Local Intrinsic Dimension Unveils Hallucinations in Diffusion Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Sobieski, Bartlomiej, Tivnan, Matthew, Płudowski, Dawid, Włodarczyk, Michał Jan, Jin, Pengfei, Biecek, Przemyslaw, Li, Quanzheng
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917464503222272
author Sobieski, Bartlomiej
Tivnan, Matthew
Płudowski, Dawid
Włodarczyk, Michał Jan
Jin, Pengfei
Biecek, Przemyslaw
Li, Quanzheng
author_facet Sobieski, Bartlomiej
Tivnan, Matthew
Płudowski, Dawid
Włodarczyk, Michał Jan
Jin, Pengfei
Biecek, Przemyslaw
Li, Quanzheng
contents Diffusion models are prone to generating structural hallucinations - samples that match the statistical properties of the training data yet defy underlying structural rules, resulting in anomalies like hands with more than five fingers. Recent research studied this failure mode from several viewpoints, offering partial explanations to their occurrence, such as mode interpolation. In this work, we propose a complementary perspective that treats hallucinations as instabilities on the model-induced manifold. We begin by showing that a hallucination filter based on such instabilities matches or exceeds the performance of the recently proposed temporal one. By tracing the source of these instabilities, we identify local intrinsic dimension (LID) as their primary driver and propose Intrinsic Quenching (IQ), a direct corrective mechanism that deflates it to alleviate hallucinations. IQ consistently outperforms standard hallucination reduction baselines across a wide array of benchmarks and offers a highly promising solution for enforcing anatomical consistency in downstream medical imaging tasks.
format Preprint
id arxiv_https___arxiv_org_abs_2605_05026
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Local Intrinsic Dimension Unveils Hallucinations in Diffusion Models
Sobieski, Bartlomiej
Tivnan, Matthew
Płudowski, Dawid
Włodarczyk, Michał Jan
Jin, Pengfei
Biecek, Przemyslaw
Li, Quanzheng
Computer Vision and Pattern Recognition
Artificial Intelligence
Diffusion models are prone to generating structural hallucinations - samples that match the statistical properties of the training data yet defy underlying structural rules, resulting in anomalies like hands with more than five fingers. Recent research studied this failure mode from several viewpoints, offering partial explanations to their occurrence, such as mode interpolation. In this work, we propose a complementary perspective that treats hallucinations as instabilities on the model-induced manifold. We begin by showing that a hallucination filter based on such instabilities matches or exceeds the performance of the recently proposed temporal one. By tracing the source of these instabilities, we identify local intrinsic dimension (LID) as their primary driver and propose Intrinsic Quenching (IQ), a direct corrective mechanism that deflates it to alleviate hallucinations. IQ consistently outperforms standard hallucination reduction baselines across a wide array of benchmarks and offers a highly promising solution for enforcing anatomical consistency in downstream medical imaging tasks.
title Local Intrinsic Dimension Unveils Hallucinations in Diffusion Models
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2605.05026