Search Results (138)

We present ESNformer, a hybrid architecture that couples an Echo State Network (ESN) reservoir with a Transformer encoder for classification of structured, multi-indicator assessment data: a fixed-order vector of complementary indicators per assessment instance rather than a repeated-measures time series. The reservoir acts as a fixed nonlinear feature map over the indicator vector, while self-attention, made position-aware over the fixed column order, learns how each indicator’s evidence contributes to the final decision, so the two components, together, capture local, indicator-level detail and global, cross-indicator interactions within a single, end-to-end trainable model. Interpretability is treated as a first-class design requirement rather than an afterthought: the architecture is paired with an explainability layer combining SHAP feature attribution (reported both globally and per class), the model’s own attention weights, a deletion/insertion faithfulness test that quantitatively verifies which inputs the model actually relies on, and counterfactual maps that translate a prediction into an actionable, inspectable recommendation. We evaluate the architecture on a concrete case study, classifying Braille-literacy instructional recommendations from 15 pedagogical indicators grouped into three categories (Mangold’s, ABKL, and Progresar), using a benchmark of 900 real assessment instances (630 used, together with a class-conditional augmentation procedure, to build a 2100-instance training set) with validation and test partitions (135 instances each) kept exclusively real. On this benchmark, the tuned model reached 85.33% accuracy, 85.90% macro-precision, 85.33% macro-recall, an F1 score of 85.25%, and an AUC of 0.95 on the real test set. SHAP attribution, attention weights, and the faithfulness test converge on the same two dominant indicators (response time and error count): removing them alone collapses accuracy to chance, while retaining only them recovers most of the model’s accuracy. We report this transparently alongside a comparison against ESN-only, Transformer-only, and tabular baselines (logistic regression, decision tree, random forest, XGBoost, and an MLP) on the same data and discuss what the hybrid architecture and its explainability pipeline add beyond what the two dominant indicators already explain and how the approach generalizes to other tabular and mixed-granularity assessment settings that require both predictive accuracy and a verifiable account of what drove each decision. Full article

Large language models (LLMs) now generate fluent natural language and source code, creating challenges for authorship attribution, academic integrity, and software supply-chain security. Most existing detectors for AI-generated content are evaluated separately on natural language or source code, often under matched train–test conditions that can overestimate real-world reliability. We present a paired-prompt benchmark for human-versus-machine detection across English text, Python code, and mixed text–code documents. The benchmark includes $\mathrm{22,141}$ instances from HC3, CodeSearchNet, MBPP, and HumanEval across training, validation, and test partitions, plus Mix-Eval, a mixed-content set of 997 Jupyter-notebook-style samples. We evaluate RoBERTa-large for text, GraphCodeBERT and CodeBERT-base for code, a unified RoBERTa-base detector trained on both modalities, and zero-shot baselines. Fine-tuned detectors achieve near-perfect in-distribution performance, with AUROC $1.0000\pm 0.0000$ and accuracy above $99.5\%$. Across five instruction-tuned generator families of varying size (3.8B–7B) and architecture, with the human and problem distributions held fixed, cross-generator transfer causes negligible degradation (AUROC spread $0.0002$; drops of at most $0.0003$). In contrast, domain shift is the main failure mode: on MBPP+HumanEval, GraphCodeBERT drops to $0.85\pm 0.02$ AUROC and CodeBERT-base to $0.67\pm 0.02$. On Mix-Eval, the unified detector outperforms a routed text–code pipeline by 21 AUROC points ($0.96$ vs. $0.75$), largely because of router failures on mixed inputs. Training-time augmentation improves low-false-positive performance, while legacy supervised detectors show systematic class inversion on modern LLM outputs. These results show that reliable deployment requires cross-domain evaluation, mixed-content testing, and calibration beyond in-distribution accuracy. Full article

Ambient Assisted Living (AAL) environments aim to enable older people to remain active and lead an autonomous and independent life for as long as possible. Among the technologies that can be incorporated into these settings, socially assistive robots (SAR) seek to establish a more natural and intuitive means of interaction with people, whilst helping them to carry out everyday tasks. One of the main challenges facing the design of these robots is how to enable them to undertake more complex tasks. Recent advances in Large Language Models (LLMs) have opened new avenues for flexible robot deliberation, yet their integration into real-time robotic systems remains challenging due to latency constraints, reasoning reliability, and the complexity of coordinating multi-step tasks. This paper proposes a hierarchical multi-agent architecture for robot deliberation that addresses these challenges by combining LLM-based planning with structured execution mechanisms within the ROS 2 ecosystem. The proposed architecture employs a supervisor agent that decomposes high-level natural language instructions into prioritised subtasks, enabling a priority-driven execution model that dynamically adapts to task relevance, temporal constraints, and environmental feedback. Subtasks are delegated to a set of Single-Purpose Agents (SPAs), orchestrated via LangGraph state machines and coordinated through a priority-aware scheduling mechanism. A key design principle is the use of Behaviour Trees (BTs) as high-level callable tools through the Model Context Protocol (MCP), encapsulating closed-loop control strategies while enabling preemptive and priority-consistent execution. This reduces the number of LLM inference steps required per task and improves robustness under dynamic conditions. A further contribution concerns the deployment of fine-tuned, lightweight LLMs—on the order of 0.6 billion parameters—specifically adapted for both the supervisor and the individual SPA roles through parameter-efficient low-rank adaptation (LoRA). These models are trained on role-specific tool-calling datasets to specialise in constrained reasoning patterns and task-specific decision-making, enabling efficient, low-latency inference directly on edge hardware. The combination of fine-tuning and hierarchical priority control enhances both the determinism and responsiveness of the system while mitigating error propagation across agent interactions. The paper presents the full software architecture, a formal characterisation of the system as a priority-aware hierarchical policy over a graph of agent workflows, and an experimental evaluation in an Ambient Assisted Living scenario assessing task success rate, inference efficiency, responsiveness under competing priorities, and overall user experience. Because SPA execution is decoupled from the supervisor’s own reasoning loop, the architecture is designed to keep accepting, processing, and queuing new user queries while previously dispatched SPAs are still executing their tasks. Full article

The growing demand for scalable and cost-effective digital learning environments has increased interest in lightweight cloud-based learning management solutions that can support instructional delivery without the complexity and infrastructure requirements of conventional Learning Management Systems (LMSs). While enterprise LMS platforms provide extensive functionality, their deployment and maintenance often require substantial technical, financial, and administrative resources. This study proposes a lightweight cloud-based LMS architecture using low-code web technologies as an alternative framework for educational content management, learner engagement, resource distribution, and instructional support. The proposed architecture integrates four functional system layers: course management, performance management, content delivery, and productivity support. These components are designed to operate within a cloud-hosted environment that leverages web-based content management, embedded digital resources, collaborative productivity tools, and centralized storage services. The architecture emphasizes accessibility, modularity, low deployment overhead, and cross-platform compatibility, enabling rapid implementation in resource-constrained educational settings. To evaluate the feasibility of the proposed architecture, a large-scale deployment was conducted involving 1765 end users interacting with the platform within an educational environment. System functionality was assessed through feature-level evaluation across the core LMS components and supported by user capability indicators related to digital engagement and platform utilization. Analytical results demonstrated strong functional performance across all architectural modules, with course management and content delivery components exhibiting the highest operational effectiveness. Findings further indicated that the cloud-based architecture successfully supported essential LMS functions through integrated web services and low-code platform technologies. The study contributes a replicable systems architecture and deployment framework for lightweight learning management environments. The proposed model offers a practical foundation for the development of cloud-based educational platforms that support scalable content delivery, learner interaction, instructional management, and future integration with learning analytics, adaptive learning engines, and intelligent educational support systems. The framework provides an engineering-oriented approach for designing accessible and sustainable digital learning infrastructures using low-code technologies. Full article

The proliferation of Edge-AI in personalized healthcare has driven significant demand for energy-efficient Systems-on-Chip (SoCs) capable of executing high-accuracy, real-time disease detection. Conventional accelerator architectures struggle to simultaneously accommodate disparate AI models. Specifically, memory-intensive Convolutional Neural Networks (CNNs) for high-fidelity feature extraction and event-driven Spiking Neural Networks (SNNs) for ultra-low-power, brain-inspired computation. To address this bottleneck, this paper presents a comprehensive survey of Reconfigurable Hybrid Neural Networks (RHNNs), an emerging paradigm that dynamically merges the strengths of CNNs and SNNs to meet the stringent resource constraints of biomedical edge devices. We establish a comprehensive taxonomy of existing RHNN architectures, categorizing them by hardware interconnection topologies, dataflow orchestration strategies, and internal structural adaptation mechanisms. Furthermore, we examine the integration of these hybrid accelerators within the open-source RISC-V processor ecosystem, evaluating how custom instruction set extensions optimize control efficiency and minimize energy overhead. The survey also analyzes commonly used datasets based on three major biomedical signal modalities, including electroencephalography (EEG), electrocardiography (ECG), and electromyography (EMG), in the context of processing systems for hardware accelerators. Finally, we highlight the open research challenges and outline future research directions to guide the development of next-generation biomedical intelligent systems. Full article

Application-specific instruction set processors (ASIPs) often employ specialized hardware to improve performance, but this introduces complexity in resource management and programming. Existing compiler solutions, including LLVM’s default schedulers, lack fine-grained structural conflict analysis for complex arithmetic logic unit (ALU) instructions, leading to suboptimal performance or runtime errors. This limitation becomes critical when targeting very long instruction word (VLIW) architectures with instruction fusion units that exhibit pipeline-stage-level resource contention. In this paper, we propose a compile-time instruction scheduling method that models sub-cycle resource usage and analyzes both data and structural dependencies at fine granularity. Unlike coarse-grained resource tables used in existing compilers, our approach tracks functional unit occupancy at the pipeline stage level, enabling precise detection of structural hazards in complex execution units. We implement this scheduler as a backend pass in the LLVM compiler framework and validate it on the Sayram VLIW processor for wireless communication. Experimental results show that our approach achieves 100% scheduling correctness while improving execution efficiency by 23% on average compared with in-order scheduling, with benefits up to 38% for highly parallel kernels such as PRACH, and reducing average running time by 66% compared with atomic execution. Full article

This paper presents a quantitative evaluation of an automatic translation system between industrial robot programming languages based on a sequence-to-sequence (Seq2Seq) neural architecture using Long Short-Term Memory (LSTM) networks. The study addresses the interoperability problem between proprietary robot programming languages by proposing a data-driven approach capable of learning correspondences between structured code instructions. A parallel dataset of 28,000 aligned instruction pairs was constructed and preprocessed through tokenization and normalization to enable structured sequence learning. The model was trained under four configurations (50,100, 150 and 200 epochs) to analyze the impact of training duration on performance and generalization capability. The system was evaluated using multiple quantitative metrics, including accuracy, loss, BLEU, and Exact Match (EM), allowing assessment of both structural similarity and exact sequence correctness. Experimental results demonstrate that the 200-epoch configuration improves the performance across all metrics, achieving an accuracy of 0.9943, a BLEU score of 0.682, and an Exact Match of 0.970 on the test set. These results indicate that the model is capable of generating both structurally consistent and syntactically correct translations. The analysis shows that while BLEU captures structural similarity, EM provides a stricter evaluation of exact sequence correctness, which is critical in structured code translation tasks where minor variations may affect execution. The proposed approach demonstrates the feasibility of applying neural machine translation techniques to industrial robot programming, contributing to improved interoperability and reduced manual effort in multi-platform robotic environments. Full article

Low-Rank Adaptation (LoRA) is a representative parameter-efficient fine-tuning method that reduces computational and memory costs without modifying the model architecture. Standard LoRA initializes matrix A from a symmetric distribution, such as Gaussian or Kaiming initialization, and matrix B to zero. Although this provides a statistically neutral starting point, it ignores the influence of task-specific input features on initialization. We propose Gamma-Moment Equalization Initialization (GME-Init), a data-aware asymmetric LoRA initialization method based on output-moment calibration. Using a small calibration set, GME-Init estimates the variance and skewness of target-layer outputs and adjusts the layer-wise initialization scale and asymmetry of LoRA weights, improving their statistical alignment with task-specific skewed representations. GME-Init operates only during initialization and does not change the LoRA architecture, trainable parameter count, training budget, or inference cost. We evaluate it on a GLUE subset with RoBERTa-base, integrate it with AdaLoRA and DoRA, and test it on VRSBench-VQA, VRSBench-Caption, and UCM-Caption using Qwen2.5-VL-3B-Instruct. Results show that GME-Init serves as a simple plug-in PEFT initialization module with no additional inference cost and that it consistently improves standard LoRA and selected LoRA-style methods across the evaluated text understanding and multimodal tasks. Full article

This paper proposes a lightweight adaptive knowledge graph tutoring system for discrete mathematics, addressing the challenges of knowledge fragmentation, homogenized pedagogical paradigms, and insufficient practical support in traditional instruction. The system formalizes disciplinary knowledge points and associated exercise repositories, constructing an integrated architecture that unifies knowledge graph representation, intelligent recommendation, adaptive assessment, and content delivery. By integrating lightweight cognitive assessment models with heuristic recommendation strategies, an adaptive learning path generation mechanism is built to dynamically quantify learner proficiency based on problem solving interaction traces. Diverging from compute-intensive deep learning paradigms, this method requires no large-scale training, making it convenient for users to adjust parameters according to actual conditions. It preserves the interpretability of learning trajectories and real-time responsiveness while enabling seamless deployment across conventional computing environments. Experiments in authentic educational settings indicate that the system facilitates learners’ problem solving performance and learning efficiency, offering a scalable and cost-effective solution for personalized supplementary instruction in discrete mathematics and broader undergraduate foundational mathematics courses. Full article

Text-to-Cypher generator systems translate natural language questions into Cypher queries, enabling intuitive interactions with graph databases such as Neo4j and Amazon Neptune. Despite recent advancements in LLM-based Cypher query generation, the vulnerabilities of the known methods—such as prompt injection attacks—are not discussed in detail. In this paper, we employ a robust Retrieval-Augmented Generation (RAG) architecture tailored specifically for text-to-Cypher tasks, leveraging dense vector retrieval to enhance query generation accuracy. We propose a dynamic and self-corrective procedure with feedback-loop-based AI architecture with Large Language Models (LLMs) for near real-time validation and correction of generated queries. We create a systematic procedure for generating datasets specifically designed to assess prompt injection robustness. Comprehensive evaluations are conducted using a diverse set of LLMs, including GPT-4o, DeepSeek R1, Claude 3.5 Sonnet and Qwen 2.5 Coder 32B Instruct. Our evaluation results indicate substantial improvements in resiliency against prompt injection attacks compared to various benchmarks. It is demonstrated that the proposed solution outperforms various training-free prompt injection defense methods. Full article

Background: Preoperative screening for methicillin-resistant Staphylococcus aureus (MRSA) and methicillin-susceptible S. aureus (MSSA) is intended to identify patients at increased risk of surgical site infection and guide decolonization and perioperative antimicrobial prophylaxis. However, many molecular assays reduce this decision to a binary positive/negative result, potentially obscuring clinically important distinctions in organism identity, methicillin resistance attribution, and mupirocin resistance risk. Methods: This structured narrative review organized direct perioperative evidence and indirect mechanistic, implementation, and economic evidence around one question: how MRSA/MSSA screening can move from organism detection to actionable molecular triage. Results: Useful preoperative reporting depends on assigning resistance markers to the correct organism. The proposed multi-target NAAT framework organizes concordant and discordant molecular patterns into provisional reportable categories, including MSSA, MRSA, methicillin-resistant non-aureus Staphylococcus/CoNS, mixed populations, SCCmec dropout patterns, mupirocin resistance marker states, and invalid or indeterminate results. No externally validated composite score, universal molecular cutoff, or prospectively validated target-to-action decision rule currently links all of these categories to specific perioperative actions. Conclusions: Preoperative MRSA/MSSA screening may benefit from moving beyond binary reporting, but the framework presented here is a development-stage rule set rather than a validated clinical decision instrument. Assay-specific analytical thresholds, locked target combination rules, and prospective clinical and implementation validation are required before the framework can be used to assign patients reproducibly to management pathways. Until such validation is completed, the proposed categories should be interpreted as a testable reporting and validation architecture rather than as universal prophylaxis or decolonization instructions. Full article

Background: Modern household appliances often present complex interfaces that can be difficult to use, especially for older adults, people with visual impairments, and users with mild cognitive difficulties. In such cases, interacting with appliances may require sustained attention, visuospatial search, working memory, and sequential action planning, while traditional user manuals often provide limited contextual support. Methods: To address this issue, this study presents a proof-of-concept augmented reality (AR) and artificial intelligence (AI)-based system for contextual appliance guidance. The proposed architecture integrates visual sensing, deep learning, and large language models to detect appliance controls, interpret user queries, retrieve relevant information from user manuals, and provide step-by-step guidance directly on the real interface. A YOLOv8 model trained on a custom dataset was used for button detection, YOLO-Seg was employed to enhance visual highlighting through segmentation, and BoT-SORT was used to maintain detection consistency across frames. A Unity-based mobile application displayed real-time AR overlays with customizable visual settings for accessibility needs, such as low vision and color blindness that may be relevant for future accessibility-oriented applications. In addition to its technical pipeline, the system is conceptually relevant as a potential form of external cognitive support because it transforms static manual instructions into situated, sequential, and visually grounded guidance.Results: Experimental results showed promising technical performance for button detection and segmentation, while a preliminary user evaluation in a non-clinical sample suggested good usability, clarity, and acceptability of the interface. Conclusions: These findings support the technical feasibility and preliminary usability of the approach, while cognitive workload, confidence, functional autonomy, and clinical benefit were not directly measured. Targeted validation in older adults, people with visual impairments, and clinical populations is therefore still needed. Future developments will include multimodal feedback, read-aloud guidance, and more specific evaluation of workload, confidence, and functional autonomy. Full article

The open RISC-V Instruction Set Architecture (ISA) is a versatile architecture with a growing number of implementations, including the aerospace sector, where tolerance to Single-Event Upset (SEU) faults is critical. This paper presents the FPGA-based design and implementation of an SEU-tolerant cache memory for integration into RISC-V cores as part of a memory subsystem protection strategy. The proposed architecture incorporates separate instruction and data caches, organized as four-way set-associative with a write-through policy, complemented by Hamming H(39, 32) encoding for single-bit error detection and correction. Two experimental platforms were developed on a ZYNQ-7000 SoC: one dedicated to functional validation of the Hamming modules through controlled error injection and another for the evaluation of the complete system integrated with an internally developed RV32IMAFE core. The results confirm the correction of single errors and the detection of multiple errors while demonstrating performance improvements between 1.42× and 1.67× compared to the baseline RISC-V core. Full article

This conceptual paper proposes Math-CEFR, a reference framework for describing postsecondary mathematical readiness in the context of remedial and gateway placement. Single-score placement tests frequently misclassify students and offer little guidance for instruction. Drawing on the architecture of the Common European Framework of Reference for Languages (CEFR), Math-CEFR separates global decision bands (M0–M5) from a multi-dimensional diagnostic profile covering procedural fluency, conceptual understanding, structural awareness, and transfer. The framework treats readiness as non-linear and pathway-sensitive, distinguishing two common course pathways in postsecondary mathematics: a quantitative reasoning/statistics route and an algebra-to-calculus route. It is accompanied by illustrative descriptors, task-family specifications, and an alignment and local standard-setting protocol. Because no validity evidence yet exists, transparency and defensibility are design goals whose attainment depends on the validation agenda outlined here—they are not established properties of the framework. Math-CEFR is offered as a structural proposal intended to make placement interpretations more transparent and future empirical evaluation feasible, rather than as a finished testing instrument. Full article

In the context of industrial digitalization and the widespread adoption of process automation systems, Knowledge Management Systems (KMS) play a key role in providing operational personnel with up-to-date instructions and regulations. However, the inherent ambiguity of natural language in technical documentation remains a serious obstacle, leading to incorrect operator actions, process deviations, and increased safety risks. This article investigates the integration of Large Language Models (LLMs) into KMS and its impact on user experience and human–machine interaction in industrial automation environments. A method called Semantic Latent Choice Detection is presented, designed to systematically identify interpretation ambiguities in process instructions and operator commands. Unlike existing approaches that require access to the internal model architecture (“white box”) or token-level logits, the proposed method is logit-free and operates with closed commercial LLMs (“black box”) via standard API interfaces. The method analyzes the semantic similarity of binary text blocks and polysemous terms within the context of a specific technological process. Using a metallurgical production case study, we demonstrate how the system detects hidden semantic collisions (e.g., the difference between “adding ferroalloys into the ladle” and “feeding ferroalloys onto the conveyor”) that are missed by traditional rule-based validation methods. Instead of arbitrarily selecting an interpretation, the system initiates a clarification request to the human operator, thereby reducing cognitive load, preventing erroneous automated decisions, and increasing trust in the KMS. An empirical evaluation conducted in a real-world industrial setting (unit control rooms and dispatch centers) shows a statistically significant reduction in errors related to misinterpretation of process regulations. The article contributes to the fields of automation engineering, knowledge management, and human-centered automation by proposing a novel method for validating operational instructions in high-risk industrial environments. Full article