SFW sampling for diffusion models via external conditioning
Its authors withdrew this paper, so there is no full text to read or summarize.
Lists
Only you see this
Abstract
Score-based generative models (SBM), also known as diffusion models, are the de facto state of the art for image synthesis. Despite their unparalleled performance, SBMs have recently been in the spotlight for being tricked into creating not-safe-for-work (NSFW) content, such as violent images and non-consensual nudity. This article proposes a safe-for-work (SFW) sampler for SBMs implementing a Conditional Trajectory Correction step that guides the samples away from undesired regions in the ambient space using external multimodal models as the source of conditioning. Furthermore, using Contrastive Language Image Pre-training (CLIP), our method admits user-defined NSFW classes, which can vary in different settings. Our experiments on the text-to-image SBM Stable Diffusion validate that the proposed SFW sampler effectively reduces the generation of explicit content, as assessed via independent NSFW detectors. Furthermore, the proposed correction comes at a minor cost in image quality and has an almost null effect on samples that do not need correction. Our study confirms the suitability of the SFW sampler towards aligned SBM models.
BibTeX
@misc{reyes2025sfw,
author = {Camilo Carvajal Reyes and Joaquin Fontbona and Felipe Tobar},
title = {{SFW sampling for diffusion models via external conditioning}},
year = {2025},
url = {https://openreview.net/forum?id=xETLME9sNq},
howpublished = {OpenReview},
note = {Submitted to ICLR 2025},
}