Dataset · lioooox
Images for T2I logic and composition evaluation
lioooox/T2I-CoReBench-Images
Generated images from complex prompts to check composition and logical relations in text-to-image models.
What's inside
Contains images generated from 1080 complex English prompts and corresponding text descriptions. Sample size ranges from 10k to 100k examples. Data covers composition, cause-and-effect relationships, object relations, and real-world logical reasoning scenarios.
What you can do with it
- Evaluating T2I models' ability to build correct composition and logical relations
- Comparing generation quality of different models on complex prompts
- Training or calibrating reward models for text-to-image
- Tasks
- Text → image
- Languages
- English
- Size
- 10K–100K rows
- Downloads this month
- 408.5K
- Likes
- 6
How to load it
from datasets import load_dataset
ds = load_dataset("lioooox/T2I-CoReBench-Images")