Rapid advances in large language models (LLMs) have expanded their use across software engineering workflows, yet most prompt engineering remains ad hoc —yielding inconsistent quality, poor transferability, and inefficient iteration. This paper introduces and evaluates Preference-Driven Refinement (PDR), a structured prompt engineering method that iteratively updates prompts by incorporating user-selected preferred and non-preferred elements from model outputs. Building on in context learning, PDR uses synthetic example generation and explicit preference incorporation to overcome common challenges, such as instruction blindness, inadequate context capture, and alignment drift. We extend PDR with PDR+Critic, a multi-LLM refinement loop where a separate Critic model evaluates generated outputs, identifies strengths and weaknesses, and recommends improvements. To assess these structured refinement methods, we conduct a controlled, simulation-based study across five software engineering tasks and eight personas using GPT-5, with decoding parameters held constant to isolate methodological effects. To address evaluation circularity concerns, we employ multi-model cross-validation using independent LLM families (GPT-4o and Gemini-2.5-Pro) as evaluators Our results reveal both methodological contributions and challenges. PDR significantly reduces refinement iterations compared to ad hoc prompting (p = 0.020, FDR-corrected), demonstrating improved convergence efficiency through fewer refinement cycles. However, PDR exhibits longer total execution time due to structured processing overhead (Ad Hoc: 67.5s, PDR: 179.6s, PDR+Critic: 209.1s on average). PDR+Critic shows mixed effectiveness: while adding explicit evaluation loops, it often increases runtime and iteration count without consistent quality improvements, and both evaluators identified cases where critic feedback degraded output quality. Critically, multi-model evaluation revealed low inter-rater agreement (r = 0.044, p = 0.794) between independent evaluators, indicating that different LLM families apply fundamentally different quality standards. This finding weakens claims about quality improvements but contributes important evidence about current limitations in LLM-as-judge methodologies. Objective metrics (iteration counts, timing) remain valid and demonstrate PDR’s efficiency trade-offs: fewer iterations but longer execution time per iteration. By formalizing PDR and PDR+Critic, evaluating their performance with rigorous cross-model validation, and transparently reporting both successes and limitations, this paper advances systematic prompt engineering research. Our findings reveal key trade-offs between structure and efficiency, document how multi-LLM Generator–Critic loops interact with software engineering tasks, and motivate future work on standardized, reliable evaluation frameworks for LLM-augmented software development.

OBJECTIVE: Returning aggregate study results is an important ethical responsibility to promote trust and inform decision making, but the practice of providing results to a lay audience is not widely adopted. Barriers include significant cost and time required to develop lay summaries and scarce infrastructure necessary for returning them to the public. Our study aims to generate, evaluate, and implement ChatGPT 4 lay summaries of scientific abstracts on a national clinical study recruitment platform, ResearchMatch, to facilitate timely and cost-effective return of study results at scale. MATERIALS AND METHODS: We engineered prompts to summarize abstracts at a literacy level accessible to the public, prioritizing succinctness, clarity, and practical relevance. Researchers and volunteers assessed ChatGPT-generated lay summaries across five dimensions: accuracy, relevance, accessibility, transparency, and harmfulness. We used precision analysis and adaptive random sampling to determine the optimal number of summaries for evaluation, ensuring high statistical precision. RESULTS: ChatGPT achieved 95.9% (95% CI, 92.1-97.9) accuracy and 96.2% (92.4-98.1) relevance across 192 summary sentences from 33 abstracts based on researcher review. 85.3% (69.9-93.6) of 34 volunteers perceived ChatGPT-generated summaries as more accessible and 73.5% (56.9-85.4) more transparent than the original abstract. None of the summaries were deemed harmful. We expanded ResearchMatch's technical infrastructure to automatically generate and display lay summaries for over 750 published studies that resulted from the platform's recruitment mechanism. DISCUSSION AND CONCLUSION: Implementing AI-generated lay summaries on ResearchMatch demonstrates the potential of a scalable framework generalizable to broader platforms for enhancing research accessibility and transparency.

There is a substantial and growing body of malware samples that evade automated analysis and detection tools. Malware may measure fingerprints (“artifacts”) of the underlying analysis tool or environment, and change their behavior when such artifacts are detected. While analysis tools can mitigate artifacts to reduce exposure, such concealment is expensive and limits scalable automated malware analysis. However, not every sample checks for every type of artifact—analysis efficiency can be improved by mitigating only those artifacts most likely to be used by a sample. Using that insight, we proposeMimosa, a system that identifies a small set of “covering” configurations that collectively and efficiently defeat most malware samples in a corpus.Mimosaidentifies a set of configurations that maximize analysis throughput and detection accuracy while minimizing manual effort, enabling scalable automation for analyzing stealthy malware. We evaluate our approach against a benchmark of 1535 meticulously labeled stealthy malware samples. We further test our approach on an additional set of 1221 stealthy malware samples and successfully analyze nearly 99% of them using only 2 VM backends.Mimosaprovides a practical, tunable method for efficiently deploying malware analysis resources.

Spoof speech can be used to try and fool speaker verification systems that determine the identity of the speaker based on voice characteristics. This paper compares popular learnable front-ends on this task. We categorize the front-ends by defining two generic architectures and then analyze the filtering stages of both types in terms of learning constraints. We pro-pose replacing fixed filterbanks with a learnable layer that can better adapt to anti-spoofing tasks. The proposed FastAudio front-end is then tested with two popular back-ends to measure the performance on the Logical Access track of the ASVspoof 2019 dataset. The FastAudio front-end achieves a relative improvement of 29.7% when compared with fixed front-ends, outperforming all other learnable front-ends on this task.

An emerging trend in audio processing is capturing low-level speech representations from raw waveforms. These representations have shown promising results on a variety of tasks, such as speech recognition and speech separation. Compared to handcrafted features, learning speech features via backpropagation can potentially provide the model greater flexibility in how it represents data for different tasks. However, results from empirical studies show that, in some tasks, such as spoof speech detection, handcrafted features still currently outperform learned features. Instead of evaluating handcrafted features and raw waveforms independently, this paper proposes an Auxiliary Rawnet model to complement handcrafted features with features learned from raw waveforms for spoof speech detection. A key benefit of the approach is that it can improve accuracy at a relatively low computational cost. The proposed Auxiliary Rawnet model is tested using the ASVspoof 2019 dataset and pooled EER and min-tDCF are 1.11% and 0.03645 respectively. Results from this dataset indicate that a lightweight waveform encoder can boost the performance of handcrafted-features-based encoders for 10 types of spoof attacks, including 3 challenging attacks, in exchange for a small amount of additional computational work.

Research in the past several years has boosted the performance of automatic speaker verification systems and countermeasure systems to deliver low Equal Error Rates (EERs) on each system. However, research on joint optimization of both systems is still limited. The Spoofing-Aware Speaker Verification (SASV) 2022 challenge was proposed to encourage the development of integrated SASV systems with new metrics to evaluate joint model performance. This paper proposes an ensemble-free end-to-end solution, known as Spoof-Aggregated-SASV (SA-SASV) to build a SASV system with multi-task classifiers, which are optimized by multiple losses and has more flexible requirements in training set. The proposed system is trained on the ASVSpoof 2019 LA dataset, a spoof verification dataset with small number of bonafide speakers. Results of SASV-EER indicate that the model performance can be further improved by training in complete automatic speaker verification and countermeasure datasets.

This paper explores the translation of natural language into Bash Commands, which developers commonly use to accomplish command-line tasks in a terminal. In our approach a terminal takes a command as a sentence in plain English and translates it into the corresponding string of Bash Commands. The paper analyzes the performance of several architectures on this translation problem using the data from the NLC2CMD competition at the NeurIPS 2020 conference. The approach presented in this paper is the best performing architecture on this problem to date and improves the current state-of-the-art accuracy on this translation task from 13.8% to 53.2%.

Data visualization has become a vital tool to help people understand the driving forces behind real-world phenomena. Although the learning curve of visualization tools have been reduced, domain experts still often require significant amounts of training to use them effectively. To reduce this learning curve even further, this paper proposes Sketch2Vis, a novel solution using deep learning techniques and tools to generate the source code for multi-platform data visualizations automatically from hand-drawn sketches provided by domain experts, which is similar to how an expert might sketch on a cocktail napkin and ask a software engineer to implement the sketched visualization.This paper explores key challenges (such as model training) in generating visualization code from hand-drawn sketches since acquiring a large dataset of sketches paired with visualization source code is often prohibitively complicated. We present solutions for these problems and conduct experiments on three baseline models that demonstrate the feasibility of generating visualizations from hand-drawn sketches. The best models tested reach a structural accuracy of 95% in generating correct data visualization code from hand-drawn sketches of visualizations.

Website visitors are performing increasingly complex computational work on the websites' behalf, such as validating forms, rendering animations, and producing data visualizations. In this article, we explore the possibility of increasing the work offloaded to web visitors' browsers. The idle computing cycles of web visitors can be turned into a large-scale distributed data processing engine, which we term gray computing. Past research has looked primarily at either volunteer computing with specialized clients or browser-based volunteer computing where the visitors keep their browsers open to a single web page for a long period of time. This article provides a comprehensive analysis of the architecture, performance, security, cost effectiveness, user experience, and other issues of gray computing distributed data processing engines with heterogeneous computing power, non-uniform page view times, and high computing pool volatility. Several real-world applications are examined and gray computing is shown to be cost effective for a number of complex tasks ranging from computer vision to bioinformatics to cryptology.

Outpatient clinics often run behind schedule due to patients who arrive late or appointments that run longer than expected. We sought to develop a generalizable method that would allow healthcare providers to diagnose problems in workflow that disrupt the schedule on any given provider clinic day. We use a constraint optimization problem to identify the least number of appointment modifications that make the rest of the schedule run on-time. We apply this method to an outpatient clinic at Vanderbilt. For patient seen in this clinic between March 27, 2017 and April 21, 2017, long cycle times tended to affect the overall schedule more than late patients. Results from this workflow diagnosis method could be used to inform interventions to help clinics run smoothly, thus decreasing patient wait times and increasing provider utilization.

Mobile health (mHealth) apps have been adopted in healthcare areas such as the management of diabetes, the monitoring of physical activities and the treatment of HIV. The users of mHealth may be patients with handicaps such as motor impairments, difficulties to remember and psychiatric conditions. Hence, assumptions normally made on other types of apps, like the user's ability to type or remember a password, might not hold in the mHealth area. This paper evaluates how different authentication approaches impact mHealth apps usability. Second, we present new metrics to evaluate ease-of-use and third, we evaluate the usability of two common authentication approaches for mHealth apps via several key process aspects and their impact on users. Based on these results, we propose a QR-Code based authentication approach for mHealth apps, which helps overcome common impediments faced by mHealth apps users.

Non-recurring traffic congestion is caused by temporary disruptions, such as accidents, sports games, adverse weather, etc. We use data related to real-time traffic speed, jam factors (a traffic congestion indicator), and events collected over a year from Nashville, TN to train a multi-layered deep neural network. The traffic dataset contains over 900 million data records. The network is thereafter used to classify the real-time data and identify anomalous operations. Compared with traditional approaches of using statistical or machine learning techniques, our model reaches an accuracy of 98.73 percent when identifying traffic congestion caused by football games. Our approach first encodes the traffic across a region as a scaled image. After that the image data from different timestamps is fused with event- and time-related data. Then a crossover operator is used as a data augmentation method to generate training datasets with more balanced classes. Finally, we use the receiver operating characteristic (ROC) analysis to tune the sensitivity of the classifier. We present the analysis of the training time and the inference time separately.

Blockchain is a decentralized, trustless protocol that combines transparency, immutability, and consensus properties to enable secure, pseudo-anonymous transactions. Smart contracts are built atop a blockchain to support on-chain storage and enable Decentralized Apps (DApps) to interact with the blockchain programatically. Programmable blockchains have generated interest in the healthcare domain as a potential solution to resolve key challenges, such as gapped communications, inefficient clinical report delivery, and fragmented health records. This paper provides evaluation metrics to assess blockchain-based DApps in terms of their feasibility, intended capability, and compliance in the healthcare domain.

The on-time arrival performance of vehicles at stops is a critical metric for both riders and city planners to evaluate the reliability of a transit system. However, it is a non-trivial task for transit agencies to adjust the existing bus schedule to optimize the on-time performance for the future. For example, severe weather conditions and special events in the city could slow down traffic and cause bus delay. Furthermore, the delay of previous trips may affect the initial departure time of consecutive trips and generate accumulated delay. In this paper, we formulate the problem as a single-objective optimization task with constraints and propose a greedy algorithm and a genetic algorithm to generate bus schedules at timepoints that improve the bus on-time performance at timepoints which is indicated by whether the arrival delay is within the desired range. We use the Nashville bus system as a case study and simulate the optimization performance using historical data. The comparative analysis of the results identifies that delay patterns change over time and reveals the efficiency of the greedy and genetic algorithms.

Public bus transit plays an important role in city transportation infrastructure. However, public bus transit is often difficult to use because of lack of real- time information about bus locations and delay time, which in the presence of operational delays and service alerts makes it difficult for riders to predict when buses will arrive and plan trips. Precisely tracking vehicle and informing riders of estimated times of arrival is challenging due to a number of factors, such as traffic congestion, operational delays, varying times taken to load passengers at each stop. In this paper, we introduce a public transportation decision support system for both short-term as well as long-term prediction of arrival bus times. The system uses streaming real-time bus position data, which is updated once every minute, and historical arrival and departure data - available for select stops to predict bus arrival times. Our approach combines clustering analysis and Kalman filters with a shared route segment model in order to produce more accurate arrival time predictions. Experiments show that compared to the basic arrival time prediction model that is currently being used by the city, our system reduces arrival time prediction errors by 25% on average when predicting the arrival delay an hour ahead and 47% when predicting within a 15 minute future time window.

Websites routinely distribute small amounts of work to visitors' browsers in order to validate forms, render animations, and perform other computations. This paper examines the feasibility, cost effectiveness, and approaches for increasing the workloads offloaded to web visitors' browsers in order to turn them into a large-scale distributed data processing engine, which we term gray computing. Past research has looked primarily at either non-browser based volunteer computing or browser-based volunteer computing where the visitors keep their browsers open to a single web page for a long period of time. This paper provides a deep analysis of the architectural, cost effectiveness, user experience, performance, security, and other issues of gray computing distributed data processing engines with high heterogeneity, non-uniform page view times, and high computing pool volatility.

This paper describes the state of the art in software frameworks for executing Software Defined Radio (SDR) components. These frameworks are catalyzing drastic changes in signal processing by enabling software engineers and signal processing engineers to work in tandem on core challenges, such as effectively processing large amounts of data in real-time on limited hardware resources. In addition to a historical perspective of this area, we showcase the REDHAWK framework as an example of a modern SDR framework which provides many facilities for distributed SDR deployment.

Large-scale cyber-physical systems (CPS) in mission-critical areas such as transportation, health care, energy, agriculture, defense, homeland security, and manufacturing, are becoming increasingly interconnected and interdependent. These types of CPS are unique in their need to combine rigorous control over timing and physical properties, as well as functional ones, while operating dynamically, reliably and affordably over significant scales of distribution, resource consumption, and utilization. As large-scale CPS continue to evolve-and grow in scale and complexity-they will impose significant and novel requirements for a new kind of cloud computing that is not supported by conventional technologies To meet these requirements, cloud computing advances are needed to establish real-time computing, communication, and control foundations rigorously at scale. Likewise, advances are needed to apply these foundations in a flexible and scalable manner to different real-world large-scale CPS challenge problems. To support both foundational and experimental R&D, a new generation of elastic infrastructure must be designed, developed, and evaluated. This paper identifies challenges, opportunities, and benefits for this work and for the largescale CPS it targets.

The articles in this special issue focus on the technology and applications supported by augmented reality.

The widespread adoption and contextually sensitive nature of smartphone devices has increased concerns over smartphone malware. Machine learning classifiers are a current method for detecting malicious applications on smartphone systems. This paper presents the evaluation of a number of existing classifiers, using a dataset containing thousands of real (i.e. not synthetic) applications. We also present our STREAM framework, which was developed to enable rapid large-scale validation of mobile malware machine learning classifiers.

Determining what physical activity the user is performing using mobile devices is an elusive task even though consumer off the shelf (COTS) mobile devices carried by most users contain a vast array of physical sensors. When conducting work improvement studies, construction crew members are typically observed by eye to record information about the physical activities being performed. Another approach to capturing crew member activities is utilizing sensors within COTS mobile devices carried by crew members. This paper presents CTrack: an approach for capturing telemetry from the array of sensors on-board COTS mobile devices. The resulting sensor telemetry is fused together and can be used to make informed decisions when characterizing crew member activities. Additionally, this paper presents results from field testing CTrack and discusses the current benefits and limitations of fusing multiple telemetry sources to characterize crew member activities.

ABSTRACT Domain‐Specific Modeling Languages (DSMLs) are playing an increasingly significant role in software development. By raising the level of abstraction using notations that are representative of a specific domain, DSMLs allow the core essence of a problem to be separated from irrelevant accidental complexities, which are typically found at the implementation level in source code. In addition to modeling the functional aspects of a system, a number of nonfunctional properties (e.g., quality of service constraints and timing requirements) also need to be integrated into models in order to reach a complete specification of a system. This is particularly true for domains that have distributed real time and embedded needs. Given a base model with functional components, maintaining the nonfunctional properties that crosscut the base model has become an essential modeling task when using DSMLs. The task of maintaining nonfunctional properties in DSMLs is traditionally supported by manual model editing or by using model transformation languages. However, these approaches are challenging to use for those unfamiliar with the specific details of a modeling transformation language and the underlying metamodel of the domain, which presents a7 steep learning curve for many users. This paper presents a demonstration‐based approach to automate the maintenance of nonfunctional properties in DSMLs. Instead of writing model transformation rules explicitly, users demonstrate how to apply the nonfunctional properties by directly editing the concrete model instances and simulating a single case of the maintenance process. By recording a user's operations, an inference engine analyzes the user's intention and generates generic model transformation patterns automatically, which can be refined by users and then reused to automate the same evolution and maintenance task in other models. Using this approach, users are able to automate the maintenance tasks without learning a complex model transformation language. In addition, because the demonstration is performed on model instances, users are isolated from the underlying abstract metamodel definitions. Our demonstration‐based approach has been applied to several scenarios, such as auto scaling and model layout. The specific contribution in this paper is the application of the demonstration‐based approach to capture crosscutting concerns representative of aspects at the modeling level. Several examples are presented across multiple modeling languages to demonstrate the benefits of our approach. Copyright © 2013 John Wiley & Sons, Ltd.

Consider non-preemptive fixed-priority scheduling of arbitrary-deadline sporadic tasks on a single processor assuming that the execution time of a job J depends on the actual schedule (sequence) of jobs executed before J. We present exact schedulability analysis for such a system.

Smartphone sales are expected to outpace desktop/laptop computer sales in 2011. It is critical for software engineers to understand the key issues of building applications for this new platform. The mobile sensing and networking capabilities of smartphones create a unique platform for building cyberphysical and other applications that sense and respond to the environment. Moreover, social networking capabilities of these platforms offer new paradigms for dissemination of knowledge, harvesting of user relationship information, and following current events. This workshop will foster new ideas, approaches, and artifacts that can be used to support the introduction of smartphones into traditional software engineering courses (e.g., a survey course, or senior capstone design course, among others).

Model evolution is an essential activity in software system modeling, which is traditionally supported by manual editing or writing model transformation rules. However, the current state of practice for model evolution presents challenges to those who are unfamiliar with model transformation languages or metamodel definitions. This demonstration presents a demonstration-based approach that assists end-users through automation of model evolution tasks (e.g., refactoring, model scaling, and aspect weaving).

Smart phones are starting to find use in mission critical applications, such as search-and-rescue operations, wherein the mission capabilities are realized by deploying a collaborating set of services across a group of smart phones involved in the mission. Since these missions are deployed in environments where replenishing resources, such as smart phone batteries, is hard, it is necessary to maximize the lifespan of the mission while also maintaining its real-time quality of service (QoS) requirements. To address these requirements, this paper presents a deployment framework called Smart Deploy, which integrates bin packing heuristics with evolutionary algorithms to produce near-optimal deployment solutions that are computationally inexpensive to compute for maximizing the lifespan of smart phone-based mission critical applications. The paper evaluates the merits of deployments produced by Smart Deploy for a search-and-rescue mission comprising a heterogeneous mix of smart phones by integrating a worst-fit bin packing heuristic with particle swarm optimization and genetic algorithm. Results of our experiments indicate that the missions deployed using Smart Deploy have a lifespan that is 20% to 162% greater than those deployed using just the bin packing heuristic or evolutionary algorithms. Although Smart Deploy is slightly slower than the other algorithms, the slower speed is acceptable for offline computations of deployment.

Distributed real-time and embedded (DRE) systems can be composed of hundreds of software components running across tens or hundreds of networked processors that are physically separated from one another. A key concern in DRE systems is determining the spatial deployment topology, which is how the software components map to the underlying hardware components. Optimizations, such as placing software components with high-frequency communications on processors that are closer together, can yield a number of important benefits, such as reduced power consumption due to decreased wireless transmission power required to communicate between the processing nodes. Determining a spatial deployment plan across a series of processors that will minimize power consumption is hard since the spatial deployment plan must respect a combination of real-time scheduling, fault-tolerance, resource, and other complex constraints. This article presents a hybrid heuristic/evolutionary algorithm, called ScatterD, for automatically generating spatial deployment plans that minimize power consumption. This work provides the following contributions to the study of spatial deployment optimization for power consumption minimization: (1) it combines heuristic bin-packing with an evolutionary algorithm to produce a hybrid algorithm with excellent deployment derivation capabilities and scalability, (2) it shows how a unique representation of the spatial deployment solution space integrates the heuristic and evolutionary algorithms, and (3) it analyzes the results of experiments performed with data derived from a large-scale avionics system that compares ScatterD with other automated deployment techniques. These results show that ScatterD reduces power consumption by between 6% and 240% more than standard bin-packing, genetic, and particle swarm optimization algorithms.

Model transformation is a core technology in Domain-Specific Modeling (DSM). While a number of model transformation languages and tools have been developed to support model transformation activities, the layout of visual models in the transformation process is not often considered. In many cases, after a transformation is performed the layout of the resulting model must be manually rearranged, which can be time consuming. The automatic layout arrangement features provided by some modeling tools usually do not take a user's preferences or the semantics of the model into consideration, and therefore could potentially alter the desired layout in an undesired manner. This paper describes a new approach to enable users to specify the model layout in a model transformation. We applied the Model Transformation By Demonstration (MTBD) approach and extended it to let users specify the layout information using the concept of "What You See Is What You Get" (WYSIWYG), so that the complex layout specification can be simplified.

Search-based software engineering is an emerging paradigm that uses automated search algorithms to help designers iteratively find solutions to complicated design problems. For example, when designing a climate monitoring satellite, designers may want to use the minimal amount of computing hardware to reduce weight and cost while supporting the image processing algorithms running onboard. A key problem in these situations is that the hardware and software designs are locked in a tightly coupled cost-constrained producer/consumer relationship that makes it hard to find a good hardware/software design configuration. Search-based software engineering can be used to apply algorithmic techniques to automate the search for hardware/software designs that maximize the image processing accuracy while respecting cost constraints. This paper provides the following contributions to research on search-based software engineering: 1) We show how a cost-constrained producer/consumer problem can be modeled as a set of two multidimensional multiple-choice knapsack problems (MMKPs), 2) we present a polynomial-time search-based software engineering technique, called the Allocation-baSed Configuration Exploration Technique (ASCENT), for finding near optimal hardware/software codesign solutions, and 3) we present empirical results showing that ASCENT's solutions average over 95 percent of the optimal solution's value.

Feature models are widely used to model software product-line (SPL) variability. SPL variants are configured by selecting feature sets that satisfy feature model constraints. Configuration of large feature models can involve multiple stages and participants, which makes it hard to avoid conflicts and errors. New techniques are therefore needed to debug invalid configurations and derive the minimal set of changes to fix flawed configurations. This paper provides three contributions to debugging feature model configurations: (1) we present a technique for transforming a flawed feature model configuration into a Constraint Satisfaction Problem (CSP) and show how a constraint solver can derive the minimal set of feature selection changes to fix an invalid configuration, (2) we show how this diagnosis CSP can automatically resolve conflicts between configuration participant decisions, and (3) we present experiment results that evaluate our technique. These results show that our technique scales to models with over 5,000 features, which is well beyond the size used to validate other automated techniques.

The decentralized process of configuring enterprise applications is complex and error-prone, involving multiple participants/roles and numerous configuration changes across multiple files, application server settings, and database decisions. This paper describes an approach to automated enterprise application configuration that uses a feature model, executes a series of probes to verify configuration properties, formalizes feature selection as a constraint satisfaction problem, and applies constraint logic programming techniques to derive a correct application configuration. To validate the approach, we developed a configuration engine, called Fresh, for enterprise Java applications and conducted experiments to measure how effectively Fresh can configure the canonical Java Pet Store application. Our results show that Fresh reduces the number of lines of hand written XML code by up to 92% and the total number of configuration steps by up to 72%.

Product-line architectures (PLAs) designed for mobile devices create a unique challenge for automated product variant selection engines since variants must be derived on-the-fly as devices are discovered. Current automation techniques do not incorporate device resource consumption constraints into variant selection and do not address how a PLA can be designed to improve automated variant selection speed. This paper presents a tool called Scatter whose input is (1) the requirements of PLA construction and (2) the resources available on a discovered mobile device and whose output is the optimal variant that can be deployed to the device. Scatter provides automatic variant selection based on configuration and resource constraints and also ensures that variant selection is optimal with regard to a configurable cost function. The paper presents our results from experiments with Scatter and how PLA design decisions affect a constraint-based variant selection engine's solving speed.