TraceDet: Hallucination Detection from the Decoding Trace of Diffusion Large Language Models Article Swipe

PDF

S. Chang , Junchi Yu , Weixing Wang , Yongqiang Chen , Jialin Yu , Philip H. S. Torr , Jindong Gu ·

YOU? · · 2025 · Open Access · · DOI: https://doi.org/10.48550/arxiv.2510.01274

Diffusion large language models (D-LLMs) have recently emerged as a promising alternative to auto-regressive LLMs (AR-LLMs). However, the hallucination problem in D-LLMs remains underexplored, limiting their reliability in real-world applications. Existing hallucination detection methods are designed for AR-LLMs and rely on signals from single-step generation, making them ill-suited for D-LLMs where hallucination signals often emerge throughout the multi-step denoising process. To bridge this gap, we propose TraceDet, a novel framework that explicitly leverages the intermediate denoising steps of D-LLMs for hallucination detection. TraceDet models the denoising process as an action trace, with each action defined as the model's prediction over the cleaned response, conditioned on the previous intermediate output. By identifying the sub-trace that is maximally informative to the hallucinated responses, TraceDet leverages the key hallucination signals in the multi-step denoising process of D-LLMs for hallucination detection. Extensive experiments on various open source D-LLMs demonstrate that TraceDet consistently improves hallucination detection, achieving an average gain in AUROC of 15.2% compared to baselines.

Related Topics

Concepts

No concepts available.

Metadata

Type: preprint
Language: en
Landing Page: http://arxiv.org/abs/2510.01274
PDF: https://arxiv.org/pdf/2510.01274
OA Status: green
OpenAlex ID: https://openalex.org/W4414814664

All OpenAlex metadata

Raw OpenAlex JSON

OpenAlex ID: https://openalex.org/W4414814664

Canonical identifier for this work in OpenAlex
DOI: https://doi.org/10.48550/arxiv.2510.01274

Digital Object Identifier
Title: TraceDet: Hallucination Detection from the Decoding Trace of Diffusion Large Language Models

Work title
Type: preprint

OpenAlex work type
Language: en

Primary language
Publication year: 2025

Year of publication
Publication date: 2025-09-30

Full publication date if available
Authors: S. Chang, Junchi Yu, Weixing Wang, Yongqiang Chen, Jialin Yu, Philip H. S. Torr, Jindong Gu

List of authors in order
Landing page: https://arxiv.org/abs/2510.01274

Publisher landing page
PDF URL: https://arxiv.org/pdf/2510.01274

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/2510.01274

Direct OA link when available
Cited by: 0

Total citation count in OpenAlex

Full payload

id	https://openalex.org/W4414814664
doi	https://doi.org/10.48550/arxiv.2510.01274
ids.doi	https://doi.org/10.48550/arxiv.2510.01274
ids.openalex	https://openalex.org/W4414814664
fwci
type	preprint
title	TraceDet: Hallucination Detection from the Decoding Trace of Diffusion Large Language Models
biblio.issue
biblio.volume
biblio.last_page
biblio.first_page
topics[0].id	https://openalex.org/T13702
topics[0].field.id	https://openalex.org/fields/17
topics[0].field.display_name	Computer Science
topics[0].score	0.9099000096321106
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	Machine Learning in Healthcare
is_xpac	False
apc_list
apc_paid
language	en
locations[0].id	pmh:oai:arXiv.org:2510.01274
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/2510.01274
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/2510.01274
locations[1].id	doi:10.48550/arxiv.2510.01274
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.2510.01274
indexed_in	arxiv, datacite
authorships[0].author.id	https://openalex.org/A5113721466
authorships[0].author.orcid
authorships[0].author.display_name	S. Chang
authorships[0].author_position	first
authorships[0].raw_author_name	Chang, Shenxu
authorships[0].is_corresponding	False
authorships[1].author.id	https://openalex.org/A5060684352
authorships[1].author.orcid	https://orcid.org/0000-0003-4118-3248
authorships[1].author.display_name	Junchi Yu
authorships[1].author_position	middle
authorships[1].raw_author_name	Yu, Junchi
authorships[1].is_corresponding	False
authorships[2].author.id	https://openalex.org/A5100739970
authorships[2].author.orcid	https://orcid.org/0000-0002-9512-3307
authorships[2].author.display_name	Weixing Wang
authorships[2].author_position	middle
authorships[2].raw_author_name	Wang, Weixing
authorships[2].is_corresponding	False
authorships[3].author.id	https://openalex.org/A5100611593
authorships[3].author.orcid	https://orcid.org/0000-0003-2309-5674
authorships[3].author.display_name	Yongqiang Chen
authorships[3].author_position	middle
authorships[3].raw_author_name	Chen, Yongqiang
authorships[3].is_corresponding	False
authorships[4].author.id	https://openalex.org/A5078597351
authorships[4].author.orcid	https://orcid.org/0000-0001-7561-5008
authorships[4].author.display_name	Jialin Yu
authorships[4].author_position	middle
authorships[4].raw_author_name	Yu, Jialin
authorships[4].is_corresponding	False
authorships[5].author.id	https://openalex.org/A5042899882
authorships[5].author.orcid
authorships[5].author.display_name	Philip H. S. Torr
authorships[5].author_position	middle
authorships[5].raw_author_name	Torr, Philip
authorships[5].is_corresponding	False
authorships[6].author.id	https://openalex.org/A5055994909
authorships[6].author.orcid	https://orcid.org/0009-0000-0574-0129
authorships[6].author.display_name	Jindong Gu
authorships[6].author_position	last
authorships[6].raw_author_name	Gu, Jindong
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/2510.01274
open_access.oa_status	green
open_access.any_repository_has_fulltext	False
created_date	2025-10-10T00:00:00
display_name	TraceDet: Hallucination Detection from the Decoding Trace of Diffusion Large Language Models
has_fulltext	False
is_retracted	False
updated_date	2025-11-06T06:51:31.235846
primary_topic.id	https://openalex.org/T13702
primary_topic.field.id	https://openalex.org/fields/17
primary_topic.field.display_name	Computer Science
primary_topic.score	0.9099000096321106
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	Machine Learning in Healthcare
cited_by_count	0
locations_count	2
best_oa_location.id	pmh:oai:arXiv.org:2510.01274
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/2510.01274
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/2510.01274
primary_location.id	pmh:oai:arXiv.org:2510.01274
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/2510.01274
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/2510.01274
publication_date	2025-09-30
publication_year	2025
referenced_works_count	0
abstract_inverted_index.a	9, 67
abstract_inverted_index.By	109
abstract_inverted_index.To	60
abstract_inverted_index.an	88, 152
abstract_inverted_index.as	8, 87, 95
abstract_inverted_index.in	20, 27, 127, 155
abstract_inverted_index.is	114
abstract_inverted_index.of	77, 132, 157
abstract_inverted_index.on	40, 104, 139
abstract_inverted_index.to	12, 117, 160
abstract_inverted_index.we	64
abstract_inverted_index.and	38
abstract_inverted_index.are	34
abstract_inverted_index.for	36, 48, 79, 134
abstract_inverted_index.key	124
abstract_inverted_index.the	17, 56, 73, 84, 96, 100, 105, 111, 118, 123, 128
abstract_inverted_index.LLMs	14
abstract_inverted_index.each	92
abstract_inverted_index.from	42
abstract_inverted_index.gain	154
abstract_inverted_index.gap,	63
abstract_inverted_index.have	5
abstract_inverted_index.open	141
abstract_inverted_index.over	99
abstract_inverted_index.rely	39
abstract_inverted_index.that	70, 113, 145
abstract_inverted_index.them	46
abstract_inverted_index.this	62
abstract_inverted_index.with	91
abstract_inverted_index.15.2%	158
abstract_inverted_index.AUROC	156
abstract_inverted_index.large	1
abstract_inverted_index.novel	68
abstract_inverted_index.often	53
abstract_inverted_index.steps	76
abstract_inverted_index.their	25
abstract_inverted_index.where	50
abstract_inverted_index.D-LLMs	21, 49, 78, 133, 143
abstract_inverted_index.action	89, 93
abstract_inverted_index.bridge	61
abstract_inverted_index.emerge	54
abstract_inverted_index.making	45
abstract_inverted_index.models	3, 83
abstract_inverted_index.source	142
abstract_inverted_index.trace,	90
abstract_inverted_index.AR-LLMs	37
abstract_inverted_index.average	153
abstract_inverted_index.cleaned	101
abstract_inverted_index.defined	94
abstract_inverted_index.emerged	7
abstract_inverted_index.methods	33
abstract_inverted_index.model's	97
abstract_inverted_index.output.	108
abstract_inverted_index.problem	19
abstract_inverted_index.process	86, 131
abstract_inverted_index.propose	65
abstract_inverted_index.remains	22
abstract_inverted_index.signals	41, 52, 126
abstract_inverted_index.various	140
abstract_inverted_index.(D-LLMs)	4
abstract_inverted_index.Existing	30
abstract_inverted_index.However,	16
abstract_inverted_index.TraceDet	82, 121, 146
abstract_inverted_index.compared	159
abstract_inverted_index.designed	35
abstract_inverted_index.improves	148
abstract_inverted_index.language	2
abstract_inverted_index.limiting	24
abstract_inverted_index.previous	106
abstract_inverted_index.process.	59
abstract_inverted_index.recently	6
abstract_inverted_index.Diffusion	0
abstract_inverted_index.Extensive	137
abstract_inverted_index.TraceDet,	66
abstract_inverted_index.achieving	151
abstract_inverted_index.denoising	58, 75, 85, 130
abstract_inverted_index.detection	32
abstract_inverted_index.framework	69
abstract_inverted_index.leverages	72, 122
abstract_inverted_index.maximally	115
abstract_inverted_index.promising	10
abstract_inverted_index.response,	102
abstract_inverted_index.sub-trace	112
abstract_inverted_index.(AR-LLMs).	15
abstract_inverted_index.baselines.	161
abstract_inverted_index.detection,	150
abstract_inverted_index.detection.	81, 136
abstract_inverted_index.explicitly	71
abstract_inverted_index.ill-suited	47
abstract_inverted_index.multi-step	57, 129
abstract_inverted_index.prediction	98
abstract_inverted_index.real-world	28
abstract_inverted_index.responses,	120
abstract_inverted_index.throughout	55
abstract_inverted_index.alternative	11
abstract_inverted_index.conditioned	103
abstract_inverted_index.demonstrate	144
abstract_inverted_index.experiments	138
abstract_inverted_index.generation,	44
abstract_inverted_index.identifying	110
abstract_inverted_index.informative	116
abstract_inverted_index.reliability	26
abstract_inverted_index.single-step	43
abstract_inverted_index.consistently	147
abstract_inverted_index.hallucinated	119
abstract_inverted_index.intermediate	74, 107
abstract_inverted_index.applications.	29
abstract_inverted_index.hallucination	18, 31, 51, 80, 125, 135, 149
abstract_inverted_index.underexplored,	23
abstract_inverted_index.auto-regressive	13
cited_by_percentile_year
countries_distinct_count	0
institutions_distinct_count	7
citation_normalized_percentile