SJY-1995/GeoTikz-Instruct

---
dataset_id: GeoTikz-Instruct
license: cc-by-nc-4.0
language:
- en
tags:
- image-to-tikz
- geometry
- mathematical-diagrams
- tikz
- plane-geometry
- analytic-geometry
- geo-tikz
pretty_name: GeoTikz-Instruct
size_category: large
task_categories:
- image-to-text
---

# GeoTikzBridge: Advancing Multimodal Code Generation for Geometric Perception and Reasoning [CVPR2026]

## Table of Contents
1. [Dataset Summary](#dataset-summary)
2. [Dataset Structure](#dataset-structure)
3. [Dataset Creation](#dataset-creation)
4. [Citation](#citation)

---

## Dataset Summary
**GeoTikz-Instruct** is a 421K **instruction-based image-to-TikZ dataset**. This dataset provides paired data consisting of original images, instructions for adding auxiliary lines, and the corresponding TikZ codes with auxliary lines added, serving as a foundational resource for research on geometric perception, mathematical reasoning, TikZ code synthesis, etc. The core operations for the construction of auxiliary lines include constructing parallel lines, constructing perpendicular lines, connecting two points, and other related operations.

## Dataset Structure
Each sample in GeoTikz-Instruct is composed of:
1. `image`: image bytes
2. `question`: instruction to adding auxiliary lines
3. `response`: generated tikz code of the figure with auxiliary lines added

## Dataset Creation
The dataset is built using the core innovative strategies:
**Localized Transformation Strategy**
Please refer to our paper for details.

## Citation
Please cite the paper if you use our data, model or code:
```bibtex
@inproceedings{
geotikzbridge,
title={GeoTikzBridge: Advancing Multimodal Code Generation for Geometric Perception and Reasoning},
author={Jiayin Sun and Caixia Sun and Boyu Yang and Hailin Li and Xiao Chen and Yi Zhang and Errui Ding and Liang Li and Chao Deng and Junlan Feng},
booktitle={2026 IEEE/CVF Conference on Computer Vision and Pattern Recognition},
year={2026}
}