AIbase
Product LibraryTool Navigation

AlignProp

Public

AlignProp uses direct reward backpropogation for the alignment of large-scale text-to-image diffusion models. Our method is 25x more sample and compute efficient than reinforcement learning methods (PPO) for finetuning Stable Diffusion

Creat2023-10-06T08:03:26
Update2025-03-20T12:19:47
https://align-prop.github.io/
281
Stars
0
Stars Increase

Related projects