https://www.mdu.se/

mdu.sePublications
System disruptions
There are recurring service disruptions on some of the search portals due to high traffic. We are working to resolve the issue, you may temporarily encounter an error message.
4243444546474845 of 78
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
Agentic Model Transformation: Service-Mediated Orchestration for Transformation Development
Mälardalen University, Faculty of Engineering and Health Sciences, Department of Computer Science & Engineering.ORCID iD: 0009-0002-7198-8514
Mälardalen University, Faculty of Engineering and Health Sciences, Department of Computer Science & Engineering.ORCID iD: 0000-0002-8027-0611
Mälardalen University, Faculty of Engineering and Health Sciences, Department of Computer Science & Engineering.ORCID iD: 0000-0003-0416-1787
2026 (English)In: MODELS '26: Proceedings of the ACM/IEEE 29th International Conference on Model Driven Engineering Languages and Systems, Association for Computing Machinery (ACM) , 2026, p. 365-371Conference paper, Published paper (Refereed)
Abstract [en]

Model transformation is a core mechanism in Model-Driven Engineering (MDE), but its use requires expertise in metamodels, transformation languages, execution engines, runtime configuration, and diagnostics. Although LLMs can generate and revise transformation artifacts, direct prompt-to-output generation does not by itself provide an executable MDE workflow grounded in metamodel conformance and transformation rules execution.

This paper proposes agentic model transformation: a service-mediated workflow in which an LLM agent generates candidate transformation specifications and coordinates MDE services for metamodel access, example retrieval, checking, execution, diagnostics, and repair. Model Context Protocol is used to expose these services to the agent. When execution succeeds, target artifacts are produced through the configured transformation infrastructure, serialized by the target modeling environment, and loadable against the registered target metamodel.

Early results from adopting the proposed framework for the UML-to-Java and SysMLv2-to-AAS cases show that constructing and repairing executable transformations is possible. The results also show that executable success and metamodel conformance are only first checks; structural completeness, mapping coverage, and semantic adequacy require additional MDE services. The framework, therefore, shifts LLM support for model transformation from one-off artifact generation towards service-supported transformation development, where specifications, diagnostics, and execution traces can be inspected, reused, and extended.

Place, publisher, year, edition, pages
Association for Computing Machinery (ACM) , 2026. p. 365-371
Keywords [en]
Agentic Model Transformation, Service-Mediated Orchestration, Large Language Models, Execution Diagnostics
National Category
Computer and Information Sciences
Identifiers
URN: urn:nbn:se:mdh:diva-79076DOI: 10.1145/3822455.3838763ISBN: 979-8-4007-2809-9 (print)OAI: oai:DiVA.org:mdh-79076DiVA, id: diva2:2104475
Conference
ACM/IEEE 29th International Conference on Model Driven Engineering Languages and Systems, 4-9 October, Malaga, Spain
Available from: 2026-09-23 Created: 2026-09-23 Last updated: 2026-09-28Bibliographically approved
In thesis
1. Toward Generative AI-assisted Model Transformation in Model-driven Engineering
Open this publication in new window or tab >>Toward Generative AI-assisted Model Transformation in Model-driven Engineering
2026 (English)Licentiate thesis, comprehensive summary (Other academic)
Abstract [en]

Model transformation is an important activity in Model-Driven Engineering (MDE), but it is often hard to use in practice. It usually requires knowledge of special languages, metamodels, and complex toolchains. Recent progress in large language models (LLMs) suggests new ways to lower this entry barrier. However, results from my work show that although LLMs can help with model transformation tasks, their performance is still unstable when the tasks involve rich meaning, complex structure, or changes across related artifacts.

This licentiate research studies how LLMs can be used to reduce the practical difficulty of model transformation in MDE while still preserving the correctness and dependability needed in engineering work. The research builds on three empirical studies covering model transformation by example, zero-shot transformation, and round-trip consistency under change. Together, these studies show both the promise of LLMs and their limits in handling structural correctness, semantic consistency, and change across connected artifacts.

Based on these findings, the next step of the research explores a hybrid direction in which LLMs are not used as stand-alone transformation engines, but instead act as coordinators of deterministic tools and services through the Model Context Protocol (MCP). The goal is to combine the flexibility of LLMs with the precision, traceability, and validation support of established MDE technologies.

The expected contribution of this licentiate is an empirical and architectural foundation for more dependable LLM-assisted model transformation. More broadly, it aims to help move from purely generative approaches toward hybrid workflows that make MDE more accessible without giving up engineering rigor.

Abstract [sv]

Modelltransformation är en viktig aktivitet inom modelldriven teknik (MDE), men den är ofta svår att använda i praktiken. Det kräver vanligtvis kunskap om specialspråk, metamodeller och komplexa verktygskedjor. Nyligen genomförda framsteg inom stora språkmodeller (LLM) tyder på nya sätt att sänka denna inträdesbarriär. Resultat från mitt arbete visar dock att även om LLM:er kan hjälpa till med modelltransformationsuppgifter, är deras prestanda fortfarande instabil när uppgifterna involverar rik betydelse, komplex struktur eller förändringar över relaterade artefakter.

Denna licentiatforskning studerar hur LLM:er kan användas för att minska den praktiska svårigheten med modelltransformation i MDE samtidigt som den korrekthet och tillförlitlighet som behövs i ingenjörsarbete bevaras. Forskningen bygger på tre empiriska studier som täcker modelltransformation med exempel, noll-skotttransformation och tur-och-retur-konsistens under förändring. Tillsammans visar dessa studier både potentialen hos LLM:er och deras begränsningar i hanteringen av strukturell korrekthet, semantisk konsistens och förändring över sammankopplade artefakter.

Baserat på dessa resultat utforskar nästa steg i forskningen en hybridriktning där LLM:er inte används som fristående transformationsmotorer, utan istället fungerar som koordinatorer för deterministiska verktyg och tjänster genom Model Context Protocol (MCP). Målet är att kombinera flexibiliteten hos LLM:er med precisionen, spårbarheten och valideringsstödet hos etablerade MDE-tekniker.

Det förväntade bidraget från denna licentiatexamen är en empirisk och arkitektonisk grund för mer pålitlig LLM-assisterad modelltransformation. Mer allmänt syftar den till att bidra till att gå från rent generativa metoder till hybrida arbetsflöden som gör MDE mer tillgänglig utan att ge avkall på teknisk stringens.

Place, publisher, year, edition, pages
Västerås: Mälardalens universitet, 2026
Series
Mälardalen University Press Licentiate Theses, ISSN 1651-9256 ; 389
National Category
Software Engineering
Identifiers
urn:nbn:se:mdh:diva-79079 (URN)978-91-7485-775-7 (ISBN)
Presentation
2026-12-16, CASE och digitalt, Mälardalens universitet, Västerås, 13:15 (English)
Opponent
Supervisors
Available from: 2026-10-07 Created: 2026-09-24 Last updated: 2026-10-07Bibliographically approved

Open Access in DiVA

fulltext(448 kB)17 downloads
File information
File name FULLTEXT01.pdfFile size 448 kBChecksum SHA-512
3f62293e0abbe4c73cf39d25718dea2af3a2e83906b9c73fdbe7c886ba45e46598dbc132d68786bf72a5dba036e81392afa142690ccc1305bda22985658fadd0
Type fulltextMimetype application/pdf

Other links

Publisher's full texthttps://doi.org/10.1145/3822455.3838763

Authority records

Dao, Khanh DuyBucaioni, AlessioCicchetti, Antonio

Search in DiVA

By author/editor
Dao, Khanh DuyBucaioni, AlessioCicchetti, Antonio
By organisation
Department of Computer Science & Engineering
Computer and Information Sciences

Search outside of DiVA

GoogleGoogle Scholar
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

doi
isbn
urn-nbn

Altmetric score

doi
isbn
urn-nbn
Total: 592 hits
4243444546474845 of 78
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf