October 25 ~ 26, 2025, Vienna, Austria
Davide AVESANI, Ammar KHEIRBEK, Isep - Institut Sup´erieur d’Electronique de Paris ´10 rue de Vanves, 92130 Issy-les-Moulineaux, France
Social media is now deeply integrated into people’s daily life, enabling rapid information exchange and global connectivity. Unfortunately, harmful content can be easily disseminated among all communities, including hate speech and biases against vulnerable groups such as people with disabilities. While social media platforms employ a mix of automated systems and skillful experts for content moderation, significant challenges remain in detecting nuanced hate speech, particularly when expressed through indirect or coded language. This paper proposes a novel approach to address these challenges through HEROL (Hate-speech Evaluation via RAG and Optimized LLM), a unified model that combines RAG-Enhanced Large Language Models with Prompt Engineering and Fine-Tuning. Experimental results, obtained through a structured evaluation methodology using annotated social media datasets, demonstrated that HEROL achieved an accuracy improvement by up to 10% compared to baseline models. This highlights its effectiveness in identifying subtle and indirect forms of hate speech and its potential to contribute to safer, more inclusive online environments.
Social Media – Hate Speech Detection – Disability – Natural Language Processing – Large Language Models – Prompt Engineering – Fine-Tuning – Retrieval-Augmented Generation – Knowledge Graph
Liya Wang, David Yi,Damien Jose,John Passarelli, James Gao, Jordan Leventis, and Kang Li, Atlassian, USA
Large Language Models (LLMs) enhance productivity through AI tools, yet existing benchmarks like Multitask Language Understanding (MMLU) inadequately assess enterprise-specific task complexities. We propose a 14-task framework grounded in Bloom’s Taxonomy to holistically evaluate LLM capabilities in enterprise contexts. To address challenges of noisy data and costly annotation, we develop a scalable pipeline combining LLM-as-a-Labeler, LLM-as-a-Judge, and corrective retrieval-augmented generation (CRAG), curating a robust 9,700-sample benchmark. Evaluation of six leading models shows open-source contenders like DeepSeek R1 rival proprietary models in reasoning tasks but lag in judgment-based scenarios, likely due to overthinking. Our benchmark reveals critical enterprise performance gaps and offers actionable insights for model optimization. This work provides enterprises a blueprint for tailored evaluations and advances practical LLM deployment.
Large Language Models (LLMs), Evaluation Benchmark, Bloom’s Taxonomy, LLM-as-a-Labeler, LLM-as-a-Judge, corrective retrieval-augmented generation (CRAG).
Philipp Seitz, Jan Schmitt, and Andreas Schiffler, Institute of Digital Engineering, Technical University of Applied Sciences W¨urzburg- Schweinfurt, Germany
For a larger set of predictions of several differently trained machine learning models, known as bagging predictors, the mean of all predictions is taken by default. Nevertheless, this proceeding can deviate from the actual ground truth in certain parameter regions. A method is presented to determine a representative value ˜yBS from such a set of predictions and to evaluate it by an associated quality criterion βBS, called Bagging Score (BS), using nonlinear regression with Neural Networks (NN). The BS reflects the confidence of the obtained ensemble prediction and also allows the construction of a prediction estimation function δ(β) for specifying deviations that are more precise than using the variance of the bagged predictors themselves.
Machine Learning, Neural Network, Bagging Predictors, Bagging Score, Nonlinear Regression, Deviation Estimation.
Hantao Wang1, Yu Cao2, 1USA, 2California State Polytechnic University, Pomona, CA, 91768
Traditional Q&A platforms are slow in responding to users’ questions, while AI responses are often unreliable and lack trustworthiness [1]. Meditrust aims to provide fast and reliable answers to users’ questions by incorporating AI in conjunction with manual Q&A and review. Meditrust contains a Q&A platform where users can post their questions and get answers. It also has an AI Chat page where people can obtain real-time responses by expressing their medical concerns and questions to a large language model [2]. To increase the trustworthiness of the app and the AI response, if the user has questions or concerns about the AI response generated, they can request a review of the content generated by medical experts. In order to assess the effectiveness of the app, we created a survey consisting of 10 questions with answers ranging from 1 (strongly disagree) to 5 (strongly agree) and sent the survey to 20 college students to obtain responses. The results of the survey proved that our app is indeed effective as it provides quick and reliable answers to users’ medical questions and receives positive feedback from users. The survey response also reveals that people do not generally trust the response of artificial intelligence and value human-to-human interaction for their medical questions and answers [4]. This finding further proves our app’s effectiveness, as our app allows users to request a review from human experts if they have concerns with AI-generated content. Furthermore, our app also provides a traditional Q&A platform for manual interaction on users’ questions and concerns. These features give Meditrust a unique edge compared to similar applications.
AI medical answers, expert review, fast Q&A, trusted health info
Jiale Zhao1, Soroush Mirzaee2, 1USA, 2California State Polytechnic University, Pomona, CA, 91768
This paper addresses the challenge of creating an affordable and effective guitar learning system. Traditional guitar learning methods rely heavily on teacher-student interaction, which can be limited in terms of feedback and accessibility [10]. To solve this problem, we propose a system that uses membrane buttons on the guitar fretboard to detect user input, combined with machine learning to provide real-time feedback and corrections. The system converts raw guitar signals into a readable format and integrates with an application to enhance the learning experience. Key technologies include RP2040 for signal conversion and machine learning for input analysis. Challenges such as signal accuracy and real-time feedback were addressed by using membrane buttons, which are more accurate and costeffective compared to other methods like video detection or audio analysis. The system was tested in various scenarios, demonstrating its potential to provide an interactive, accessible, and personalized guitar learning experience that can improve how students learn the instrument.
Adaptive, Assist, Guitar Learning, Music Creation, Machine Learning
Kaitlyn Wei1, Rodrigo Onate2, 1USA, 2California State Polytechnic University, Pomona, CA, 91768
The increased use of fast-fashion lately and the detrimental environmental impacts it causes is a very prevalent issue in society. Not only does this impact on the environment, but it means that the homeless are receiving poor quality clothing and live in areas full of waste. My program idea is an app that promotes sustainable clothing and donating clothes. The key technologies are authentication, digital closet, and the StyleMate AI. The digital closet is a neat and organized way for people to store their clothes and also receive an in-detail analysis for the eco-friendliness of the clothing. A ChatGPT API call is also used to maximize the capability of the app [11]. The experiment was performed on the AI, where a series of questions were asked and the results were rated based on the similarity of the actual response to the expected response. The donations page is an extremely helpful map that has markers placed on donation centers near a person’s location, and they can learn more just by clicking on the marker. My app is a fun way to organize clothes on a digital platform, but even more than that it promotes sustainable clothing and donating clothes to those in need.
Sustainable fashion, Digital closet, AI-powered app, Clothing donation
Iván Martínez-Murillo, Paloma Moreda, Elena Lloret, University of Alicante, Spain
This paper explores the influence of external knowledge integration in Natural Language Generation (NLG), focusing on a commonsense generation task. We examine how semantic relations from knowledge bases influence the generated text by creating a benchmark dataset that pairs input data with related retrieved knowledge and includes manually annotated outputs. Additionally, we conduct a detailed interpretability analysis to better understand these effects. By selectively removing relevant knowledge, we assess its impact on sentence quality and coherence. Our interpretability analysis shows that well-integrated external knowledge significantly enhances commonsense reasoning and concept coverage when generating a sentence. In contrast, filtering out key knowledge components leads to notable performance degradation, highlighting the critical role of relevant knowledge in guiding coherent generation. These findings underscore the value of interpretable, knowledge-enhanced NLG systems and call for evaluation frameworks that go beyond surface-level metrics to assess the underlying reasoning capabilities.
natural language generation, interpretability, knowledge-enhanced, commonsense generation.
Stepan Plotytsia, Delivery Manager, Grid Dynamics Holdings, Inc., Schaumburg, Illinois, USA
This research presents a comprehensive proof of concept (PoC) study comparing migration strategies from monolithic SAP Commerce platforms to cloud-native microservices architectures. I propose an innovative six-phase transformation methodology that incorporates composable commerce patterns, AI-driven personalization, and event-first integration. Through simulation modelling, theoretical analysis, and a comparative evaluation against alternative approaches (modular monolith, lift-and-shift, and phased SOA), I project potential improvements of 60% in latency reduction, 4× deployment frequency, and 282% ROI over five years. The study employs mixed-methods research, combining quantitative modelling with qualitative analysis of organizational readiness factors. This framework includes detailed architectural blueprints, comprehensive risk mitigation strategies, implementation roadmaps, and decision matrices validated through industry benchmarks and theoretical modelling. This research aims to provide organizations with data-driven insights and actionable guidance for evaluating modernization strategies before committing to full-scale transformation, addressing the critical gap in empirical comparison of migration approaches for enterprise e-commerce platforms.
Proof of concept, Migration strategy comparison, SAP Commerce modernization, Microservices architecture, ROI modelling, Risk assessment, Cloud transformation, Composable commerce, Event-driven architecture, Digital transformation
Jingbo Yang1, Garret Washburn2, 1USA, 2California State Polytechnic University, Pomona, CA, 91768
This paper presents an augmented reality (AR) system that overlays multimedia content on a Rubik-type cube using Vuforia Engine and Firebase services [1]. The system addresses the challenge of combining secure authentication, event-driven unlocking, and cloud-based content delivery. After user login through Firebase Authentication, cube interactions detected by Vuforia trigger the UnlockEventSystem, which updates Firestore to track progression [2]. Media files are retrieved from Firebase Storage and displayed in AR via Unity’s VideoPlayer and ImagePlayer managers [3]. Experiments tested tracking reliability under varying lighting conditions and media load times across network environments. Results demonstrated strong accuracy in normal settings and low latency on modern networks, though performance declined in poor lighting and weak connectivity. Methodology comparisons showed that while prior research identified AR’s educational potential, our work contributes a functional prototype that directly integrates progression, gamification, and cloud persistence. Ultimately, the project demonstrates a scalable, engaging, and secure AR framework for interactive learning and training.
Augmented Reality (AR), Vuforia Engine, Rubik’s Cube Tracking, Unity3D, Firebase Authentication.
Yutong Huo1, Moddwyn Andaya2, 1USA, 2California State Polytechnic University, Pomona, CA, 91768
This project introduces an AI-powered debate practice app designed to help users improve their public speakingand debate skills [9]. The app simulates real public forum debates by letting users input topics, choose roles, andengage with an AI opponent in various debate phases. It allows for flexible, on-demand practice and gives usersinstant feedback based on their choices [1]. To test the app’s effectiveness, five users participated in a survey afterusing it. The results showed an average score of 8.0 in both preparedness and confidence, proving the app helpedusers feel more ready and self-assured. The app also stands out when compared to other public speaking methods,such as therapy, solo prep, or structured classes [2]. Unlike those, this app offers an interactive, real-timeexperience that helps users practice impromptu responses under pressure. Overall, this tool provides a practicaland accessible way to build communication confidence and improve debate performance.
AI debate app, Speaking skills, Instant feedback, Confidence building.
Nasreddine Aoumeur1, Kamel Barkaoui2, Gunter Saake3, 1University of Science and Technology (USTO), Algeria, 2Laboratoire CEDRIC, CNAM, 292 Saint Martin, 75003 Paris - FRANCE, 3ITI, FIN, Otto-von-Guericke-Universit¨at Magdeburg, Germany
Whereas Artificial Intelligence (AI) with its Machine-Learning (ML) vertiginous advances are significantly reshaping our way of developing software either as (prompting) GenerativeAI- or purely ML-based ones, any significant influence of decades of investigations and findings around software-engineering (SE) concepts, principles and methods on such new AI-Era software is unfortunately almost desperately missing. The resulting is plethora of GenerativeAI- and MLbased software: Black-boxed rigid ill-conceptually and completely isolated from our ”ordinary” yet mostly disciplined software landscape. The aim of this paper is to contribute in leveraging such unsatisfactory Promptingand ML-based software form to be well-conceptually, dynamically adaptable by intrinsically fitting it within our ”ordinary” domain-oriented software landscape: We refer generically to as AI-Powered (knowledge-intensive) applications software; thereby reconciliating Domain- and AI-Experts instead of contemporarily miserable ’confrontation’. We achieve such promising endeavour by exactly capitalizing on best advanced SE concepts and principles More precisely, we are putting forward an innovative stepwise integrated modeldriven approach that smoothly exhibits the following conceptual, founded and technological milestones. Firstly, any structural features are semi-formally modelled as UML components intrinsically thereafter mapped into (ordinary and MLbased) Web-Services. Behavioural crucial features are then captured as intuitive business rules mostly at the inter-service interactions. Secondly, for the precise conceptualization of such inter-service behavioural rules, we are proposing tailored graphically appealing stereotyped primitives as ECA-driven architectural (interservice) connector glues, we refer to as ECA-driven interaction laws. Thirdly, for the rigorous certification, while staying ECA-Compliant we are tailoring Meseguer’s true-concurrent rewriting logic and its strategies-enabled Maude language for that purpose. Last but not least, for the efficient implementation we are proposing a four-level implementation still ECA-Compliant architecture, by relying on modern software technologies including python-empowered API with Django and its REST framework and Visual-Studio enterprise as advanced IDE. All approach milestones and steps are extensively illustrated using a quite realistic AI-powered software application dealing with Brain Tumor diagnostics while stressing all its benefits, with at-top reliability, dynamic-adaptability, self-learning and understandability.
ECA-driven architectural interaction laws, UML and Service-orientation, Machine-Learning (ML), KNN, Brain-Tumor, Reliability and Adaptability, RewritingLogic, Domain- and AI-Experts, Django REST and Python API.
Luz-Viviana Cobaleda-Estepa1, Julián Carvajal2, Paola Vallejo3, Andrés López4, Raúl Mazo5, 1Facultad de Ingeniería, Universidad de Antioquia, Medellín, Colombia, 2Facultad de Ingeniería, Universidad de Antioquia, Medellín, Colombia, 3Escuela de Ciencias Aplicadas e Ingeniería, Universidad EAFIT, Medellín, Colombia, 4Facultad de Ingeniería - Universidad de Antioquia, Medellín, Colombia, 5Lab-STICC, ENSTA, Brest, Francia.
Modern software systems increasingly integrate machine learning (ML) due to its advancements and ability to enhance data-driven decision-making. However, this integration introduces significant challenges for software engineering, especially in software product lines (SPLs), where managing variability and reuse becomes more complex with the inclusion of ML components. Although existing approaches have addressed variability management in SPLs and the integration of ML components in isolated systems, few have explored the intersection of both domains. Specifically, there is limited support for modeling and managing variability in SPLs that incorporate ML components. This article addresses this gap by proposing a structured framework that enhances SPL to support the inclusion of ML components. It facilitates the design of SPLs with ML capabilities by enabling systematic modeling of variability and reuse. The proposal has been partially implemented with the VariaMos tool.
Machine Learning (ML), Software Product Lines (SPL), ML-based systems, variability modeling.
Tanay Kamlesh Pate, Niraj Anil Babar, Deep Pujara, Glen Uehara, JeanLarson, Andreas Spanias, Arizona State University, Tempe, USA
This study describes the development of the Research Experience for Undergraduate students (REU) training program on Quantum Machine Learning (QML), hosted by the SenSIP Center at Arizona State University. In 2025, the REU hosted several projects that engaged quantum computing in signal processing, audio analysis, computer vision, medical diagnostics, anomaly detection, and generative machine learning. The objectives of the training program are to a) engage students in QML research by immersing them in government and industry projects, b) train students in quantum information processing and machine learning simulations, c) encourage students to pursue graduate research, d) increase awareness of career opportunities in QML, and e) provide professional development training. As part of professional development, students presented to stakeholders and received training in preparing publications, building awareness on social implications, ethics, and privacy. The program is evaluated by the Center for Evaluating the Research Pipeline (CERP) and an independent evaluator. This paper describes the importance of introducing QML research at the undergraduate level, recruitment, program structure, summaries of REU projects, and preliminary evaluations.
REU, Quantum Computing, Quantum Machine Learning, Qubits, Workforce Development
Yiyao Zhang1, Anne Yunsheng Zhang1, Jonathan Sahagun2, 1USA, 2California State Polytechnic University, Pomona, CA, 91768
Visual impairments affect millions worldwide, creating significant barriers to safe navigation and independence. Traditional tools such as white canes and guide dogs offer limited range and capabilities, underscoring the need for advanced assistive technologies. This project proposes a wearable smart vest that integrates four VL53L1X distance sensors, four DRV2605L haptic motors, and an ESP32-S3 Feather microcontroller with a PCA9548A multiplexer. The system delivers directional haptic cues and optional voice alerts through a mobile application. Experiments tested the vest under different lighting conditions and target angles, revealing high accuracy in most cases, though performance degraded in harsh sunlight and at steep angles. Compared with related methodologies, our design reduces reliance on auditory overload, external semantic data, or GPS connectivity, ensuring reliability in both indoor and outdoor settings [3]. Ultimately, the smart vest demonstrates a practical, user-centered solution that enhances safety, independence, and quality of life for visually impaired individuals.
Visually Impaired individual, Navigation, Bluetooth, Vest.
Osama S. Qatrani, Independent Researcher, UK
This paper introduces a symbolic governance model that maps how political authorities (P1–P2, L, I), capital (F, C), and layered technical infrastructures (T1–T4) translate directives into recursive algorithmic control. Rather than treating AI as an autonomous or neutral technology, the model reframes it as an encoded structure of power operating across data (D), social interfaces (S1–S2), and global arenas (G1–G3). By compressing complex system interactions into an accessible symbolic language, the framework helps non-technical stakeholders trace influence from policy and finance to code, platforms, and behavioral feedback (R). Brief case snapshots illustrate how algorithmic logics shape information, decisions, and public life. The contribution is a practical lens for diagnosing power in AI ecosystems and a policy-oriented roadmap for oversight, transparency in optimization targets, and public-interest safeguards against algorithmic domination.
Artificial Intelligence; Algorithmic Governance; Political Power; Digital Politics; Algorithmic Control.
Siyu Jiang1 , Rodrigo Onate2, 1USA, 2California State Polytechnic University, Pomona, CA, 91768
This paper discusses how childrens low emotional literacy showed as a major issue that traditional mental health services ignore and that schools fail to adequately address or educate young age groups to recognize. We suggest SelfGen, a multimedia mobile application that helps kids identify, categorize, and control their emotions by fusing hobbies from journaling, music, and art. Through gamified and imaginative activities, SelfGen promotes daily emotional check-ins through Firebase authentication, AI image generation, and real-time survey tracking [10]. Secured logins, moderated content, and adaptive feedback loops helped to address issues with community safety, privacy, and emotion recognition accuracy. SelfGens AI and survey system were tested in two experiments. The first demonstrated an accuracy of 70%+ in AI-generated emotional interpretations, and the second verified that user reflections and daily survey scores were strongly correlated. These findings suggest that SelfGen is capable of effectively recognizing emotional patterns and empowering users. SelfGen provides a dynamic, habit-forming substitute for strict school-based SEL programs by fusing security, creativity, and psychological insight—making emotional growth both interesting and approachable [11].
Psychology, Social Media, Teens, Data Analysis.
Le Chen1, Yu Cao2, 1USA, 2California State Polytechnic University, Pomona, CA, 91768
This research paper explores the development of Pet’s Mind, a mobile application designed to improve pet carethrough real-time AI assistance, health tracking, and community interaction [1]. The problem addressed is the lackof af ordable, comprehensive, and accessible tools for new pet owners, many of whom struggle to provide propercare due to limited knowledge or resources. Our methodology involved designing three core systems: the AI Nutrition Expert, the Health Tracking system, and the Community Forum. These systems were implemented usingFlutter and Firebase, ensuring accessibility across platforms [2]. To evaluate ef ectiveness, we conducted ausersurvey that tested navigation, usefulness, and satisfaction. The results showed high averages across most categories, particularly in usability and AI responses, with some room for improvement in design aesthetics and trust inAI accuracy. Overall, Pet’s Mind of ers a free, ad-free, and supportive platform that enables pet owners to makeinformed decisions and build healthier lives for their pets.
Pet Care, Mobile Application, AI Assistance, Health Tracking.
Anne Chen1, Ang Li2, 1USA, 2California State Polytechnic University, Pomona, CA, 91768
The inspiration for creating this program stemmed from seeing the large number of homeless individuals in my community, leading me to design an app that could provide essential support, raise awareness, and facilitate help. Seeing the prominent number of homeless people in my community inspired me to create this program so that they can have better support, spread awareness, and make it easier for other people to help [1]. The app is aimed at assisting the homeless population in San Diego by offering immediate access to resources such as food, medical, mental health services, and shelter [2]. It also serves as a platform for donations and volunteer opportunities for people looking to help combat homelessness. The app focuses on accessibility and simplicity, ensuring its easy for anyone to use by making it available on kiosks around San Diego and on the app store [3]. A key feature is the AI-powered chatbot, created using OpenAI, which helps address any specific questions or concerns that users may have beyond the standard resources. The app uses Firestore to manage a comprehensive database of locations and organizations, including contact details, hours of operation, and descriptions, which helps users select the most suitable support services [4]. The app also includes a map feature that guides users to nearby organizations, with locations sorted by proximity. To ensure the apps resources are up to date, it will be refreshed every two months to incorporate any changes. The app was designed with simplicity in mind, featuring straightforward buttons and clear instructions for ease of use, with the AI chatbot providing additional help for more complex questions. As part of the testing process, I created a usability survey with 10 questions focusing on interface, navigation, map functionality, and the chatbots effectiveness [5]. Results showed an average score of 4.04/5, with the highest rating given to the apps potential for volunteers and donors. However, the map feature received the lowest ratings, suggesting some users might struggle with it. Despite minor issues, the app is performing as intended, and feedback indicates it successfully meets its primary goal of encouraging people to download it, especially for those seeking to help or looking for support in homelessness [6].
Homelessness Support, Mobile Application, AI Chatbot, Volunteer and Donation Platform.