{"ID":2863165,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2510.02390","arxiv_id":"2510.02390","title":"F-scheduler: illuminating the free-lunch design space for fast sampling of diffusion models","abstract":"Diffusion models are the state-of-the-art generative models for high-resolution images, but sampling from pretrained models is computationally expensive, motivating interest in fast sampling. Although Free-U Net is a training-free enhancement for improving image quality, we find it ineffective under few-step ($\u003c10$) sampling. We analyze the discrete diffusion ODE and propose F-scheduler, a scheduler designed for ODE solvers with Free-U Net. Our proposed scheduler consists of a special time schedule that does not fully denoise the feature to enable the use of the KL-term in the $β$-VAE decoder, and the schedule of a proper inference stage for modifying the U-Net skip-connection via Free-U Net. Via information theory, we provide insights into how the better scheduled ODE solvers for the diffusion model can outperform the training-based diffusion distillation model. The newly proposed scheduler is compatible with most of the few-step ODE solvers and can sample a 1024 x 1024-resolution image in 6 steps and a 512 x 512-resolution image in 5 steps when it applies to DPM++ 2m and UniPC, with an FID result that outperforms the SOTA distillation models and the 20-step DPM++ 2m solver, respectively. Codebase: https://github.com/TheLovesOfLadyPurple/F-scheduler","short_abstract":"Diffusion models are the state-of-the-art generative models for high-resolution images, but sampling from pretrained models is computationally expensive, motivating interest in fast sampling. Although Free-U Net is a training-free enhancement for improving image quality, we find it ineffective under few-step ($\u003c10$) sa...","url_abs":"https://arxiv.org/abs/2510.02390","url_pdf":"https://arxiv.org/pdf/2510.02390v3","authors":"[\"Zilai Li\",\"Lujia Bai\"]","published":"2025-09-30T23:27:09Z","proceeding":"cs.GR","tasks":"[\"cs.GR\",\"cs.AI\",\"eess.IV\"]","methods":"[\"Diffusion Model\",\"Variational Autoencoder\"]","has_code":false,"code_links":[{"ID":608979,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_id":2863165,"paper_url":"https://arxiv.org/abs/2510.02390","paper_title":"F-scheduler: illuminating the free-lunch design space for fast sampling of diffusion models","repo_url":"https://github.com/TheLovesOfLadyPurple/F-scheduler","is_official":false,"mentioned_in_paper":false,"mentioned_in_github":true,"github_stars":0}]}