Video Generation Latent Rewards

Use pre-trained video generation models (VGMs) as efficient reward models by evaluating video quality directly in latent space at any denoising timestep, enabling process-reward learning across the entire generation trajectory without expensive VAE decoding to RGB.

adu2021 Updated 7 repo stars

File contents

adu2021/skillxiv commit f554e249a3

Frequently asked questions

npx get-skill add adu2021/video-generation-latent-rewards