From Failures to Fixes: LLM-Driven Scenario Repair for Self-Evolving Autonomous Driving Article Swipe

PDF

Xin Xia , Xingjun Ma , Dafang Zhuang , Ting Qu , Chen Hong , Xun Gong ·

YOU? · · 2025 · Open Access · · DOI: https://doi.org/10.48550/arxiv.2505.22067

Ensuring robust and generalizable autonomous driving requires not only broad scenario coverage but also efficient repair of failure cases, particularly those related to challenging and safety-critical scenarios. However, existing scenario generation and selection methods often lack adaptivity and semantic relevance, limiting their impact on performance improvement. In this paper, we propose \textbf{SERA}, an LLM-powered framework that enables autonomous driving systems to self-evolve by repairing failure cases through targeted scenario recommendation. By analyzing performance logs, SERA identifies failure patterns and dynamically retrieves semantically aligned scenarios from a structured bank. An LLM-based reflection mechanism further refines these recommendations to maximize relevance and diversity. The selected scenarios are used for few-shot fine-tuning, enabling targeted adaptation with minimal data. Experiments on the benchmark show that SERA consistently improves key metrics across multiple autonomous driving baselines, demonstrating its effectiveness and generalizability under safety-critical conditions.

Related Topics

Escape From New York

Tales From Earthsea (Film)

The Girl From Ipanema

From Here To Eternity

List Of United States Senators From New York

Death From Laughter

Transition From Ming To Qing

Postcards From The Edge (Film)

Formula One Drivers From The United States

Plan 9 From Bell Labs

Escape From Sobibor

We Got It From Here... Thank You 4 Your Service

From The Earth To The Moon (Miniseries)

From Mars To Sirius

Civilian Casualties From U.S. Drone Strikes

Concepts

No concepts available.

Metadata

Type: preprint
Language: en
Landing Page: http://arxiv.org/abs/2505.22067
PDF: https://arxiv.org/pdf/2505.22067
OA Status: green
OpenAlex ID: https://openalex.org/W4416047818

All OpenAlex metadata

Raw OpenAlex JSON

OpenAlex ID: https://openalex.org/W4416047818

Canonical identifier for this work in OpenAlex
DOI: https://doi.org/10.48550/arxiv.2505.22067

Digital Object Identifier
Title: From Failures to Fixes: LLM-Driven Scenario Repair for Self-Evolving Autonomous Driving

Work title
Type: preprint

OpenAlex work type
Language: en

Primary language
Publication year: 2025

Year of publication
Publication date: 2025-05-28

Full publication date if available
Authors: Xin Xia, Xingjun Ma, Dafang Zhuang, Ting Qu, Chen Hong, Xun Gong

List of authors in order
Landing page: https://arxiv.org/abs/2505.22067

Publisher landing page
PDF URL: https://arxiv.org/pdf/2505.22067

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/2505.22067

Direct OA link when available
Cited by: 0

Total citation count in OpenAlex

Full payload

id	https://openalex.org/W4416047818
doi	https://doi.org/10.48550/arxiv.2505.22067
ids.doi	https://doi.org/10.48550/arxiv.2505.22067
ids.openalex	https://openalex.org/W4416047818
fwci
type	preprint
title	From Failures to Fixes: LLM-Driven Scenario Repair for Self-Evolving Autonomous Driving
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:2505.22067
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/2505.22067
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/2505.22067
locations[1].id	doi:10.48550/arxiv.2505.22067
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.2505.22067
indexed_in	arxiv, datacite
authorships[0].author.id	https://openalex.org/A5067980386
authorships[0].author.orcid	https://orcid.org/0000-0002-7177-5028
authorships[0].author.display_name	Xin Xia
authorships[0].author_position	first
authorships[0].raw_author_name	Xia, Xinyu
authorships[0].is_corresponding	False
authorships[1].author.id	https://openalex.org/A5078711649
authorships[1].author.orcid	https://orcid.org/0000-0003-2099-4973
authorships[1].author.display_name	Xingjun Ma
authorships[1].author_position	middle
authorships[1].raw_author_name	Ma, Xingjun
authorships[1].is_corresponding	False
authorships[2].author.id	https://openalex.org/A5067926448
authorships[2].author.orcid	https://orcid.org/0000-0002-6219-6251
authorships[2].author.display_name	Dafang Zhuang
authorships[2].author_position	middle
authorships[2].raw_author_name	Hu, Yunfeng
authorships[2].is_corresponding	False
authorships[3].author.id	https://openalex.org/A5021504878
authorships[3].author.orcid	https://orcid.org/0000-0003-1012-2856
authorships[3].author.display_name	Ting Qu
authorships[3].author_position	middle
authorships[3].raw_author_name	Qu, Ting
authorships[3].is_corresponding	False
authorships[4].author.id	https://openalex.org/A5100785302
authorships[4].author.orcid	https://orcid.org/0000-0001-7540-3021
authorships[4].author.display_name	Chen Hong
authorships[4].author_position	middle
authorships[4].raw_author_name	Chen, Hong
authorships[4].is_corresponding	False
authorships[5].author.id	https://openalex.org/A5000673554
authorships[5].author.orcid	https://orcid.org/0000-0001-7111-736X
authorships[5].author.display_name	Xun Gong
authorships[5].author_position	last
authorships[5].raw_author_name	Gong, Xun
authorships[5].is_corresponding	False
has_content.pdf	True
has_content.grobid_xml	False
is_paratext	False
open_access.is_oa	True
open_access.oa_url	https://arxiv.org/pdf/2505.22067
open_access.oa_status	green
open_access.any_repository_has_fulltext	False
created_date	2025-10-10T00:00:00
display_name	From Failures to Fixes: LLM-Driven Scenario Repair for Self-Evolving Autonomous Driving
has_fulltext	False
is_retracted	False
updated_date	2025-11-28T11:00:00.990440
primary_topic
cited_by_count	0
locations_count	2
best_oa_location.id	pmh:oai:arXiv.org:2505.22067
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/2505.22067
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/2505.22067
primary_location.id	pmh:oai:arXiv.org:2505.22067
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/2505.22067
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/2505.22067
publication_date	2025-05-28
publication_year	2025
referenced_works_count	0
abstract_inverted_index.a	85
abstract_inverted_index.An	88
abstract_inverted_index.By	70
abstract_inverted_index.In	46
abstract_inverted_index.an	52
abstract_inverted_index.by	62
abstract_inverted_index.of	16
abstract_inverted_index.on	43, 116
abstract_inverted_index.to	22, 60, 96
abstract_inverted_index.we	49
abstract_inverted_index.The	101
abstract_inverted_index.and	2, 24, 31, 37, 78, 99, 134
abstract_inverted_index.are	104
abstract_inverted_index.but	12
abstract_inverted_index.for	106
abstract_inverted_index.its	132
abstract_inverted_index.key	124
abstract_inverted_index.not	7
abstract_inverted_index.the	117
abstract_inverted_index.SERA	74, 121
abstract_inverted_index.also	13
abstract_inverted_index.from	84
abstract_inverted_index.lack	35
abstract_inverted_index.only	8
abstract_inverted_index.show	119
abstract_inverted_index.that	55, 120
abstract_inverted_index.this	47
abstract_inverted_index.used	105
abstract_inverted_index.with	112
abstract_inverted_index.bank.	87
abstract_inverted_index.broad	9
abstract_inverted_index.cases	65
abstract_inverted_index.data.	114
abstract_inverted_index.logs,	73
abstract_inverted_index.often	34
abstract_inverted_index.their	41
abstract_inverted_index.these	94
abstract_inverted_index.those	20
abstract_inverted_index.under	136
abstract_inverted_index.across	126
abstract_inverted_index.cases,	18
abstract_inverted_index.impact	42
abstract_inverted_index.paper,	48
abstract_inverted_index.repair	15
abstract_inverted_index.robust	1
abstract_inverted_index.aligned	82
abstract_inverted_index.driving	5, 58, 129
abstract_inverted_index.enables	56
abstract_inverted_index.failure	17, 64, 76
abstract_inverted_index.further	92
abstract_inverted_index.methods	33
abstract_inverted_index.metrics	125
abstract_inverted_index.minimal	113
abstract_inverted_index.propose	50
abstract_inverted_index.refines	93
abstract_inverted_index.related	21
abstract_inverted_index.systems	59
abstract_inverted_index.through	66
abstract_inverted_index.Ensuring	0
abstract_inverted_index.However,	27
abstract_inverted_index.coverage	11
abstract_inverted_index.enabling	109
abstract_inverted_index.existing	28
abstract_inverted_index.few-shot	107
abstract_inverted_index.improves	123
abstract_inverted_index.limiting	40
abstract_inverted_index.maximize	97
abstract_inverted_index.multiple	127
abstract_inverted_index.patterns	77
abstract_inverted_index.requires	6
abstract_inverted_index.scenario	10, 29, 68
abstract_inverted_index.selected	102
abstract_inverted_index.semantic	38
abstract_inverted_index.targeted	67, 110
abstract_inverted_index.LLM-based	89
abstract_inverted_index.analyzing	71
abstract_inverted_index.benchmark	118
abstract_inverted_index.efficient	14
abstract_inverted_index.framework	54
abstract_inverted_index.mechanism	91
abstract_inverted_index.relevance	98
abstract_inverted_index.repairing	63
abstract_inverted_index.retrieves	80
abstract_inverted_index.scenarios	83, 103
abstract_inverted_index.selection	32
abstract_inverted_index.adaptation	111
abstract_inverted_index.adaptivity	36
abstract_inverted_index.autonomous	4, 57, 128
abstract_inverted_index.baselines,	130
abstract_inverted_index.diversity.	100
abstract_inverted_index.generation	30
abstract_inverted_index.identifies	75
abstract_inverted_index.reflection	90
abstract_inverted_index.relevance,	39
abstract_inverted_index.scenarios.	26
abstract_inverted_index.structured	86
abstract_inverted_index.Experiments	115
abstract_inverted_index.LLM-powered	53
abstract_inverted_index.challenging	23
abstract_inverted_index.conditions.	138
abstract_inverted_index.dynamically	79
abstract_inverted_index.performance	44, 72
abstract_inverted_index.self-evolve	61
abstract_inverted_index.consistently	122
abstract_inverted_index.fine-tuning,	108
abstract_inverted_index.improvement.	45
abstract_inverted_index.particularly	19
abstract_inverted_index.semantically	81
abstract_inverted_index.demonstrating	131
abstract_inverted_index.effectiveness	133
abstract_inverted_index.generalizable	3
abstract_inverted_index.\textbf{SERA},	51
abstract_inverted_index.recommendation.	69
abstract_inverted_index.recommendations	95
abstract_inverted_index.safety-critical	25, 137
abstract_inverted_index.generalizability	135
cited_by_percentile_year
countries_distinct_count	0
institutions_distinct_count	6
citation_normalized_percentile