ZiyiZhang27/sdpo

[IEEE TPAMI] Code for the paper "Aligning Few-Step Diffusion Models with Dense Reward Difference Learning"

32
/ 100
Emerging

This project helps machine learning researchers and practitioners who work with diffusion models to improve the quality of images generated by these models, especially when they need results quickly. It takes an existing few-step diffusion model and fine-tunes it using a specialized reinforcement learning technique. The output is a more refined diffusion model that produces images better aligned with specific quality objectives like aesthetics or human preference.

Use this if you are a machine learning engineer or researcher developing high-performance image generation models and need to align few-step diffusion models with specific quality metrics efficiently.

Not ideal if you are looking for an out-of-the-box image generation tool without needing to fine-tune diffusion model architectures or understand reinforcement learning concepts.

Diffusion Models Image Generation Reinforcement Learning AI Alignment Deep Learning Research
No Package No Dependents
Maintenance 10 / 25
Adoption 6 / 25
Maturity 16 / 25
Community 0 / 25

How are scores calculated?

Stars

19

Forks

Language

Python

License

MIT

Last pushed

Feb 25, 2026

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/diffusion/ZiyiZhang27/sdpo"

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