BatsResearch/csp

Learning to compose soft prompts for compositional zero-shot learning.

36
/ 100
Emerging

This project helps machine learning researchers improve how well vision-language models can understand novel combinations of attributes and objects, like a 'striped blue chair' they've never seen before. It takes existing image-text datasets and a pre-trained vision-language model, then outputs a more accurate model capable of recognizing complex, unseen visual concepts. This is for researchers and practitioners working on advanced computer vision and natural language understanding tasks.

No commits in the last 6 months.

Use this if you need to boost the ability of large pre-trained vision-language models to recognize novel compositions (like 'spotted yellow') without fully retraining the entire model, saving significant computational resources.

Not ideal if you are looking for a pre-packaged solution for a specific business application rather than a research tool for model improvement.

zero-shot learning compositional AI vision-language models computer vision research prompt engineering
Stale 6m No Package No Dependents
Maintenance 2 / 25
Adoption 9 / 25
Maturity 16 / 25
Community 9 / 25

How are scores calculated?

Stars

94

Forks

6

Language

Python

License

BSD-3-Clause

Last pushed

Sep 13, 2025

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/ml-frameworks/BatsResearch/csp"

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