TeleLoRA: Teleporting Model-Specific Alignment Across LLMs Article Swipe

PDF

Xiao Lin , Manoj Acharya , Anirban Roy , Susmit Jha ·

YOU? · · 2025 · Open Access · · DOI: https://doi.org/10.48550/arxiv.2503.20228

Mitigating Trojans in Large Language Models (LLMs) is one of many tasks where alignment data is LLM specific, as different LLMs have different Trojan triggers and trigger behaviors to be removed. In this paper, we introduce TeleLoRA (Teleporting Low-Rank Adaptation), a novel framework that synergizes model-specific alignment data across multiple LLMs to enable zero-shot Trojan mitigation on unseen LLMs without alignment data. TeleLoRA learns a unified generator of LoRA adapter weights by leveraging local activation information across multiple LLMs. This generator is designed to be permutation symmetric to generalize across models with different architectures and sizes. We optimize the model design for memory efficiency, making it feasible to learn with large-scale LLMs with minimal computational resources. Experiments on LLM Trojan mitigation benchmarks demonstrate that TeleLoRA effectively reduces attack success rates while preserving the benign performance of the models.

Related Topics

Concepts

Political science

Metadata

Type: preprint
Language: en
Landing Page: http://arxiv.org/abs/2503.20228
PDF: https://arxiv.org/pdf/2503.20228
OA Status: green
Related Works: 10
OpenAlex ID: https://openalex.org/W4409051414

All OpenAlex metadata

Raw OpenAlex JSON

OpenAlex ID: https://openalex.org/W4409051414

Canonical identifier for this work in OpenAlex
DOI: https://doi.org/10.48550/arxiv.2503.20228

Digital Object Identifier
Title: TeleLoRA: Teleporting Model-Specific Alignment Across LLMs

Work title
Type: preprint

OpenAlex work type
Language: en

Primary language
Publication year: 2025

Year of publication
Publication date: 2025-03-26

Full publication date if available
Authors: Xiao Lin, Manoj Acharya, Anirban Roy, Susmit Jha

List of authors in order
Landing page: https://arxiv.org/abs/2503.20228

Publisher landing page
PDF URL: https://arxiv.org/pdf/2503.20228

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/2503.20228

Direct OA link when available
Concepts: Political science

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/W4409051414
doi	https://doi.org/10.48550/arxiv.2503.20228
ids.doi	https://doi.org/10.48550/arxiv.2503.20228
ids.openalex	https://openalex.org/W4409051414
fwci
type	preprint
title	TeleLoRA: Teleporting Model-Specific Alignment Across LLMs
biblio.issue
biblio.volume
biblio.last_page
biblio.first_page
topics[0].id	https://openalex.org/T10703
topics[0].field.id	https://openalex.org/fields/14
topics[0].field.display_name	Business, Management and Accounting
topics[0].score	0.98089998960495
topics[0].domain.id	https://openalex.org/domains/2
topics[0].domain.display_name	Social Sciences
topics[0].subfield.id	https://openalex.org/subfields/1404
topics[0].subfield.display_name	Management Information Systems
topics[0].display_name	Business Process Modeling and Analysis
topics[1].id	https://openalex.org/T10215
topics[1].field.id	https://openalex.org/fields/17
topics[1].field.display_name	Computer Science
topics[1].score	0.9722999930381775
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	Semantic Web and Ontologies
topics[2].id	https://openalex.org/T10679
topics[2].field.id	https://openalex.org/fields/17
topics[2].field.display_name	Computer Science
topics[2].score	0.9485999941825867
topics[2].domain.id	https://openalex.org/domains/3
topics[2].domain.display_name	Physical Sciences
topics[2].subfield.id	https://openalex.org/subfields/1710
topics[2].subfield.display_name	Information Systems
topics[2].display_name	Service-Oriented Architecture and Web Services
is_xpac	False
apc_list
apc_paid
concepts[0].id	https://openalex.org/C17744445
concepts[0].level	0
concepts[0].score	0.32051074504852295
concepts[0].wikidata	https://www.wikidata.org/wiki/Q36442
concepts[0].display_name	Political science
keywords[0].id	https://openalex.org/keywords/political-science
keywords[0].score	0.32051074504852295
keywords[0].display_name	Political science
language	en
locations[0].id	pmh:oai:arXiv.org:2503.20228
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/2503.20228
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/2503.20228
locations[1].id	doi:10.48550/arxiv.2503.20228
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
locations[1].pdf_url
locations[1].version
locations[1].raw_type	article
locations[1].license_id
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.2503.20228
indexed_in	arxiv, datacite
authorships[0].author.id	https://openalex.org/A5116868324
authorships[0].author.orcid
authorships[0].author.display_name	Xiao Lin
authorships[0].author_position	first
authorships[0].raw_author_name	Lin, Xiao
authorships[0].is_corresponding	False
authorships[1].author.id	https://openalex.org/A5116868325
authorships[1].author.orcid
authorships[1].author.display_name	Manoj Acharya
authorships[1].author_position	middle
authorships[1].raw_author_name	Acharya, Manoj
authorships[1].is_corresponding	False
authorships[2].author.id	https://openalex.org/A5101531634
authorships[2].author.orcid	https://orcid.org/0000-0002-0702-4553
authorships[2].author.display_name	Anirban Roy
authorships[2].author_position	middle
authorships[2].raw_author_name	Roy, Anirban
authorships[2].is_corresponding	False
authorships[3].author.id	https://openalex.org/A5035902535
authorships[3].author.orcid	https://orcid.org/0000-0001-5983-9095
authorships[3].author.display_name	Susmit Jha
authorships[3].author_position	last
authorships[3].raw_author_name	Jha, Susmit
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/2503.20228
open_access.oa_status	green
open_access.any_repository_has_fulltext	False
created_date	2025-10-10T00:00:00
display_name	TeleLoRA: Teleporting Model-Specific Alignment Across LLMs
has_fulltext	False
is_retracted	False
updated_date	2025-11-06T06:51:31.235846
primary_topic.id	https://openalex.org/T10703
primary_topic.field.id	https://openalex.org/fields/14
primary_topic.field.display_name	Business, Management and Accounting
primary_topic.score	0.98089998960495
primary_topic.domain.id	https://openalex.org/domains/2
primary_topic.domain.display_name	Social Sciences
primary_topic.subfield.id	https://openalex.org/subfields/1404
primary_topic.subfield.display_name	Management Information Systems
primary_topic.display_name	Business Process Modeling and Analysis
related_works	https://openalex.org/W4391375266, https://openalex.org/W2899084033, https://openalex.org/W2748952813, https://openalex.org/W2949263084, https://openalex.org/W2743539335, https://openalex.org/W594353338, https://openalex.org/W2922049016, https://openalex.org/W4390697879, https://openalex.org/W2070214669, https://openalex.org/W2724734218
cited_by_count	0
locations_count	2
best_oa_location.id	pmh:oai:arXiv.org:2503.20228
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/2503.20228
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/2503.20228
primary_location.id	pmh:oai:arXiv.org:2503.20228
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/2503.20228
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/2503.20228
publication_date	2025-03-26
publication_year	2025
referenced_works_count	0
abstract_inverted_index.a	40, 64
abstract_inverted_index.In	31
abstract_inverted_index.We	96
abstract_inverted_index.as	18
abstract_inverted_index.be	29, 84
abstract_inverted_index.by	71
abstract_inverted_index.in	2
abstract_inverted_index.is	7, 15, 81
abstract_inverted_index.it	105
abstract_inverted_index.of	9, 67, 135
abstract_inverted_index.on	56, 117
abstract_inverted_index.to	28, 51, 83, 87, 107
abstract_inverted_index.we	34
abstract_inverted_index.LLM	16, 118
abstract_inverted_index.and	25, 94
abstract_inverted_index.for	101
abstract_inverted_index.one	8
abstract_inverted_index.the	98, 132, 136
abstract_inverted_index.LLMs	20, 50, 58, 111
abstract_inverted_index.LoRA	68
abstract_inverted_index.This	79
abstract_inverted_index.data	14, 47
abstract_inverted_index.have	21
abstract_inverted_index.many	10
abstract_inverted_index.that	43, 123
abstract_inverted_index.this	32
abstract_inverted_index.with	91, 109, 112
abstract_inverted_index.LLMs.	78
abstract_inverted_index.Large	3
abstract_inverted_index.data.	61
abstract_inverted_index.learn	108
abstract_inverted_index.local	73
abstract_inverted_index.model	99
abstract_inverted_index.novel	41
abstract_inverted_index.rates	129
abstract_inverted_index.tasks	11
abstract_inverted_index.where	12
abstract_inverted_index.while	130
abstract_inverted_index.(LLMs)	6
abstract_inverted_index.Models	5
abstract_inverted_index.Trojan	23, 54, 119
abstract_inverted_index.across	48, 76, 89
abstract_inverted_index.attack	127
abstract_inverted_index.benign	133
abstract_inverted_index.design	100
abstract_inverted_index.enable	52
abstract_inverted_index.learns	63
abstract_inverted_index.making	104
abstract_inverted_index.memory	102
abstract_inverted_index.models	90
abstract_inverted_index.paper,	33
abstract_inverted_index.sizes.	95
abstract_inverted_index.unseen	57
abstract_inverted_index.Trojans	1
abstract_inverted_index.adapter	69
abstract_inverted_index.minimal	113
abstract_inverted_index.models.	137
abstract_inverted_index.reduces	126
abstract_inverted_index.success	128
abstract_inverted_index.trigger	26
abstract_inverted_index.unified	65
abstract_inverted_index.weights	70
abstract_inverted_index.without	59
abstract_inverted_index.Language	4
abstract_inverted_index.Low-Rank	38
abstract_inverted_index.TeleLoRA	36, 62, 124
abstract_inverted_index.designed	82
abstract_inverted_index.feasible	106
abstract_inverted_index.multiple	49, 77
abstract_inverted_index.optimize	97
abstract_inverted_index.removed.	30
abstract_inverted_index.triggers	24
abstract_inverted_index.alignment	13, 46, 60
abstract_inverted_index.behaviors	27
abstract_inverted_index.different	19, 22, 92
abstract_inverted_index.framework	42
abstract_inverted_index.generator	66, 80
abstract_inverted_index.introduce	35
abstract_inverted_index.specific,	17
abstract_inverted_index.symmetric	86
abstract_inverted_index.zero-shot	53
abstract_inverted_index.Mitigating	0
abstract_inverted_index.activation	74
abstract_inverted_index.benchmarks	121
abstract_inverted_index.generalize	88
abstract_inverted_index.leveraging	72
abstract_inverted_index.mitigation	55, 120
abstract_inverted_index.preserving	131
abstract_inverted_index.resources.	115
abstract_inverted_index.synergizes	44
abstract_inverted_index.Experiments	116
abstract_inverted_index.demonstrate	122
abstract_inverted_index.effectively	125
abstract_inverted_index.efficiency,	103
abstract_inverted_index.information	75
abstract_inverted_index.large-scale	110
abstract_inverted_index.performance	134
abstract_inverted_index.permutation	85
abstract_inverted_index.(Teleporting	37
abstract_inverted_index.Adaptation),	39
abstract_inverted_index.architectures	93
abstract_inverted_index.computational	114
abstract_inverted_index.model-specific	45
cited_by_percentile_year
countries_distinct_count	0
institutions_distinct_count	4
citation_normalized_percentile