Dataset · TencentARC
Chart-to-code generation benchmark for MLLMs
TencentARC/Plot2Code
Chart images and English texts to evaluate how multimodal models write code.
What's inside
This benchmark from Tencent ARC Lab contains fewer than 1,000 examples with images and English text. Designed to comprehensively evaluate multimodal language models (MLLMs) on visual programming tasks—generating code from chart images.
What you can do with it
- Evaluating multimodal models' ability to reconstruct plotting code from images
- Comparing visual programming quality across different VLMs and MLLMs
- Tasks
- Text generation, Text → image, Image → text, Image → image
- Languages
- English
- Size
- up to 1K rows
- Downloads this month
- 3.1K
- Likes
- 36
How to load it
from datasets import load_dataset
ds = load_dataset("TencentARC/Plot2Code")