AI holds great promise for industrial applications, but its complexity and opacity can hinder trust and limit adoption, particularly in high-stakes sectors, where transparency and accountability are essential. Question Answering with Explanation (QAE) helps address this challenge by enabling natural interaction with AI and providing justifications that allow users to assess, verify, and better understand AI-generated answers, thus improving perceived credibility. However, despite its clear benefits, QAE remains underexplored in industrial contexts. To address this gap, we employ a user-centric framework for QAE, and study it in system information modelling (SIM), a realistic and valuable industrial use case spanning disciplines, lifecycle stages, and downstream engineering tasks. Our framework integrates retrieval-augmented generation (RAG), with instruction, chain-of-thought, and few-shot prompting to guide LLMs in producing reasoning-rich explanations, and incorporates both automatic and user-perspective human evaluation. Extensive experiments on PDF and RDF datasets covering five key SIM topics demonstrate near-human-level performance. Case studies further illustrate our strengths in delivering strong argumentation, contextual evidence, and complex reasoning chains.

As AI-embodied autonomous agents increasingly operate in dynamic environments, especially autonomous driving, explaining their decisions is vital for transparency and trust. While generating natural language explanations for visually grounded actions is the goal, current methods often analyse individual frames or aggregate temporal data without explicitly modelling crucial visual changes between frames. This paper introduces a change-instructed approach to generate more temporally grounded action explanations. Our key contribution is the explicit incorporation of visual changes between frames, summarised offline by a large multimodal model, to guide a vision language model (VLM) in producing both the action and its justification. In this preliminary work, we evaluate the effectiveness of directly prompting a pre-trained VLM with these change summaries on a driving dataset. Our initial findings suggest that this approach enhances the accuracy and interpretability of action explanations by focussing on relevant temporal dynamics, laying the groundwork for future fine-tuning of VLMs with change-augmented data for improved performance and practicality.

Feature model evolution plans (FMEPs) describe how feature models for software product lines (SPLs) evolve over time. While different feature models can exist for different points in time over the lifetime of the product line, an FMEP describes how to compute a feature model for a given time point. SPLs capitalise on the variability and reusability of the software through combining optional and mandatory features. As business requirements change over time, FMEPs should support intermediate update. A plan hence contains updates to an initial model by adding, deleting, moving or changing elements at different points in time, in line with the evolving business requirements on the SPL, potentially affecting feature models that should be derived in the future from the plan. A recurring challenge in maintaining FMEPs is that updates may lead to inconsistent intermediate feature models, most notably so-called paradoxes. A paradox may not materialise at the first point in time an update on the plan is performed to obtain a particular feature model, but may only in combination with a later modification prescribed by the plan create a structurally invalid model. Correspondingly, a single modification to a plan may require multiple checks over the liftetime of the affected elements to rule out paradoxes. Current approaches require the analysis from the point in time an update is applied to an FMEP throughout the entire lifetime of the plan. In this paper, we define a so-called interval-based feature model (IBFM) to represent FMEPs, with a precise definition of spatial and temporal scopes that narrow the time interval and the sub-models that an update can affect. We propose a rule system for updating IBFMs, and also prove the soundness of the proposed rules and show their modularity, i.e., that each rule operates strictly within its temporal and spatial scopes. We have conducted a detailed evaluation on our modular approach and present the experimental results, which show that we outperform an existing linear approach.

Recent studies have revealed that Machine Learning (ML) models are vulnerable to adversarial perturbations. Such perturbations can be intentionally or accidentally added to the original inputs, evading the classifier’s behavior to misclassify the crafted samples. A widely-used solution is to retrain the model using data points generated by various attack strategies. However, this creates a classifier robust to some particular evasions and can not defend unknown or universal perturbations. Counterfactual explanations are a specific class of post-hoc explanation methods that provide minimal modification to the input features in order to obtain a particular outcome from the model. In addition to the resemblance of counterfactual explanations to the universal perturbations, the possibility of generating instances from specific classes makes such approaches suitable for analyzing and improving the model’s robustness. Rather than explaining the model’s decisions in the deployment phase, we utilize the distance information obtained from counterfactuals and propose novel metrics to analyze the robustness of tabular classifiers. Further, we introduce a decision boundary modification approach using customized counterfactual data points to improve the robustness of the models without compromising their accuracy. Our framework addresses the robustness of black-box classifiers in the tabular setting, which is considered an under-explored research area. Through several experiments and evaluations, we demonstrate the efficacy of our approach in analyzing and improving the robustness of black-box tabular classifiers.

The research around developing methods for debugging and refining Machine Learning (ML) models is still in its infancy. We believe employing tailored tools in the development process can help developers in creating more trustworthy and reliable models. This is particularly essential for creating black-box models such as deep neural networks and random forests, as their opaque decision-making and complex structure prevent detailed investigations. Although many explanation techniques provide interpretability in terms of predictive features for a mispredicted instance, it would be beneficial for a developer to find a partition of the training data that significantly influences the anomaly. Such responsible partitions can be subjected to data visualization and data engineering in the development phase to improve the model's accuracy. In this paper, we propose a systematic debugging framework for the development of ML models that guides the data engineering process using the model's decision boundary. Our approach finds the influential neighborhood of anomalous data points using observation-level feature importance and explains them via a novel quasi-global explanation technique. It is also equipped with a robust global explanation approach to reveal general trends and expose potential biases in the neighborhoods. We demonstrate the efficacy of the devised framework through several experiments on standard data sets and black-box models and propose various guidelines on how the framework's components can be practically useful from a developer's perspective.

Defining a representative locality is an urgent challenge in perturbation-based explanation methods, which influences the fidelity and soundness of explanations. We address this issue by proposing a robust and intuitive approach for EXPLaining black-box classifiers using Adaptive Neighborhood generation (EXPLAN). EXPLAN is a module-based algorithm consisted of dense data generation, representative data selection, data balancing, and rule-based interpretable model. It takes into account the adjacency information derived from the black-box decision function and the structure of the data for creating a representative neighborhood for the instance being explained. As a local model-agnostic explanation method, EXPLAN generates explanations in the form of logical rules that are highly interpretable and well-suited for qualitative analysis of the model's behavior. We discuss fidelity-interpretability trade-offs and demonstrate the performance of the proposed algorithm by a comprehensive comparison with state-of-the-art explanation methods LIME, LORE, and Anchor. The conducted experiments on real-world data sets show our method achieves solid empirical results in terms of fidelity, precision, and stability of explanations.

As more and more devices connect to Internet of Things, unbounded streams of data will be generated, which have to be processed "on the fly" in order to trigger automated actions and deliver real-time services. Spark Streaming is a popular realtime stream processing framework. To make efficient use of Spark Streaming and achieve stable stream processing, it requires a careful interplay between different parameter configurations. Mistakes may lead to significant resource overprovisioning and bad performance. To alleviate such issues, this paper develops an executable and configurable model named SSP (stands for Spark Streaming Processing) to model and simulate Spark Streaming. SSP is written in ABS, which is a formal, executable, and object-oriented language for modeling distributed systems by means of concurrent object groups. SSP allows users to rapidly evaluate and compare different parameter configurations without deploying their applications on a cluster/cloud. The simulation results show that SSP is able to mimic Spark Streaming in different scenarios.

Software Product Lines (SPLs) are a common technique to capture families of software products in terms of commonalities and variabilities. On a conceptual level, functionality of an SPL is modeled in terms of features in Feature Models (FMs). As other software systems, SPLs and their FMs are subject to evolution that may lead to the introduction of anomalies (e.g., non-selectable features). To fix such anomalies, developers need to understand the cause for them. However, for large evolution histories and large SPLs, explanations may become very long and, as a consequence, hard to understand. In this paper, we present a method for anomaly detection and explanation that, by encoding the entire evolution history, identifies the evolution step of anomaly introduction and explains which of the performed evolution operations lead to it. In our evaluation, we show that our method significantly reduces the complexity of generated explanations.

Metamodelling and model transformation play important roles in model-driven engineering as they can be used to define domain-specific modelling languages. During the modelling phase, modellers encode domain knowledge into models which may include both structural and behavioral aspects of a system. The contribution of this paper is a new web-based metamodelling and model transformation tool called WebDPF based on the Diagram Predicate Framework (DPF). WebDPF supports multilevel diagrammatic metamodelling and specification of model constraints, and it supports diagrammatic development and analysis of model transformation systems. We show how the support for model transformation systems in WebDPF can be exploited to (i) support auto-completion of partial models thereby enhancing modelling efficiency, and (ii) provide execution semantics for workflow models. Furthermore, we illustrate how WebDPF incorporates a scalable model navigation facility designed to enable users to inspect and query large models.

Supporting adaptation of metamodels is essential for realising Model-Driven Engineering. However, adapting and changing metamodels impact other artefacts of the metamodelling ecosystem. In particular, conformant models will no longer be valid instances of their changed metamodel. This gives rise to co-evolution issues where metamodels and models are no longer synchronised. This is critical as systems become inconsistent. A typical approach for re-establishing conformance is to manually craft transformations which update existing models for the new metamodel variant. In this paper we present an analysis-based approach that addresses this concern. The approach enables an arbitrary number of metamodels to evolve based on an adaptation strategy. During analysis we accumulate information required to automatically transform existing models to ensure conformance. We formalise the approach and prove model conformance.

Delta-oriented programming is a modular, yet flexible technique to implement software product lines. To efficiently verify the specifications of all possible product variants of a product line, it is usually infeasible to generate all product variants and to verify them individually. To counter this problem, we propose a transformational proof system in which the specifications in a delta module describe changes to previous specifications. Our approach allows each delta module to be verified in isolation, based on symbolic assumptions for calls to methods which may be in other delta modules. When product variants are generated from delta modules, these assumptions are instantiated by the actual guarantees of the methods in the considered product variant and used to derive the specifications of this product variant.

A new method for rapid construction of large-scale executable railway models is presented. Computer systems for railway systems suffer from poor integration and lack of explicit understanding of the large amount of static and dynamic information in the railway. In this paper, we give solutions to both problems. It is shown how a component-oriented approach makes it easy to construct and refine basic railway models by effective methods, such that a variety of models with important properties can be maintained within the same framework. Basic railway nets are refined into several new kinds: nets that aresafe, permitcollision detection, includetime, and aresensitiveto its surroundings. Since the underlying implementation language is Petri nets, large expressibility is combined with simplicity, and in addition, the analysis of the behavior of railway models comes gently.