Workshop program

Welcome and Opening: 20 Years of Models and Evolution

Authors: Wael Kessentini, Juri Di Rocco and Djamel Eddine Khelladi

Presenter: ME 2026 organizers

9:15-9:25

Opening of the 20th edition of the Models and Evolution workshop: scope of the anniversary edition, special theme on Digital Twins, and practical information about the day.

Keynote: Evolution in Model-Driven Engineering: From Knowing What Changed to Knowing What to Change

Authors: Gabriele Taentzer (Philipps-Universitat Marburg, Germany)

Presenter: Gabriele Taentzer

9:25-10:30

Evolution is a fundamental challenge in model-driven engineering (MDE), affecting models, metamodels, transformations, and the systems and processes built around them. This talk provides an overview of the many facets of evolution in MDE, ranging from linear model evolution and the co-evolution of coupled artifacts to planned changes and configuration management. For selected challenges, we will revisit classical approaches and examine how they can be complemented, extended, or transformed by recent advances in artificial intelligence (AI). In particular, the talk will explore how AI can take us from knowing what changed to knowing what to change, while examining the opportunities and challenges of combining AI with established model-driven techniques.

A System-Level Perspective on Drift in Digital Twins

Authors: Taguhi Mesropyan, Paul Groth and Victoria Degeler (University of Amsterdam, Netherlands)

11:00-11:25

Digital Twins (DTs) operate in dynamic environments where changes over time can affect the validity and interpretation of their outputs. Despite this, the notion of drift in DT systems remains ambiguous and inconsistently defined. In this work, we define drift in DT systems as changes that affect the validity, consistency, or interpretation of their data, models, representations, or outputs over time, and propose a taxonomy for its systematic characterization, addressing ambiguities in existing terminology and providing a unified representation of drift phenomena. The proposed framework characterizes drift along three complementary dimensions, capturing where drift occurs, what type of change it represents, and its level of intentionality. We demonstrate its applicability in a case study with various drift scenarios. Finally, we discuss how the proposed characterization enables the systematic analysis of drift, supports the identification of common patterns across domains, and facilitates the design of more robust and adaptable DT systems.

Assessing the Evolvability of Greenhouse Digital Twins on an Architectural Level

Authors: Oscar van Zelm (Wageningen University, Netherlands), Joachim Denil (University of Antwerp, Belgium), Onder Babur (Wageningen University and Eindhoven University of Technology, Netherlands) and Tarek Alskaif (Wageningen University, Netherlands)

11:25-11:50

Improving agricultural productivity increasingly relies on data-driven software solutions. Among these, digital twins (DTs) provide farmers with detailed insights into crop development, enabling more efficient farming practices and higher yields. These software systems are not static, as they continuously evolve to accommodate changes in their requirements and environmental conditions. Evolutions can be motivated by a multitude of triggers, ranging from expansion of the size of the greenhouse, to user-interface updates of the software. The difficulty of implementing such evolutions depends on both the nature of the evolution, and the software architecture's evolvability, its ability to support evolution. Although software evolution is often discussed as a concept, empirical research on software evolution in applied domains is lacking. This study addresses this knowledge gap by applying the scenario-based Architecture-Level Modifiability Analysis (ALMA) to two commercial greenhouse DT architectures, demonstrating the applicability of ALMA to evaluate DT evolvability. It also provides a list of twelve greenhouse-specific DT evolution triggers, a set of twenty domain-specific evolution scenarios, and elaborate ALMA assessments of two commercial DT architectures. The scenarios for the ALMA were identified through a bottom-up approach, interviewing six relevant stakeholders. The ALMA revealed that both architectures were flexible for evolution regarding the layout and size of the greenhouse, but were inflexible for updating the DT's prediction models with new inputs and outputs, such as integrating new types of greenhouse sensors and actuators. These findings provide insight into the state-of-the-art of greenhouse DT evolvability and provide direction for future advancements.

Towards Illusions Awareness in Cyber-Physical System's Design

Authors: Anna Di Placido, Nicolas Ferry and Julien Deantoni (Universite Cote d'Azur, I3S/INRIA Kairos, France)

11:50-12:15

Cyber-Physical Systems (CPS) operate through a continuous sense-compute-act loop within an open context environment, making it impossible to anticipate all the situations the system will face. To cope with this openness, stakeholders rely on assumptions, formalized into design models. However, these assumptions may no longer hold once the system is confronted with runtime reality, resulting in a discrepancy between expected and observed behaviour known in literature as the reality gap. Existing approaches mainly focus on reducing or overcoming it by making simulations more faithful to reality, with no unified methodology to structure and exploit invalidated assumptions that give rise to this gap as reusable design knowledge. We refer to the persistent reliance on invalidated assumptions, and the resulting false confidence in the design model's operational validity, as design illusions, and argue that they need to be made explicit, structured, and exploited as knowledge to support better design decisions. We propose a conceptual pipeline for illusions-awareness that identifies, classifies, characterizes, and leverages illusions to transform them into actionable design knowledge.

Special Theme Discussion: Model Evolution in Digital Twins

Authors: All participants

12:15-12:40

Open discussion on the special theme of the 2026 edition. Starting from the three talks of the session, we collect open problems in drift detection, co-evolution and evolvability assessment for Digital Twin systems, and identify candidate topics for future community work.

Morning Wrap-Up

Authors: ME 2026 organizers

Presenter: ME 2026 organizers

12:40-12:45

Short wrap-up of the morning sessions before the lunch break.

SMDSL: A Domain-Specific Language for Describing Software Platform Migrations

Authors: Mohamadreza Sabeghi and Richard F. Paige (McMaster University, Canada) and Dimitris Kolovos (University of York, United Kingdom)

14:30-14:55

Software platform migration involves more than executing transformations. It requires reasoning about feature mismatches, semantic changes, automation limits, and manual decisions. To address this gap, this paper introduces SMDSL, a domain-specific language for representing migration knowledge. We apply SMDSL to four migration scenarios and provide Picto-based views and a Java analyzer for visualizing, querying, and summarizing SMDSL models. The examples and discussion suggest that SMDSL makes migration knowledge explicit, traceable, and analyzable, while introducing modeling effort and requiring further empirical validation.

Continuous Verification of Evolving Architecture Models: A Testing Framework for MontiArc

Authors: Tom Bursch, Adrian Marin, Bernhard Rumpe and David Schmalzing (RWTH Aachen University, Germany)

14:55-15:20

Domain-specific languages and modeling languages benefit from dedicated testing frameworks that enable modelers to specify, organize, and execute tests at the same level of abstraction as the models under test. This paper presents MaUnit, a framework for specifying and executing model-level tests for MontiArc. MaUnit provides a dedicated testing notation together with an execution platform that supports reusable test components, parameterized test cases, and automated result reporting. We describe the key design decisions underlying the framework and evaluate its applicability by re-implementing the majority of MontiArc's simulator integration tests.

Speeding Up Incremental Model-to-Text Transformations with File-Level Hashing

Authors: Adam Blanchet, Dimitris Kolovos, Antonio Garcia-Dominguez and Simos Gerasimou (University of York, United Kingdom)

15:20-15:45

As model-to-text (M2T) transformations are among the most widely used modelling activities, repeatedly executing them in full after small changes to their source model can become unnecessarily costly. Identifying which parts of a transformation need to be re-executed following such modifications can significantly reduce this cost. We present an enhancement to an existing offline approach for incremental M2T transformations, targeting models split across multiple files. Our approach accelerates impact analysis through file-level hashing, enabling unchanged files to be excluded from further analysis. Preliminary experimental results indicate that our approach can reduce M2T transformation execution time by up to 20% in certain cases.

An Experience Report from Evolving a JetBrains MPS-based Domain Specific Modeling Tool over 75 Versions and 5 Years

Authors: Daniel Ratiu and Leonie Koehler (Cariad, Germany)

16:15-16:40

Domain specific modeling languages (DSMLs) allow users to directly express their content using abstractions close to their application domain. Ideally, the resulting models are clean, concise and amenable to automation. The needs of domain experts are evolving continuously and, in order to remain relevant, the DSMLs need to keep up. In this paper we present our experiences with developing and evolving a large eco-system of DSMLs which make up our architecture modeling tool ArchE, built on top of JetBrains' MPS language workbench. ArchE development features a continuous delivery pipeline with a new release every four weeks. ArchE DSLs are comprised of more than 5000 meta-classes. Over one hundred ArchE users work in parallel in several major git repositories, with dependencies among each other, contributing to their continuous delivery pipelines. Each users repository contains between several hundreds of thousands and a few millions model elements. We present our approach of implementing over 400 languages evolutions and migrating the associated user models over a period of 5 years and 75 versions of the tool. We present specific requirements we had to address, the mechanisms provided by MPS which support the evolution, the set of custom technical solutions we additionally developed and the processes we put in place to ensure discipline and rigour. We conclude the paper with a set of lessons learned.

TBA

16:40-17:05

Title and authors to be announced.

Panel and Closing Remarks: 20 Years of Models and Evolution

Authors: Gabriele Taentzer and panelists TBA

Presenter: ME 2026 organizers

17:05-17:30

Panel discussion closing the anniversary edition: which evolution problems the community considers solved, which remain open after twenty years, and how AI-based techniques change the research agenda. Followed by the closing remarks of the organizers.