scatter Plot | 0.47% |
| | Box Plot | 0.23% |
| | 3D Plot | 0.81% |
| Function-related Figures (4.7%) | — | 4.65% |
| Diagrams (4.5%) | Flow Chart | 1.74% |
| | Relationship Diagram | 2.79% |
| Equations & Texts (7.9%) | — | 7.91% |
| Tables (2.7%) | — | 2.67% |
| Grid-based Graphs (3.8%) | — | 3.84% |
| Other Figures (2.3%) | — | 2.33% |

## **Citation**

Please cite the paper if you use our data, model or code.

```
@inproceedings{
wang2025mathcodervl,
title={MathCoder-{VL}: Bridging Vision and Code for Enhanced Multimodal Mathematical Reasoning},
author={Ke Wang and Junting Pan and Linda Wei and Aojun Zhou and Weikang Shi and Zimu Lu and Han Xiao and Yunqiao Yang and Houxing Ren and Mingjie Zhan and Hongsheng Li},
booktitle={The 63rd Annual Meeting of the Association for Computational Linguistics},
year={2025},
url={https://openreview.net/forum?id=nuvtX1imAb}
}

@inproceedings{
wang2024mathcoder,
title={MathCoder: Seamless Code Integration in {LLM}s for Enhanced Mathematical Reasoning},
author={Ke Wang and Houxing Ren and Aojun Zhou and Zimu Lu and Sichun Luo and Weikang Shi and Renrui Zhang and Linqi Song and Mingjie Zhan and Hongsheng Li},
booktitle={The Twelfth International Conference on Learning Representations},
year={2024},
url={https://openreview.net/forum?id=z8TW0ttBPp}
}
```