Efficient Controllable Diffusion via Optimal Classifier Guidance Article Swipe

PDF

Owen Oertell , Shikun Sun , Yiding Chen , Jin Peng Zhou , Zhiyong Wang , Wen Sun ·

YOU? · · 2025 · Open Access · · DOI: https://doi.org/10.48550/arxiv.2505.21666

The controllable generation of diffusion models aims to steer the model to generate samples that optimize some given objective functions. It is desirable for a variety of applications including image generation, molecule generation, and DNA/sequence generation. Reinforcement Learning (RL) based fine-tuning of the base model is a popular approach but it can overfit the reward function while requiring significant resources. We frame controllable generation as a problem of finding a distribution that optimizes a KL-regularized objective function. We present SLCD -- Supervised Learning based Controllable Diffusion, which iteratively generates online data and trains a small classifier to guide the generation of the diffusion model. Similar to the standard classifier-guided diffusion, SLCD's key computation primitive is classification and does not involve any complex concepts from RL or control. Via a reduction to no-regret online learning analysis, we show that under KL divergence, the output from SLCD provably converges to the optimal solution of the KL-regularized objective. Further, we empirically demonstrate that SLCD can generate high quality samples with nearly the same inference time as the base model in both image generation with continuous diffusion and biological sequence generation with discrete diffusion. Our code is available at https://github.com/Owen-Oertell/slcd

Related Topics

Concepts

No concepts available.

Metadata

Type: preprint
Language: en
Landing Page: http://arxiv.org/abs/2505.21666
PDF: https://arxiv.org/pdf/2505.21666
OA Status: green
OpenAlex ID: https://openalex.org/W4416046262

All OpenAlex metadata

Raw OpenAlex JSON

OpenAlex ID: https://openalex.org/W4416046262

Canonical identifier for this work in OpenAlex
DOI: https://doi.org/10.48550/arxiv.2505.21666

Digital Object Identifier
Title: Efficient Controllable Diffusion via Optimal Classifier Guidance

Work title
Type: preprint

OpenAlex work type
Language: en

Primary language
Publication year: 2025

Year of publication
Publication date: 2025-05-27

Full publication date if available
Authors: Owen Oertell, Shikun Sun, Yiding Chen, Jin Peng Zhou, Zhiyong Wang, Wen Sun

List of authors in order
Landing page: https://arxiv.org/abs/2505.21666

Publisher landing page
PDF URL: https://arxiv.org/pdf/2505.21666

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/2505.21666

Direct OA link when available
Cited by: 0

Total citation count in OpenAlex

Full payload

id	https://openalex.org/W4416046262
doi	https://doi.org/10.48550/arxiv.2505.21666
ids.doi	https://doi.org/10.48550/arxiv.2505.21666
ids.openalex	https://openalex.org/W4416046262
fwci
type	preprint
title	Efficient Controllable Diffusion via Optimal Classifier Guidance
biblio.issue
biblio.volume
biblio.last_page
biblio.first_page
is_xpac	False
apc_list
apc_paid
language	en
locations[0].id	pmh:oai:arXiv.org:2505.21666
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
locations[0].pdf_url	https://arxiv.org/pdf/2505.21666
locations[0].version	submittedVersion
locations[0].raw_type	text
locations[0].license_id
locations[0].is_accepted	False
locations[0].is_published	False
locations[0].raw_source_name
locations[0].landing_page_url	http://arxiv.org/abs/2505.21666
locations[1].id	doi:10.48550/arxiv.2505.21666
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.2505.21666
indexed_in	arxiv, datacite
authorships[0].author.id	https://openalex.org/A5095045747
authorships[0].author.orcid
authorships[0].author.display_name	Owen Oertell
authorships[0].author_position	first
authorships[0].raw_author_name	Oertell, Owen
authorships[0].is_corresponding	False
authorships[1].author.id	https://openalex.org/A5101231779
authorships[1].author.orcid
authorships[1].author.display_name	Shikun Sun
authorships[1].author_position	middle
authorships[1].raw_author_name	Sun, Shikun
authorships[1].is_corresponding	False
authorships[2].author.id	https://openalex.org/A5006044524
authorships[2].author.orcid	https://orcid.org/0000-0003-1759-9112
authorships[2].author.display_name	Yiding Chen
authorships[2].author_position	middle
authorships[2].raw_author_name	Chen, Yiding
authorships[2].is_corresponding	False
authorships[3].author.id	https://openalex.org/A5070503869
authorships[3].author.orcid	https://orcid.org/0000-0001-8407-1110
authorships[3].author.display_name	Jin Peng Zhou
authorships[3].author_position	middle
authorships[3].raw_author_name	Zhou, Jin Peng
authorships[3].is_corresponding	False
authorships[4].author.id	https://openalex.org/A5100443922
authorships[4].author.orcid	https://orcid.org/0000-0002-3574-6078
authorships[4].author.display_name	Zhiyong Wang
authorships[4].author_position	middle
authorships[4].raw_author_name	Wang, Zhiyong
authorships[4].is_corresponding	False
authorships[5].author.id	https://openalex.org/A5032266950
authorships[5].author.orcid	https://orcid.org/0000-0001-7086-4910
authorships[5].author.display_name	Wen Sun
authorships[5].author_position	last
authorships[5].raw_author_name	Sun, Wen
authorships[5].is_corresponding	False
has_content.pdf	False
has_content.grobid_xml	False
is_paratext	False
open_access.is_oa	True
open_access.oa_url	https://arxiv.org/pdf/2505.21666
open_access.oa_status	green
open_access.any_repository_has_fulltext	False
created_date	2025-10-10T00:00:00
display_name	Efficient Controllable Diffusion via Optimal Classifier Guidance
has_fulltext	False
is_retracted	False
updated_date	2025-11-28T11:01:23.469902
primary_topic
cited_by_count	0
locations_count	2
best_oa_location.id	pmh:oai:arXiv.org:2505.21666
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
best_oa_location.pdf_url	https://arxiv.org/pdf/2505.21666
best_oa_location.version	submittedVersion
best_oa_location.raw_type	text
best_oa_location.license_id
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/2505.21666
primary_location.id	pmh:oai:arXiv.org:2505.21666
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
primary_location.pdf_url	https://arxiv.org/pdf/2505.21666
primary_location.version	submittedVersion
primary_location.raw_type	text
primary_location.license_id
primary_location.is_accepted	False
primary_location.is_published	False
primary_location.raw_source_name
primary_location.landing_page_url	http://arxiv.org/abs/2505.21666
publication_date	2025-05-27
publication_year	2025
referenced_works_count	0
abstract_inverted_index.a	24, 46, 65, 69, 73, 93, 128
abstract_inverted_index.--	80
abstract_inverted_index.It	20
abstract_inverted_index.KL	139
abstract_inverted_index.RL	124
abstract_inverted_index.We	60, 77
abstract_inverted_index.as	64, 172
abstract_inverted_index.at	194
abstract_inverted_index.in	176
abstract_inverted_index.is	21, 45, 114, 192
abstract_inverted_index.it	50
abstract_inverted_index.of	3, 26, 41, 67, 100, 151
abstract_inverted_index.or	125
abstract_inverted_index.to	7, 11, 96, 105, 130, 147
abstract_inverted_index.we	135, 156
abstract_inverted_index.Our	190
abstract_inverted_index.The	0
abstract_inverted_index.Via	127
abstract_inverted_index.and	33, 91, 116, 183
abstract_inverted_index.any	120
abstract_inverted_index.but	49
abstract_inverted_index.can	51, 161
abstract_inverted_index.for	23
abstract_inverted_index.key	111
abstract_inverted_index.not	118
abstract_inverted_index.the	9, 42, 53, 98, 101, 106, 141, 148, 152, 168, 173
abstract_inverted_index.(RL)	38
abstract_inverted_index.SLCD	79, 144, 160
abstract_inverted_index.aims	6
abstract_inverted_index.base	43, 174
abstract_inverted_index.both	177
abstract_inverted_index.code	191
abstract_inverted_index.data	90
abstract_inverted_index.does	117
abstract_inverted_index.from	123, 143
abstract_inverted_index.high	163
abstract_inverted_index.same	169
abstract_inverted_index.show	136
abstract_inverted_index.some	16
abstract_inverted_index.that	14, 71, 137, 159
abstract_inverted_index.time	171
abstract_inverted_index.with	166, 180, 187
abstract_inverted_index.based	39, 83
abstract_inverted_index.frame	61
abstract_inverted_index.given	17
abstract_inverted_index.guide	97
abstract_inverted_index.image	29, 178
abstract_inverted_index.model	10, 44, 175
abstract_inverted_index.small	94
abstract_inverted_index.steer	8
abstract_inverted_index.under	138
abstract_inverted_index.which	86
abstract_inverted_index.while	56
abstract_inverted_index.SLCD's	110
abstract_inverted_index.model.	103
abstract_inverted_index.models	5
abstract_inverted_index.nearly	167
abstract_inverted_index.online	89, 132
abstract_inverted_index.output	142
abstract_inverted_index.reward	54
abstract_inverted_index.trains	92
abstract_inverted_index.Similar	104
abstract_inverted_index.complex	121
abstract_inverted_index.finding	68
abstract_inverted_index.involve	119
abstract_inverted_index.optimal	149
abstract_inverted_index.overfit	52
abstract_inverted_index.popular	47
abstract_inverted_index.present	78
abstract_inverted_index.problem	66
abstract_inverted_index.quality	164
abstract_inverted_index.samples	13, 165
abstract_inverted_index.variety	25
abstract_inverted_index.Further,	155
abstract_inverted_index.Learning	37, 82
abstract_inverted_index.approach	48
abstract_inverted_index.concepts	122
abstract_inverted_index.control.	126
abstract_inverted_index.discrete	188
abstract_inverted_index.function	55
abstract_inverted_index.generate	12, 162
abstract_inverted_index.learning	133
abstract_inverted_index.molecule	31
abstract_inverted_index.optimize	15
abstract_inverted_index.provably	145
abstract_inverted_index.sequence	185
abstract_inverted_index.solution	150
abstract_inverted_index.standard	107
abstract_inverted_index.analysis,	134
abstract_inverted_index.available	193
abstract_inverted_index.converges	146
abstract_inverted_index.desirable	22
abstract_inverted_index.diffusion	4, 102, 182
abstract_inverted_index.function.	76
abstract_inverted_index.generates	88
abstract_inverted_index.including	28
abstract_inverted_index.inference	170
abstract_inverted_index.no-regret	131
abstract_inverted_index.objective	18, 75
abstract_inverted_index.optimizes	72
abstract_inverted_index.primitive	113
abstract_inverted_index.reduction	129
abstract_inverted_index.requiring	57
abstract_inverted_index.Diffusion,	85
abstract_inverted_index.Supervised	81
abstract_inverted_index.biological	184
abstract_inverted_index.classifier	95
abstract_inverted_index.continuous	181
abstract_inverted_index.diffusion,	109
abstract_inverted_index.diffusion.	189
abstract_inverted_index.functions.	19
abstract_inverted_index.generation	2, 63, 99, 179, 186
abstract_inverted_index.objective.	154
abstract_inverted_index.resources.	59
abstract_inverted_index.computation	112
abstract_inverted_index.demonstrate	158
abstract_inverted_index.divergence,	140
abstract_inverted_index.empirically	157
abstract_inverted_index.fine-tuning	40
abstract_inverted_index.generation,	30, 32
abstract_inverted_index.generation.	35
abstract_inverted_index.iteratively	87
abstract_inverted_index.significant	58
abstract_inverted_index.Controllable	84
abstract_inverted_index.DNA/sequence	34
abstract_inverted_index.applications	27
abstract_inverted_index.controllable	1, 62
abstract_inverted_index.distribution	70
abstract_inverted_index.Reinforcement	36
abstract_inverted_index.KL-regularized	74, 153
abstract_inverted_index.classification	115
abstract_inverted_index.classifier-guided	108
abstract_inverted_index.https://github.com/Owen-Oertell/slcd	195
cited_by_percentile_year
countries_distinct_count	0
institutions_distinct_count	6
citation_normalized_percentile