Search Results (168)

With the increasing deployment of large language models (LLMs) in diverse applications, security vulnerability attacks pose significant risks, such as prompt injection. Despite growing awareness, structured, hands-on educational platforms for systematically studying these threats are lacking. In this study, we present an interactive training framework designed to teach, assess, and mitigate prompt injection attacks through a structured, challenge-based approach. The platform provides progressively complex scenarios that allow users to exploit and analyze LLM vulnerabilities using both rule-based adversarial testing and Open Worldwide Application Security Project-inspired methodologies, specifically focusing on the LLM01:2025 prompt injection risk. By integrating attack simulations and guided defensive mechanisms, this platform equips security professionals, artificial intelligence researchers, and educators to understand, detect, and prevent adversarial prompt manipulations. The platform highlights the effectiveness of experiential learning in AI security, emphasizing the need for robust defenses against evolving LLM threats. Full article

The rapid proliferation of Chat Generative Pre-trained Transformer (ChatGPT) marks a pivotal moment in artificial intelligence, eliciting responses from academic shock to industrial awe. As these technologies advance from passive tools toward proactive, agentic systems, their transformative potential and inherent risks are magnified globally. This paper presents a comprehensive, critical review of ChatGPT’s impact across five key domains: natural language understanding (NLU), content generation, knowledge discovery, education, and engineering. While ChatGPT demonstrates profound capabilities, significant challenges remain in factual accuracy, bias, and the inherent opacity of its reasoning—a core issue termed the “Black Box Conundrum”. To analyze these evolving dynamics and the implications of this shift toward autonomous agency, this review introduces a series of conceptual frameworks, each specifically designed to illuminate the complex interactions and trade-offs within these domains: the “Specialization vs. Generalization” tension in NLU; the “Quality–Scalability–Ethics Trilemma” in content creation; the “Pedagogical Adaptation Imperative” in education; and the emergence of “Human–LLM Cognitive Symbiosis” in engineering. The analysis reveals an urgent need for proactive adaptation across sectors. Educational paradigms must shift to cultivate higher-order cognitive skills, while professional practices (including practices within education sector) must evolve to treat AI as a cognitive partner, leveraging techniques like Retrieval-Augmented Generation (RAG) and sophisticated prompt engineering. Ultimately, this paper argues for an overarching “Ethical–Technical Co-evolution Imperative”, charting a forward-looking research agenda that intertwines technological innovation with vigorous ethical and methodological standards to ensure responsible AI development and integration. Ultimately, the analysis reveals that the challenges of factual accuracy, bias, and opacity are interconnected and acutely magnified by the emergence of agentic systems, demanding a unified, proactive approach to adaptation across all sectors. Full article

Emotion recognition capabilities in multimodal AI systems are crucial for developing culturally responsive educational technologies yet remain underexplored for Arabic language contexts, where culturally appropriate learning tools are critically needed. This study evaluated the emotion recognition performance of two advanced multimodal large language models, GPT-4o and Gemini 1.5 Pro, when processing Arabic children’s storybook illustrations. We assessed both models across three prompting strategies (zero-shot, few-shot, and chain-of-thought) using 75 images from seven Arabic storybooks, comparing model predictions with human annotations based on Plutchik’s emotional framework. GPT-4o consistently outperformed Gemini across all conditions, achieving the highest macro F1-score of 59% with chain-of-thought prompting compared to Gemini’s best performance of 43%. Error analysis revealed systematic misclassification patterns, with valence inversions accounting for 60.7% of errors, while both models struggled with culturally nuanced emotions and ambiguous narrative contexts. These findings highlight fundamental limitations in current models’ cultural understanding and emphasize the need for culturally sensitive training approaches to develop effective emotion-aware educational technologies for Arabic-speaking learners. Full article

Large language models (LLMs) have the potential to transform healthcare by assisting in documentation, diagnosis, patient communication, and medical education. However, their integration into clinical practice remains a challenge. This perspective explores the barriers to implementation by synthesizing recent evidence across five challenge domains: workflow misalignment and diagnostic safety, bias and equity, regulatory and legal governance, technical vulnerabilities such as hallucinations or data poisoning, and the preservation of patient trust and human connection. While the perspective focuses on barriers, LLM capabilities and mitigation strategies are advancing rapidly, raising the likelihood of near-term clinical impact. Drawing on recent empirical studies, we propose a framework for understanding the key technical, ethical, and practical challenges associated with deploying LLMs in clinical environments and provide directions for future research, governance, and responsible deployment. Full article

Online education has become an important channel for extensive, inclusive and flexible learning experiences. However, significant gaps persist in providing truly accessible, personalized and adaptable e-learning environments, especially for students with disabilities, varied language backgrounds, or limited bandwidth. This paper presents AccessiLearnAI, an AI-driven platform, which converges accessibility-first design, multi-format content delivery, advanced personalization, and Progressive Web App (PWA) offline capabilities. Our solution is compliant with semantic HTML5 and ARIA standards, and incorporates features such as automatic alt-text generation for images using Large Language Models (LLMs), real-time functionality for summarization, translation, and text-to-speech capabilities. The platform, built on top of a modular MVC and microservices-based architecture, also integrates robust security, GDPR-aligned data protection, and a human-in-the-loop to ensure the accuracy and reliability of AI-generated outputs. Early evaluations indicate that AccessiLearnAI improves engagement and learning outcomes across multiple ranges of users, suggesting that responsible AI and universal design can successfully coexist to bring equity through digital education. Full article

The influence of large language models (LLMs) has permeated education, too. We explored the opportunities and challenges of LLMs in mathematics teaching. In mathematics education, the generative nature of LLMs is appropriate for teachers as it enables an understanding of mathematical knowledge rather than students who lack discernment. Additionally, we combined programming languages with LLMs, using the example of geometric models, to integrate mathematics and visual representation in a new way. Through a comparison of problem-solving between ChatGPT and MathGPT and an analysis of their logical reasoning, teachers can exercise with large models as auxiliary tools to enhance the quality of mathematics teaching. Full article

This case study examines the integration of artificial intelligence (AI) into undergraduate architectural education through a 2024–25 core studio teaching experiment at Zhejiang University. A dual-module framework was implemented, comprising a 20 h AI skills training module and in-class ethics discussions, without altering the existing studio structure. The AI skills module introduced deep learning models, LLMs, AIGC image models, LoRA fine-tuning, and ComfyUI, supported by a dedicated technical instructor. Student feedback indicated phase-dependent and tool-sensitive engagement, and students expressed a preference for embedded ethical discussion within the design studio rather than separate formal instruction. The experiment demonstrated that modular AI education is both scalable and practical, highlighting the importance of phase-sensitive guidance, balanced technical and ethical framing, and institutional support such as cloud platforms and research-based AI tools. The integration enhanced students’ digital adaptability and strategic thinking while prompting reflection on issues such as authorship, algorithmic bias, and accountability in human–AI collaboration. These findings offer a replicable model for AI-integrated design pedagogy that balances technical training with critical awareness. Full article

The rapid advancement of AI technologies and the emergence of large language models (LLMs) such as ChatGPT have facilitated the integration of intelligent question-answering systems into education. However, students often hesitate to ask questions, which negatively affects learning outcomes. To address this issue, this study proposes a closed, locally deployed generative AI teaching assistant system that enables instructors to upload course PDFs to generate customized Q&A platforms. The system is based on a Retrieval-Augmented Generation (RAG) architecture and was developed through a comparative evaluation of components, including open-source large language models, embedding models, and vector databases to determine the optimal setup. The implementation integrates RAG with responsive web technologies and is evaluated using a standardized test question bank. Experimental results demonstrate that the system achieves an average answer accuracy of up to 86%, indicating a strong performance in an educational context. These findings suggest the feasibility of the system as an effective, privacy-preserving AI teaching aid, offering a scalable technical solution to improve digital learning in on-premise environments. Full article

General large language models (LLMs) often suffer from semantic misinterpretation, information redundancy, and hallucinated content when applied to educational question-answering tasks. These issues hinder their effectiveness in supporting students’ specialized course learning and self-directed study. To address these challenges, this study proposes an intelligent tutoring model that integrates a knowledge graph with a large language model (KG-CQ). Focusing on the Data Structures (C Language) course, the model constructs a course-specific knowledge graph stored in a Neo4j graph database. It incorporates modules for knowledge retrieval, domain-specific question answering, and knowledge extraction, forming a closed-loop system designed to enhance semantic comprehension and domain adaptability. A total of 30 students majoring in Educational Technology at H University were randomly assigned to either an experimental group or a control group, with 15 students in each. The experimental group utilized the KG-CQ model during the answering process, while the control group relied on traditional learning methods. A total of 1515 data points were collected. Experimental results show that the KG-CQ model performs well in both answer accuracy and domain relevance, accompanied by high levels of student satisfaction. The model effectively promotes self-directed learning and provides a valuable reference for the development of knowledge-enhanced question-answering systems in educational settings. Full article

This review provides a comprehensive overview of the evolution and application of artificial intelligence (AI) and large language models (LLMs) in engineering, with a specific focus on chemical engineering. The review traces the historical development of LLMs, from early rule-based systems and statistical models like N-grams to the transformative introduction of neural networks and transformer architecture. It examines the pivotal role of models like BERT and the GPT series in advancing natural language processing and enabling sophisticated applications across various engineering disciplines. For example, GPT-3 (175B parameters) demonstrates up to 87.7% accuracy in structured information extraction, while GPT-4 introduces multimodal reasoning with estimated token limits exceeding 32k. The review synthesizes recent research on the use of LLMs in software, mechanical, civil, and electrical engineering, highlighting their impact on automation, design, and decision-making. A significant portion is dedicated to the burgeoning applications of LLMs in chemical engineering, including their use as educational tools, process simulation and modelling, reaction optimization, and molecular design. The review delves into specific case studies on distillation column and reactor design, showcasing how LLMs can assist in generating initial parameters and optimizing processes while also underscoring the necessity of validating their outputs against traditional methods. Finally, the review addresses the challenges and future considerations of integrating LLMs into engineering workflows, emphasizing the need for domain-specific adaptations, ethical guidelines, and robust validation frameworks. Full article

Student behavior analysis plays a critical role in enhancing educational quality and enabling personalized learning. While previous studies have utilized machine learning models to analyze campus card consumption data, few have integrated multi-source behavioral data with large language models (LLMs) to provide deeper insights. This study proposes an intelligent educational system that examines the relationship between student consumption behavior and academic performance. The system is built upon a dataset collected from students of three majors at Xinjiang Normal University, containing exam scores and campus card transaction records. We designed an artificial intelligence (AI) agent that incorporates LLMs, SageGNN-based graph embeddings, and time-series regularity analysis to generate individualized behavior reports. Experimental evaluations demonstrate that the system effectively captures both temporal consumption patterns and academic fluctuations, offering interpretable and accurate outputs. Compared to baseline LLMs, our model achieves lower perplexity while maintaining high report consistency. The system supports early identification of potential learning risks and enables data-driven decision-making for educational interventions. Furthermore, the constructed multi-source dataset serves as a valuable resource for advancing research in educational data mining, behavioral analytics, and intelligent tutoring systems. Full article

As large language model (LLM) tutors evolve from scripted helpers into adaptive educational partners, their capacity for self-regulation, ethical decision-making, and internal monitoring will become increasingly critical. This paper introduces the Needs-Driven Consciousness Framework (NDCF) as a novel, integrative architecture that combines Dennett’s multiple drafts model, Damasio’s somatic marker hypothesis, and Tulving’s tripartite memory system into a unified motivational design for synthetic consciousness. The NDCF defines three core regulators, specifically Survive (system stability and safety), Thrive (autonomy, competence, relatedness), and Excel (creativity, ethical reasoning, long-term purpose). In addition, there is a proposed supervisory Protect layer that detects value drift and overrides unsafe behaviours. The core regulators compute internal need satisfaction states and urgency gradients, feeding into a softmax-based control system for context-sensitive action selection. The framework proposes measurable internal signals (e.g., utility gradients, conflict intensity Ω), behavioural signatures (e.g., metacognitive prompts, pedagogical shifts), and three falsifiable predictions for educational AI testbeds. By embedding these layered needs directly into AI governance, the NDCF offers (i) a psychologically and biologically grounded model of emergent machine consciousness, (ii) a practical approach to building empathetic, self-regulating AI tutors, and (iii) a testable platform for comparing competing consciousness theories through implementation. Ultimately, the NDCF provides a path toward the development of AI tutors that are capable of transparent reasoning, dynamic adaptation, and meaningful human-like relationships, while maintaining safety, ethical coherence, and long-term alignment with human well-being. Full article

With the accelerated digital transformation in education, the efficient integration of massive multimodal instructional resources and the support for interactive question answering (QA) remains a prominent challenge. This study introduces Multimodal Disciplinary Knowledge-Augmented Generation (MDKAG), a framework integrating retrieval-augmented generation (RAG) with a multimodal disciplinary knowledge graph (MDKG). MDKAG first extracts high-precision entities from digital textbooks, lecture slides, and classroom videos by using the Enhanced Representation through Knowledge Integration 3.0 (ERNIE 3.0) model and then links them into a graph that supports fine-grained retrieval. At inference time, the framework retrieves graph-adjacent passages, integrates multimodal data, and feeds them into a large language model (LLM) to generate context-aligned answers. An answer-verification module checks semantic overlap and entity coverage to filter hallucinations and triggers incremental graph updates when new concepts appear. Experiments on three university courses show that MDKAG reduces hallucination rates by up to 23% and increases answer accuracy by 11% over text-only RAG and knowledge-augmented generation (KAG) baselines, demonstrating strong adaptability across subject domains. The results indicate that MDKAG offers an effective route for scalable knowledge organization and reliable interactive QA in education. Full article

Background: Dentistry is undergoing a digital transformation driven by emerging technologies such as Artificial Intelligence (AI), Internet of Things (IoT), Digital Twins (DTs), and Large Language Models (LLMs). These advancements offer new paradigms in clinical diagnostics, patient monitoring, treatment planning, and medical education. However, integrating these technologies also raises critical questions around security, privacy, ethics, and trust. Objective: This review aims to provide a structured synthesis of the recent literature exploring AI, IoT, DTs, and LLMs in dentistry, with a specific focus on their application domains and the associated ethical, privacy, and security concerns. Methods: A comprehensive literature search was conducted across PubMed, IEEE Xplore, and SpringerLink using a custom Boolean query string targeting publications from 2020 to 2025. Articles were screened based on defined inclusion and exclusion criteria. In total, 146 peer-reviewed articles and 18 technology platforms were selected. Each article was critically evaluated and categorized by technology domain, application type, evaluation metrics, and ethical considerations. Results: AI-based diagnostic systems and LLM-driven patient support tools were the most prominent technologies, primarily applied in image analysis, decision-making, and health communication. While numerous studies reported high performance, significant methodological gaps exist in evaluation design, sample size, and real-world validation. Ethical and privacy concerns were mentioned frequently, but were substantively addressed in only a few works. Notably, IoT and Digital Twin implementations remained largely conceptual or in pilot stages, highlighting a technology gap in dental deployment. Conclusions: The review identifies significant potential for converged intelligent dental systems but also reveals gaps in integration, security, ethical frameworks, and clinical validation. Future work must prioritize cross-disciplinary development, transparency, and regulatory alignment to realize responsible and patient-centered digital transformation in dentistry. Full article

Educational assessment relies on well-constructed test items to measure student learning accurately, yet traditional item development is time-consuming and demands specialized psychometric expertise. Automatic item generation (AIG) offers template-based scalability, and recent large language model (LLM) advances promise to democratize item creation. However, fully automated approaches risk introducing factual errors, bias, and uneven difficulty. To address these challenges, we propose and evaluate a hybrid human-in-the-loop (HITL) framework for AIG that combines psychometric rigor with the linguistic flexibility of LLMs. In a Spring 2025 case study at Franklin University Switzerland, the instructor collaborated with ChatGPT (o4-mini-high) to generate parallel exam variants for two undergraduate business courses: Quantitative Reasoning and Data Mining. The instructor began by defining “radical” and “incidental” parameters to guide the model. Through iterative cycles of prompt, review, and refinement, the instructor validated content accuracy, calibrated difficulty, and mitigated bias. All interactions (including prompt templates, AI outputs, and human edits) were systematically documented, creating a transparent audit trail. Our findings demonstrate that a HITL approach to AIG can produce diverse, psychometrically equivalent exam forms with reduced development time, while preserving item validity and fairness, and potentially reducing cheating. This offers a replicable pathway for harnessing LLMs in educational measurement without sacrificing quality, equity, or accountability. Full article