Interpretable Diffusion Models with B-cos Networks

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Bernold, Nicola, Vandenhirtz, Moritz, Bizeul, Alice, Vogt, Julia E.
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866915372279529472
author Bernold, Nicola
Vandenhirtz, Moritz
Bizeul, Alice
Vogt, Julia E.
author_facet Bernold, Nicola
Vandenhirtz, Moritz
Bizeul, Alice
Vogt, Julia E.
contents Text-to-image diffusion models generate images by iteratively denoising random noise, conditioned on a prompt. While these models have enabled impressive progress in image generation, they often fail to accurately reflect all semantic information described in the prompt -- failures that are difficult to detect automatically. In this work, we introduce a diffusion model architecture built with B-cos modules that offers inherent interpretability. Our approach provides insight into how individual prompt tokens affect the generated image by producing explanations that highlight the pixel regions influenced by each token. We demonstrate that B-cos diffusion models can produce high-quality images while providing meaningful insights into prompt-image alignment.
format Preprint
id arxiv_https___arxiv_org_abs_2507_03846
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Interpretable Diffusion Models with B-cos Networks
Bernold, Nicola
Vandenhirtz, Moritz
Bizeul, Alice
Vogt, Julia E.
Computer Vision and Pattern Recognition
Machine Learning
Text-to-image diffusion models generate images by iteratively denoising random noise, conditioned on a prompt. While these models have enabled impressive progress in image generation, they often fail to accurately reflect all semantic information described in the prompt -- failures that are difficult to detect automatically. In this work, we introduce a diffusion model architecture built with B-cos modules that offers inherent interpretability. Our approach provides insight into how individual prompt tokens affect the generated image by producing explanations that highlight the pixel regions influenced by each token. We demonstrate that B-cos diffusion models can produce high-quality images while providing meaningful insights into prompt-image alignment.
title Interpretable Diffusion Models with B-cos Networks
topic Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2507.03846