Squeezed Diffusion Models Article Swipe

PDF

Jarnail Singh , Sudhir Khanna , James Burgess ·

YOU? · · 2025 · Open Access · · DOI: https://doi.org/10.48550/arxiv.2508.14871

Diffusion models typically inject isotropic Gaussian noise, disregarding structure in the data. Motivated by the way quantum squeezed states redistribute uncertainty according to the Heisenberg uncertainty principle, we introduce Squeezed Diffusion Models (SDM), which scale noise anisotropically along the principal component of the training distribution. As squeezing enhances the signal-to-noise ratio in physics, we hypothesize that scaling noise in a data-dependent manner can better assist diffusion models in learning important data features. We study two configurations: (i) a Heisenberg diffusion model that compensates the scaling on the principal axis with inverse scaling on orthogonal directions and (ii) a standard SDM variant that scales only the principal axis. Counterintuitively, on CIFAR-10/100 and CelebA-64, mild antisqueezing - i.e. increasing variance on the principal axis - consistently improves FID by up to 15% and shifts the precision-recall frontier toward higher recall. Our results demonstrate that simple, data-aware noise shaping can deliver robust generative gains without architectural changes.

Related Topics

List Of Iphone Models

Stable Diffusion

List Of Ipad Models

Role Models

Fick's Laws Of Diffusion

Reaction–Diffusion System

Diffusion-Weighted Magnetic Resonance Imaging

Concepts

No concepts available.

Metadata

Type: preprint
Language: en
Landing Page: http://arxiv.org/abs/2508.14871
PDF: https://arxiv.org/pdf/2508.14871
OA Status: green
OpenAlex ID: https://openalex.org/W4415241229

All OpenAlex metadata

Raw OpenAlex JSON

OpenAlex ID: https://openalex.org/W4415241229

Canonical identifier for this work in OpenAlex
DOI: https://doi.org/10.48550/arxiv.2508.14871

Digital Object Identifier
Title: Squeezed Diffusion Models

Work title
Type: preprint

OpenAlex work type
Language: en

Primary language
Publication year: 2025

Year of publication
Publication date: 2025-08-20

Full publication date if available
Authors: Jarnail Singh, Sudhir Khanna, James Burgess

List of authors in order
Landing page: https://arxiv.org/abs/2508.14871

Publisher landing page
PDF URL: https://arxiv.org/pdf/2508.14871

Direct link to full text PDF
Open access: Yes

Whether a free full text is available
OA status: green

Open access status per OpenAlex
OA URL: https://arxiv.org/pdf/2508.14871

Direct OA link when available
Cited by: 0

Total citation count in OpenAlex

Full payload

id	https://openalex.org/W4415241229
doi	https://doi.org/10.48550/arxiv.2508.14871
ids.doi	https://doi.org/10.48550/arxiv.2508.14871
ids.openalex	https://openalex.org/W4415241229
fwci
type	preprint
title	Squeezed Diffusion Models
biblio.issue
biblio.volume
biblio.last_page
biblio.first_page
topics[0].id	https://openalex.org/T11379
topics[0].field.id	https://openalex.org/fields/15
topics[0].field.display_name	Chemical Engineering
topics[0].score	0.7617999911308289
topics[0].domain.id	https://openalex.org/domains/3
topics[0].domain.display_name	Physical Sciences
topics[0].subfield.id	https://openalex.org/subfields/1507
topics[0].subfield.display_name	Fluid Flow and Transfer Processes
topics[0].display_name	Rheology and Fluid Dynamics Studies
topics[1].id	https://openalex.org/T12100
topics[1].field.id	https://openalex.org/fields/17
topics[1].field.display_name	Computer Science
topics[1].score	0.6758000254631042
topics[1].domain.id	https://openalex.org/domains/3
topics[1].domain.display_name	Physical Sciences
topics[1].subfield.id	https://openalex.org/subfields/1703
topics[1].subfield.display_name	Computational Theory and Mathematics
topics[1].display_name	Advanced Mathematical Modeling in Engineering
is_xpac	False
apc_list
apc_paid
language	en
locations[0].id	pmh:oai:arXiv.org:2508.14871
locations[0].is_oa	True
locations[0].source.id	https://openalex.org/S4306400194
locations[0].source.issn
locations[0].source.type	repository
locations[0].source.is_oa	True
locations[0].source.issn_l
locations[0].source.is_core	False
locations[0].source.is_in_doaj	False
locations[0].source.display_name	arXiv (Cornell University)
locations[0].source.host_organization	https://openalex.org/I205783295
locations[0].source.host_organization_name	Cornell University
locations[0].source.host_organization_lineage	https://openalex.org/I205783295
locations[0].license	cc-by
locations[0].pdf_url	https://arxiv.org/pdf/2508.14871
locations[0].version	submittedVersion
locations[0].raw_type	text
locations[0].license_id	https://openalex.org/licenses/cc-by
locations[0].is_accepted	False
locations[0].is_published	False
locations[0].raw_source_name
locations[0].landing_page_url	http://arxiv.org/abs/2508.14871
locations[1].id	doi:10.48550/arxiv.2508.14871
locations[1].is_oa	True
locations[1].source.id	https://openalex.org/S4306400194
locations[1].source.issn
locations[1].source.type	repository
locations[1].source.is_oa	True
locations[1].source.issn_l
locations[1].source.is_core	False
locations[1].source.is_in_doaj	False
locations[1].source.display_name	arXiv (Cornell University)
locations[1].source.host_organization	https://openalex.org/I205783295
locations[1].source.host_organization_name	Cornell University
locations[1].source.host_organization_lineage	https://openalex.org/I205783295
locations[1].license	cc-by
locations[1].pdf_url
locations[1].version
locations[1].raw_type	article
locations[1].license_id	https://openalex.org/licenses/cc-by
locations[1].is_accepted	False
locations[1].is_published
locations[1].raw_source_name
locations[1].landing_page_url	https://doi.org/10.48550/arxiv.2508.14871
indexed_in	arxiv, datacite
authorships[0].author.id	https://openalex.org/A5101447227
authorships[0].author.orcid	https://orcid.org/0000-0001-6032-2376
authorships[0].author.display_name	Jarnail Singh
authorships[0].author_position	first
authorships[0].raw_author_name	Singh, Jyotirmai
authorships[0].is_corresponding	False
authorships[1].author.id	https://openalex.org/A5113814254
authorships[1].author.orcid
authorships[1].author.display_name	Sudhir Khanna
authorships[1].author_position	middle
authorships[1].raw_author_name	Khanna, Samar
authorships[1].is_corresponding	False
authorships[2].author.id	https://openalex.org/A5064226280
authorships[2].author.orcid	https://orcid.org/0000-0002-0823-2848
authorships[2].author.display_name	James Burgess
authorships[2].author_position	last
authorships[2].raw_author_name	Burgess, James
authorships[2].is_corresponding	False
has_content.pdf	True
has_content.grobid_xml	True
is_paratext	False
open_access.is_oa	True
open_access.oa_url	https://arxiv.org/pdf/2508.14871
open_access.oa_status	green
open_access.any_repository_has_fulltext	False
created_date	2025-10-16T00:00:00
display_name	Squeezed Diffusion Models
has_fulltext	True
is_retracted	False
updated_date	2025-11-06T06:51:31.235846
primary_topic.id	https://openalex.org/T11379
primary_topic.field.id	https://openalex.org/fields/15
primary_topic.field.display_name	Chemical Engineering
primary_topic.score	0.7617999911308289
primary_topic.domain.id	https://openalex.org/domains/3
primary_topic.domain.display_name	Physical Sciences
primary_topic.subfield.id	https://openalex.org/subfields/1507
primary_topic.subfield.display_name	Fluid Flow and Transfer Processes
primary_topic.display_name	Rheology and Fluid Dynamics Studies
cited_by_count	0
locations_count	2
best_oa_location.id	pmh:oai:arXiv.org:2508.14871
best_oa_location.is_oa	True
best_oa_location.source.id	https://openalex.org/S4306400194
best_oa_location.source.issn
best_oa_location.source.type	repository
best_oa_location.source.is_oa	True
best_oa_location.source.issn_l
best_oa_location.source.is_core	False
best_oa_location.source.is_in_doaj	False
best_oa_location.source.display_name	arXiv (Cornell University)
best_oa_location.source.host_organization	https://openalex.org/I205783295
best_oa_location.source.host_organization_name	Cornell University
best_oa_location.source.host_organization_lineage	https://openalex.org/I205783295
best_oa_location.license	cc-by
best_oa_location.pdf_url	https://arxiv.org/pdf/2508.14871
best_oa_location.version	submittedVersion
best_oa_location.raw_type	text
best_oa_location.license_id	https://openalex.org/licenses/cc-by
best_oa_location.is_accepted	False
best_oa_location.is_published	False
best_oa_location.raw_source_name
best_oa_location.landing_page_url	http://arxiv.org/abs/2508.14871
primary_location.id	pmh:oai:arXiv.org:2508.14871
primary_location.is_oa	True
primary_location.source.id	https://openalex.org/S4306400194
primary_location.source.issn
primary_location.source.type	repository
primary_location.source.is_oa	True
primary_location.source.issn_l
primary_location.source.is_core	False
primary_location.source.is_in_doaj	False
primary_location.source.display_name	arXiv (Cornell University)
primary_location.source.host_organization	https://openalex.org/I205783295
primary_location.source.host_organization_name	Cornell University
primary_location.source.host_organization_lineage	https://openalex.org/I205783295
primary_location.license	cc-by
primary_location.pdf_url	https://arxiv.org/pdf/2508.14871
primary_location.version	submittedVersion
primary_location.raw_type	text
primary_location.license_id	https://openalex.org/licenses/cc-by
primary_location.is_accepted	False
primary_location.is_published	False
primary_location.raw_source_name
primary_location.landing_page_url	http://arxiv.org/abs/2508.14871
publication_date	2025-08-20
publication_year	2025
referenced_works_count	0
abstract_inverted_index.-	114, 122
abstract_inverted_index.a	59, 77, 97
abstract_inverted_index.As	45
abstract_inverted_index.We	72
abstract_inverted_index.by	13, 126
abstract_inverted_index.in	9, 51, 58, 67
abstract_inverted_index.of	41
abstract_inverted_index.on	85, 92, 108, 118
abstract_inverted_index.to	22, 128
abstract_inverted_index.up	127
abstract_inverted_index.we	27, 53
abstract_inverted_index.(i)	76
abstract_inverted_index.15%	129
abstract_inverted_index.FID	125
abstract_inverted_index.Our	138
abstract_inverted_index.SDM	99
abstract_inverted_index.and	95, 110, 130
abstract_inverted_index.can	62, 146
abstract_inverted_index.the	10, 14, 23, 38, 42, 48, 83, 86, 104, 119, 132
abstract_inverted_index.two	74
abstract_inverted_index.way	15
abstract_inverted_index.(ii)	96
abstract_inverted_index.axis	88, 121
abstract_inverted_index.data	70
abstract_inverted_index.i.e.	115
abstract_inverted_index.mild	112
abstract_inverted_index.only	103
abstract_inverted_index.that	55, 81, 101, 141
abstract_inverted_index.with	89
abstract_inverted_index.along	37
abstract_inverted_index.axis.	106
abstract_inverted_index.data.	11
abstract_inverted_index.gains	150
abstract_inverted_index.model	80
abstract_inverted_index.noise	35, 57, 144
abstract_inverted_index.ratio	50
abstract_inverted_index.scale	34
abstract_inverted_index.study	73
abstract_inverted_index.which	33
abstract_inverted_index.(SDM),	32
abstract_inverted_index.Models	31
abstract_inverted_index.assist	64
abstract_inverted_index.better	63
abstract_inverted_index.higher	136
abstract_inverted_index.inject	3
abstract_inverted_index.manner	61
abstract_inverted_index.models	1, 66
abstract_inverted_index.noise,	6
abstract_inverted_index.robust	148
abstract_inverted_index.scales	102
abstract_inverted_index.shifts	131
abstract_inverted_index.states	18
abstract_inverted_index.toward	135
abstract_inverted_index.deliver	147
abstract_inverted_index.inverse	90
abstract_inverted_index.quantum	16
abstract_inverted_index.recall.	137
abstract_inverted_index.results	139
abstract_inverted_index.scaling	56, 84, 91
abstract_inverted_index.shaping	145
abstract_inverted_index.simple,	142
abstract_inverted_index.variant	100
abstract_inverted_index.without	151
abstract_inverted_index.Gaussian	5
abstract_inverted_index.Squeezed	29
abstract_inverted_index.changes.	153
abstract_inverted_index.enhances	47
abstract_inverted_index.frontier	134
abstract_inverted_index.improves	124
abstract_inverted_index.learning	68
abstract_inverted_index.physics,	52
abstract_inverted_index.squeezed	17
abstract_inverted_index.standard	98
abstract_inverted_index.training	43
abstract_inverted_index.variance	117
abstract_inverted_index.Diffusion	0, 30
abstract_inverted_index.Motivated	12
abstract_inverted_index.according	21
abstract_inverted_index.component	40
abstract_inverted_index.diffusion	65, 79
abstract_inverted_index.features.	71
abstract_inverted_index.important	69
abstract_inverted_index.introduce	28
abstract_inverted_index.isotropic	4
abstract_inverted_index.principal	39, 87, 105, 120
abstract_inverted_index.squeezing	46
abstract_inverted_index.structure	8
abstract_inverted_index.typically	2
abstract_inverted_index.CelebA-64,	111
abstract_inverted_index.Heisenberg	24, 78
abstract_inverted_index.data-aware	143
abstract_inverted_index.directions	94
abstract_inverted_index.generative	149
abstract_inverted_index.increasing	116
abstract_inverted_index.orthogonal	93
abstract_inverted_index.principle,	26
abstract_inverted_index.compensates	82
abstract_inverted_index.demonstrate	140
abstract_inverted_index.hypothesize	54
abstract_inverted_index.uncertainty	20, 25
abstract_inverted_index.CIFAR-10/100	109
abstract_inverted_index.consistently	123
abstract_inverted_index.disregarding	7
abstract_inverted_index.redistribute	19
abstract_inverted_index.antisqueezing	113
abstract_inverted_index.architectural	152
abstract_inverted_index.distribution.	44
abstract_inverted_index.data-dependent	60
abstract_inverted_index.anisotropically	36
abstract_inverted_index.configurations:	75
abstract_inverted_index.signal-to-noise	49
abstract_inverted_index.precision-recall	133
abstract_inverted_index.Counterintuitively,	107
cited_by_percentile_year
countries_distinct_count	0
institutions_distinct_count	3
citation_normalized_percentile