Skip to content

Dataset · Rapidata

Human pairwise evaluations for SVG graphics generation

Rapidata/text-to-svg-human-feedback-benchmark

Over half a million vector graphic comparisons of 75 models across text prompts with 3.4 million human votes.

What's inside

The English dataset contains 517,590 rows in parquet format and over 3.4 million evaluations from real humans collected via the Rapidata Python SDK. Each record represents a pairwise comparison of static SVG images generated by two of the 75 advanced LLMs for a single text prompt.

What you can do with it

  • Training reward models and alignment (RLHF) of neural networks for vector graphics generation
  • Benchmarking SVG generation quality from text prompts
Tasks
Text → image, Image classification, Reinforcement learning
Languages
English
Size
100K–1M rows
Downloads this month
5.6K
Likes
39

How to load it

from datasets import load_dataset

ds = load_dataset("Rapidata/text-to-svg-human-feedback-benchmark")