Laplacian Score Sharpening for Mitigating Hallucination in Diffusion Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: C, Barath Chandran., Anumasa, Srinivas, Liu, Dianbo
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908647004569600
author C, Barath Chandran.
Anumasa, Srinivas
Liu, Dianbo
author_facet C, Barath Chandran.
Anumasa, Srinivas
Liu, Dianbo
contents Diffusion models, though successful, are known to suffer from hallucinations that create incoherent or unrealistic samples. Recent works have attributed this to the phenomenon of mode interpolation and score smoothening, but they lack a method to prevent their generation during sampling. In this paper, we propose a post-hoc adjustment to the score function during inference that leverages the Laplacian (or sharpness) of the score to reduce mode interpolation hallucination in unconditional diffusion models across 1D, 2D, and high-dimensional image data. We derive an efficient Laplacian approximation for higher dimensions using a finite-difference variant of the Hutchinson trace estimator. We show that this correction significantly reduces the rate of hallucinated samples across toy 1D/2D distributions and a high-dimensional image dataset. Furthermore, our analysis explores the relationship between the Laplacian and uncertainty in the score.
format Preprint
id arxiv_https___arxiv_org_abs_2511_07496
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Laplacian Score Sharpening for Mitigating Hallucination in Diffusion Models
C, Barath Chandran.
Anumasa, Srinivas
Liu, Dianbo
Computer Vision and Pattern Recognition
Artificial Intelligence
Machine Learning
Diffusion models, though successful, are known to suffer from hallucinations that create incoherent or unrealistic samples. Recent works have attributed this to the phenomenon of mode interpolation and score smoothening, but they lack a method to prevent their generation during sampling. In this paper, we propose a post-hoc adjustment to the score function during inference that leverages the Laplacian (or sharpness) of the score to reduce mode interpolation hallucination in unconditional diffusion models across 1D, 2D, and high-dimensional image data. We derive an efficient Laplacian approximation for higher dimensions using a finite-difference variant of the Hutchinson trace estimator. We show that this correction significantly reduces the rate of hallucinated samples across toy 1D/2D distributions and a high-dimensional image dataset. Furthermore, our analysis explores the relationship between the Laplacian and uncertainty in the score.
title Laplacian Score Sharpening for Mitigating Hallucination in Diffusion Models
topic Computer Vision and Pattern Recognition
Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2511.07496