nllg/DaTikZ-V4

---
license: apache-2.0
pipeline_tag: text-generation
task_categories:
- text-generation
size_categories:
- 100K<n<1M
tags:
- tikz
- latex
- code-generation
- scientific-figures
---

# Dataset Card for DaTikZ-V4

DaTikZ-V4 is the dataset used to train **TikZilla-3B**, **TikZilla-3B-RL**, **TikZilla-8B**, and **TikZilla-8B-RL** for generating TikZ/LaTeX figures from natural language descriptions.

The TikZ code has been sourced from **ArXiv**, **GitHub**, and **TeXStackExchange**. Scientific figure descriptions were generated using **Qwen2.5-VL-7B-Instruct**.

## Dataset fields

Each sample contains:

- `file_id`: unique identifier
- `caption`: original caption
- `vlm_description`: detailed visual description generated by a VLM
- `tikz_code`: full LaTeX/TikZ source code
- `source`: data source (e.g., arxiv, github, tex, synthetic)
- `png_image`: rendered image of the TikZ figure

## Installation

```bash
pip install datasets
```

## Usage
```python
from datasets import load_dataset

dataset_id = "nllg/DaTikZ-V4"

ds = load_dataset(dataset_id, split="train")

sample = ds[0]

print(sample["file_id"])
print(sample["caption"])
print(sample["vlm_description"])
print(sample["tikz_code"])
image = sample["png_image"]
```