Enhancing Robustness of LLM-Driven Multi-Agent Systems through Randomized Smoothing Article Swipe

PDF

Jinwei Hu , Yi Dong , Zhengtao Ding , Xiaowei Huang ·

YOU? · · 2025 · Open Access ·

This paper presents a defense framework for enhancing the safety of large language model (LLM) empowered multi-agent systems (MAS) in safety-critical domains such as aerospace. We apply randomized smoothing, a statistical robustness certification technique, to the MAS consensus context, enabling probabilistic guarantees on agent decisions under adversarial influence. Unlike traditional verification methods, our approach operates in black-box settings and employs a two-stage adaptive sampling mechanism to balance robustness and computational efficiency. Simulation results demonstrate that our method effectively prevents the propagation of adversarial behaviors and hallucinations while maintaining consensus performance. This work provides a practical and scalable path toward safe deployment of LLM-based MAS in real-world, high-stakes environments.

Related Topics

Truth And Reconciliation Commission Of Canada

Alanis Morissette

2025 Nba Draft

28 Years Later

Concepts

No concepts available.

Metadata

Type: article
Language: en
Landing Page: http://arxiv.org/abs/2507.04105
PDF: https://arxiv.org/pdf/2507.04105
OA Status: green
OpenAlex ID: https://openalex.org/W4415345678

All OpenAlex metadata

Raw OpenAlex JSON

OpenAlex ID: https://openalex.org/W4415345678

Canonical identifier for this work in OpenAlex
Title: Enhancing Robustness of LLM-Driven Multi-Agent Systems through Randomized Smoothing

Work title
Type: article

OpenAlex work type
Language: en

Primary language
Publication year: 2025

Year of publication
Publication date: 2025-07-05

Full publication date if available
Authors: Jinwei Hu, Yi Dong, Zhengtao Ding, Xiaowei Huang

List of authors in order
Landing page: https://arxiv.org/abs/2507.04105

Publisher landing page
PDF URL: https://arxiv.org/pdf/2507.04105

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/2507.04105

Direct OA link when available
Cited by: 0

Total citation count in OpenAlex

Full payload

id	https://openalex.org/W4415345678
doi
ids.openalex	https://openalex.org/W4415345678
fwci	0.0
type	article
title	Enhancing Robustness of LLM-Driven Multi-Agent Systems through Randomized Smoothing
biblio.issue
biblio.volume
biblio.last_page
biblio.first_page
topics[0].id	https://openalex.org/T10462
topics[0].field.id	https://openalex.org/fields/17
topics[0].field.display_name	Computer Science
topics[0].score	0.9419000148773193
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	Reinforcement Learning in Robotics
topics[1].id	https://openalex.org/T10320
topics[1].field.id	https://openalex.org/fields/17
topics[1].field.display_name	Computer Science
topics[1].score	0.9071000218391418
topics[1].domain.id	https://openalex.org/domains/3
topics[1].domain.display_name	Physical Sciences
topics[1].subfield.id	https://openalex.org/subfields/1702
topics[1].subfield.display_name	Artificial Intelligence
topics[1].display_name	Neural Networks and Applications
is_xpac	False
apc_list
apc_paid
language	en
locations[0].id	pmh:oai:arXiv.org:2507.04105
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/2507.04105
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/2507.04105
indexed_in	arxiv
authorships[0].author.id	https://openalex.org/A5101242822
authorships[0].author.orcid
authorships[0].author.display_name	Jinwei Hu
authorships[0].author_position	first
authorships[0].raw_author_name	Hu, Jinwei
authorships[0].is_corresponding	False
authorships[1].author.id	https://openalex.org/A5100650289
authorships[1].author.orcid	https://orcid.org/0000-0003-3047-7777
authorships[1].author.display_name	Yi Dong
authorships[1].author_position	middle
authorships[1].raw_author_name	Dong, Yi
authorships[1].is_corresponding	False
authorships[2].author.id	https://openalex.org/A5086465750
authorships[2].author.orcid	https://orcid.org/0000-0003-0690-7853
authorships[2].author.display_name	Zhengtao Ding
authorships[2].author_position	middle
authorships[2].raw_author_name	Ding, Zhengtao
authorships[2].is_corresponding	False
authorships[3].author.id	https://openalex.org/A5083538605
authorships[3].author.orcid
authorships[3].author.display_name	Xiaowei Huang
authorships[3].author_position	last
authorships[3].raw_author_name	Huang, Xiaowei
authorships[3].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/2507.04105
open_access.oa_status	green
open_access.any_repository_has_fulltext	False
created_date	2025-10-20T00:00:00
display_name	Enhancing Robustness of LLM-Driven Multi-Agent Systems through Randomized Smoothing
has_fulltext	False
is_retracted	False
updated_date	2025-11-06T04:12:42.849631
primary_topic.id	https://openalex.org/T10462
primary_topic.field.id	https://openalex.org/fields/17
primary_topic.field.display_name	Computer Science
primary_topic.score	0.9419000148773193
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	Reinforcement Learning in Robotics
cited_by_count	0
locations_count	1
best_oa_location.id	pmh:oai:arXiv.org:2507.04105
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/2507.04105
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/2507.04105
primary_location.id	pmh:oai:arXiv.org:2507.04105
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/2507.04105
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/2507.04105
publication_date	2025-07-05
publication_year	2025
referenced_works_count	0
abstract_inverted_index.a	3, 29, 60, 93
abstract_inverted_index.We	25
abstract_inverted_index.as	23
abstract_inverted_index.in	19, 55, 104
abstract_inverted_index.of	10, 81, 101
abstract_inverted_index.on	42
abstract_inverted_index.to	34, 65
abstract_inverted_index.MAS	36, 103
abstract_inverted_index.and	58, 68, 84, 95
abstract_inverted_index.for	6
abstract_inverted_index.our	52, 75
abstract_inverted_index.the	8, 35, 79
abstract_inverted_index.This	0, 90
abstract_inverted_index.path	97
abstract_inverted_index.safe	99
abstract_inverted_index.such	22
abstract_inverted_index.that	74
abstract_inverted_index.work	91
abstract_inverted_index.(LLM)	14
abstract_inverted_index.(MAS)	18
abstract_inverted_index.agent	43
abstract_inverted_index.apply	26
abstract_inverted_index.large	11
abstract_inverted_index.model	13
abstract_inverted_index.paper	1
abstract_inverted_index.under	45
abstract_inverted_index.while	86
abstract_inverted_index.Unlike	48
abstract_inverted_index.method	76
abstract_inverted_index.safety	9
abstract_inverted_index.toward	98
abstract_inverted_index.balance	66
abstract_inverted_index.defense	4
abstract_inverted_index.domains	21
abstract_inverted_index.employs	59
abstract_inverted_index.results	72
abstract_inverted_index.systems	17
abstract_inverted_index.adaptive	62
abstract_inverted_index.approach	53
abstract_inverted_index.context,	38
abstract_inverted_index.enabling	39
abstract_inverted_index.language	12
abstract_inverted_index.methods,	51
abstract_inverted_index.operates	54
abstract_inverted_index.presents	2
abstract_inverted_index.prevents	78
abstract_inverted_index.provides	92
abstract_inverted_index.sampling	63
abstract_inverted_index.scalable	96
abstract_inverted_index.settings	57
abstract_inverted_index.LLM-based	102
abstract_inverted_index.behaviors	83
abstract_inverted_index.black-box	56
abstract_inverted_index.consensus	37, 88
abstract_inverted_index.decisions	44
abstract_inverted_index.empowered	15
abstract_inverted_index.enhancing	7
abstract_inverted_index.framework	5
abstract_inverted_index.mechanism	64
abstract_inverted_index.practical	94
abstract_inverted_index.two-stage	61
abstract_inverted_index.Simulation	71
abstract_inverted_index.aerospace.	24
abstract_inverted_index.deployment	100
abstract_inverted_index.guarantees	41
abstract_inverted_index.influence.	47
abstract_inverted_index.randomized	27
abstract_inverted_index.robustness	31, 67
abstract_inverted_index.smoothing,	28
abstract_inverted_index.technique,	33
abstract_inverted_index.adversarial	46, 82
abstract_inverted_index.demonstrate	73
abstract_inverted_index.effectively	77
abstract_inverted_index.efficiency.	70
abstract_inverted_index.high-stakes	106
abstract_inverted_index.maintaining	87
abstract_inverted_index.multi-agent	16
abstract_inverted_index.propagation	80
abstract_inverted_index.real-world,	105
abstract_inverted_index.statistical	30
abstract_inverted_index.traditional	49
abstract_inverted_index.performance.	89
abstract_inverted_index.verification	50
abstract_inverted_index.certification	32
abstract_inverted_index.computational	69
abstract_inverted_index.environments.	107
abstract_inverted_index.probabilistic	40
abstract_inverted_index.hallucinations	85
abstract_inverted_index.safety-critical	20
cited_by_percentile_year
countries_distinct_count	0
institutions_distinct_count	4
citation_normalized_percentile.value	0.22721802
citation_normalized_percentile.is_in_top_1_percent	False
citation_normalized_percentile.is_in_top_10_percent	True