Concept-based Analysis of Neural Networks via Vision-Language Models Article Swipe

PDF

Ravi Mangal , Nina Narodytska , Divya Gopinath , Boyue Caroline Hu , Anirban Roy , Susmit Jha , Corina S. Păsăreanu ·

YOU? · · 2024 · Open Access · · DOI: https://doi.org/10.48550/arxiv.2403.19837

The analysis of vision-based deep neural networks (DNNs) is highly desirable but it is very challenging due to the difficulty of expressing formal specifications for vision tasks and the lack of efficient verification procedures. In this paper, we propose to leverage emerging multimodal, vision-language, foundation models (VLMs) as a lens through which we can reason about vision models. VLMs have been trained on a large body of images accompanied by their textual description, and are thus implicitly aware of high-level, human-understandable concepts describing the images. We describe a logical specification language $\texttt{Con}_{\texttt{spec}}$ designed to facilitate writing specifications in terms of these concepts. To define and formally check $\texttt{Con}_{\texttt{spec}}$ specifications, we build a map between the internal representations of a given vision model and a VLM, leading to an efficient verification procedure of natural-language properties for vision models. We demonstrate our techniques on a ResNet-based classifier trained on the RIVAL-10 dataset using CLIP as the multimodal model.

Related Topics

Computer Science

Artificial Intelligence

Concepts

Computer science Artificial neural network Artificial intelligence Natural language processing

Metadata

Type: preprint
Language: en
Landing Page: http://arxiv.org/abs/2403.19837
PDF: https://arxiv.org/pdf/2403.19837
OA Status: green
Related Works: 10
OpenAlex ID: https://openalex.org/W4393399151

All OpenAlex metadata

Raw OpenAlex JSON

OpenAlex ID: https://openalex.org/W4393399151

Canonical identifier for this work in OpenAlex
DOI: https://doi.org/10.48550/arxiv.2403.19837

Digital Object Identifier
Title: Concept-based Analysis of Neural Networks via Vision-Language Models

Work title
Type: preprint

OpenAlex work type
Language: en

Primary language
Publication year: 2024

Year of publication
Publication date: 2024-03-28

Full publication date if available
Authors: Ravi Mangal, Nina Narodytska, Divya Gopinath, Boyue Caroline Hu, Anirban Roy, Susmit Jha, Corina S. Păsăreanu

List of authors in order
Landing page: https://arxiv.org/abs/2403.19837

Publisher landing page
PDF URL: https://arxiv.org/pdf/2403.19837

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/2403.19837

Direct OA link when available
Concepts: Computer science, Artificial neural network, Artificial intelligence, Natural language processing

Top concepts (fields/topics) attached by OpenAlex
Cited by: 0

Total citation count in OpenAlex
Related works (count): 10

Other works algorithmically related by OpenAlex

Full payload

id	https://openalex.org/W4393399151
doi	https://doi.org/10.48550/arxiv.2403.19837
ids.doi	https://doi.org/10.48550/arxiv.2403.19837
ids.openalex	https://openalex.org/W4393399151
fwci
type	preprint
title	Concept-based Analysis of Neural Networks via Vision-Language Models
biblio.issue
biblio.volume
biblio.last_page
biblio.first_page
topics[0].id	https://openalex.org/T10215
topics[0].field.id	https://openalex.org/fields/17
topics[0].field.display_name	Computer Science
topics[0].score	0.710099995136261
topics[0].domain.id	https://openalex.org/domains/3
topics[0].domain.display_name	Physical Sciences
topics[0].subfield.id	https://openalex.org/subfields/1702
topics[0].subfield.display_name	Artificial Intelligence
topics[0].display_name	Semantic Web and Ontologies
is_xpac	False
apc_list
apc_paid
concepts[0].id	https://openalex.org/C41008148
concepts[0].level	0
concepts[0].score	0.5981757640838623
concepts[0].wikidata	https://www.wikidata.org/wiki/Q21198
concepts[0].display_name	Computer science
concepts[1].id	https://openalex.org/C50644808
concepts[1].level	2
concepts[1].score	0.5425792932510376
concepts[1].wikidata	https://www.wikidata.org/wiki/Q192776
concepts[1].display_name	Artificial neural network
concepts[2].id	https://openalex.org/C154945302
concepts[2].level	1
concepts[2].score	0.4686683118343353
concepts[2].wikidata	https://www.wikidata.org/wiki/Q11660
concepts[2].display_name	Artificial intelligence
concepts[3].id	https://openalex.org/C204321447
concepts[3].level	1
concepts[3].score	0.37652909755706787
concepts[3].wikidata	https://www.wikidata.org/wiki/Q30642
concepts[3].display_name	Natural language processing
keywords[0].id	https://openalex.org/keywords/computer-science
keywords[0].score	0.5981757640838623
keywords[0].display_name	Computer science
keywords[1].id	https://openalex.org/keywords/artificial-neural-network
keywords[1].score	0.5425792932510376
keywords[1].display_name	Artificial neural network
keywords[2].id	https://openalex.org/keywords/artificial-intelligence
keywords[2].score	0.4686683118343353
keywords[2].display_name	Artificial intelligence
keywords[3].id	https://openalex.org/keywords/natural-language-processing
keywords[3].score	0.37652909755706787
keywords[3].display_name	Natural language processing
language	en
locations[0].id	pmh:oai:arXiv.org:2403.19837
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/2403.19837
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/2403.19837
locations[1].id	doi:10.48550/arxiv.2403.19837
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.2403.19837
indexed_in	arxiv, datacite
authorships[0].author.id	https://openalex.org/A5002052018
authorships[0].author.orcid	https://orcid.org/0000-0001-6267-6995
authorships[0].author.display_name	Ravi Mangal
authorships[0].author_position	first
authorships[0].raw_author_name	Mangal, Ravi
authorships[0].is_corresponding	False
authorships[1].author.id	https://openalex.org/A5034897040
authorships[1].author.orcid	https://orcid.org/0000-0002-5181-4560
authorships[1].author.display_name	Nina Narodytska
authorships[1].author_position	middle
authorships[1].raw_author_name	Narodytska, Nina
authorships[1].is_corresponding	False
authorships[2].author.id	https://openalex.org/A5055334891
authorships[2].author.orcid	https://orcid.org/0000-0002-4279-7420
authorships[2].author.display_name	Divya Gopinath
authorships[2].author_position	middle
authorships[2].raw_author_name	Gopinath, Divya
authorships[2].is_corresponding	False
authorships[3].author.id	https://openalex.org/A5028103264
authorships[3].author.orcid	https://orcid.org/0000-0002-7090-6276
authorships[3].author.display_name	Boyue Caroline Hu
authorships[3].author_position	middle
authorships[3].raw_author_name	Hu, Boyue Caroline
authorships[3].is_corresponding	False
authorships[4].author.id	https://openalex.org/A5101531634
authorships[4].author.orcid	https://orcid.org/0000-0002-0702-4553
authorships[4].author.display_name	Anirban Roy
authorships[4].author_position	middle
authorships[4].raw_author_name	Roy, Anirban
authorships[4].is_corresponding	False
authorships[5].author.id	https://openalex.org/A5035902535
authorships[5].author.orcid	https://orcid.org/0000-0001-5983-9095
authorships[5].author.display_name	Susmit Jha
authorships[5].author_position	middle
authorships[5].raw_author_name	Jha, Susmit
authorships[5].is_corresponding	False
authorships[6].author.id	https://openalex.org/A5053134485
authorships[6].author.orcid	https://orcid.org/0000-0002-5579-6961
authorships[6].author.display_name	Corina S. Păsăreanu
authorships[6].author_position	last
authorships[6].raw_author_name	Pasareanu, Corina
authorships[6].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/2403.19837
open_access.oa_status	green
open_access.any_repository_has_fulltext	False
created_date	2025-10-10T00:00:00
display_name	Concept-based Analysis of Neural Networks via Vision-Language Models
has_fulltext	False
is_retracted	False
updated_date	2025-11-06T06:51:31.235846
primary_topic.id	https://openalex.org/T10215
primary_topic.field.id	https://openalex.org/fields/17
primary_topic.field.display_name	Computer Science
primary_topic.score	0.710099995136261
primary_topic.domain.id	https://openalex.org/domains/3
primary_topic.domain.display_name	Physical Sciences
primary_topic.subfield.id	https://openalex.org/subfields/1702
primary_topic.subfield.display_name	Artificial Intelligence
primary_topic.display_name	Semantic Web and Ontologies
related_works	https://openalex.org/W2748952813, https://openalex.org/W2390279801, https://openalex.org/W2358668433, https://openalex.org/W2376932109, https://openalex.org/W2001405890, https://openalex.org/W2382290278, https://openalex.org/W2478288626, https://openalex.org/W4391913857, https://openalex.org/W2350741829, https://openalex.org/W3204019825
cited_by_count	0
locations_count	2
best_oa_location.id	pmh:oai:arXiv.org:2403.19837
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/2403.19837
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/2403.19837
primary_location.id	pmh:oai:arXiv.org:2403.19837
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/2403.19837
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/2403.19837
publication_date	2024-03-28
publication_year	2024
referenced_works_count	0
abstract_inverted_index.a	48, 63, 87, 111, 118, 123, 142
abstract_inverted_index.In	34
abstract_inverted_index.To	102
abstract_inverted_index.We	85, 137
abstract_inverted_index.an	127
abstract_inverted_index.as	47, 152
abstract_inverted_index.by	69
abstract_inverted_index.in	97
abstract_inverted_index.is	8, 13
abstract_inverted_index.it	12
abstract_inverted_index.of	2, 20, 30, 66, 78, 99, 117, 131
abstract_inverted_index.on	62, 141, 146
abstract_inverted_index.to	17, 39, 93, 126
abstract_inverted_index.we	37, 52, 109
abstract_inverted_index.The	0
abstract_inverted_index.and	27, 73, 104, 122
abstract_inverted_index.are	74
abstract_inverted_index.but	11
abstract_inverted_index.can	53
abstract_inverted_index.due	16
abstract_inverted_index.for	24, 134
abstract_inverted_index.map	112
abstract_inverted_index.our	139
abstract_inverted_index.the	18, 28, 83, 114, 147, 153
abstract_inverted_index.CLIP	151
abstract_inverted_index.VLM,	124
abstract_inverted_index.VLMs	58
abstract_inverted_index.been	60
abstract_inverted_index.body	65
abstract_inverted_index.deep	4
abstract_inverted_index.have	59
abstract_inverted_index.lack	29
abstract_inverted_index.lens	49
abstract_inverted_index.this	35
abstract_inverted_index.thus	75
abstract_inverted_index.very	14
abstract_inverted_index.about	55
abstract_inverted_index.aware	77
abstract_inverted_index.build	110
abstract_inverted_index.check	106
abstract_inverted_index.given	119
abstract_inverted_index.large	64
abstract_inverted_index.model	121
abstract_inverted_index.tasks	26
abstract_inverted_index.terms	98
abstract_inverted_index.their	70
abstract_inverted_index.these	100
abstract_inverted_index.using	150
abstract_inverted_index.which	51
abstract_inverted_index.(DNNs)	7
abstract_inverted_index.(VLMs)	46
abstract_inverted_index.define	103
abstract_inverted_index.formal	22
abstract_inverted_index.highly	9
abstract_inverted_index.images	67
abstract_inverted_index.model.	155
abstract_inverted_index.models	45
abstract_inverted_index.neural	5
abstract_inverted_index.paper,	36
abstract_inverted_index.reason	54
abstract_inverted_index.vision	25, 56, 120, 135
abstract_inverted_index.between	113
abstract_inverted_index.dataset	149
abstract_inverted_index.images.	84
abstract_inverted_index.leading	125
abstract_inverted_index.logical	88
abstract_inverted_index.models.	57, 136
abstract_inverted_index.propose	38
abstract_inverted_index.textual	71
abstract_inverted_index.through	50
abstract_inverted_index.trained	61, 145
abstract_inverted_index.writing	95
abstract_inverted_index.RIVAL-10	148
abstract_inverted_index.analysis	1
abstract_inverted_index.concepts	81
abstract_inverted_index.describe	86
abstract_inverted_index.designed	92
abstract_inverted_index.emerging	41
abstract_inverted_index.formally	105
abstract_inverted_index.internal	115
abstract_inverted_index.language	90
abstract_inverted_index.leverage	40
abstract_inverted_index.networks	6
abstract_inverted_index.concepts.	101
abstract_inverted_index.desirable	10
abstract_inverted_index.efficient	31, 128
abstract_inverted_index.procedure	130
abstract_inverted_index.classifier	144
abstract_inverted_index.describing	82
abstract_inverted_index.difficulty	19
abstract_inverted_index.expressing	21
abstract_inverted_index.facilitate	94
abstract_inverted_index.foundation	44
abstract_inverted_index.implicitly	76
abstract_inverted_index.multimodal	154
abstract_inverted_index.properties	133
abstract_inverted_index.techniques	140
abstract_inverted_index.accompanied	68
abstract_inverted_index.challenging	15
abstract_inverted_index.demonstrate	138
abstract_inverted_index.high-level,	79
abstract_inverted_index.multimodal,	42
abstract_inverted_index.procedures.	33
abstract_inverted_index.ResNet-based	143
abstract_inverted_index.description,	72
abstract_inverted_index.verification	32, 129
abstract_inverted_index.vision-based	3
abstract_inverted_index.specification	89
abstract_inverted_index.specifications	23, 96
abstract_inverted_index.representations	116
abstract_inverted_index.specifications,	108
abstract_inverted_index.natural-language	132
abstract_inverted_index.vision-language,	43
abstract_inverted_index.human-understandable	80
abstract_inverted_index.$\texttt{Con}_{\texttt{spec}}$	91, 107
cited_by_percentile_year
countries_distinct_count	0
institutions_distinct_count	7
citation_normalized_percentile