zai-org/ImageReward

[NeurIPS 2023] ImageReward: Learning and Evaluating Human Preferences for Text-to-image Generation

48
/ 100
Emerging

Trains on 137k expert preference comparisons to score text-to-image outputs, outperforming CLIP and BLIP-based metrics by 30-40%. Provides ReFL (Reward Feedback Learning) for direct diffusion model fine-tuning via reinforcement learning, improving Stable Diffusion win rates by 58.4% in human evaluation. Integrates with Stable Diffusion Web UI and HuggingFace ecosystem via a lightweight PyPI package with three-line inference API.

1,649 stars.

No Package No Dependents
Maintenance 6 / 25
Adoption 10 / 25
Maturity 16 / 25
Community 16 / 25

How are scores calculated?

Stars

1,649

Forks

89

Language

Python

License

Apache-2.0

Last pushed

Oct 29, 2025

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/diffusion/zai-org/ImageReward"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.