A-SHOJAEI/preference-guided-image-captioning-alignment

A novel multimodal alignment system that combines image-caption contrastive learning with human preference optimization. By training a vision-language model on Conceptual Captions and then fine-tuning caption generation using UltraFeedback-style preference learning, we create captions that are not just accurate but aligned with human preferences fo

19
/ 100
Experimental
No Package No Dependents
Maintenance 10 / 25
Adoption 0 / 25
Maturity 9 / 25
Community 0 / 25

How are scores calculated?

Stars

Forks

Language

Python

License

MIT

Category

image-captioning

Last pushed

Feb 21, 2026

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/generative-ai/A-SHOJAEI/preference-guided-image-captioning-alignment"

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