Skip to content

Dataset · Qwen

Qwen Image Generation Benchmark

Qwen/Qwen-Image-Bench

An English and Chinese image-text dataset for deep evaluation of Text-to-Image models.

What's inside

Containing between 1,000 and 10,000 examples, this dataset features English and Chinese texts and images. Created by the Qwen authors, it assesses image generators from a content creator's perspective, going beyond simple prompt adherence.

What you can do with it

  • Evaluating the quality of text-to-image generation models
  • Testing multimodal models and judge models (Q-Judger)
  • Comparing image generators on English and Chinese prompts
Tasks
Image → text
Languages
English, Chinese
Size
1K–10K rows
Downloads this month
15K
Likes
53

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("Qwen/Qwen-Image-Bench")