The application of standardised psychometric tools, such as the Big Five Questionnaire-2 (BFQ-2), to Artificial Intelligence (AI) models raises critical questions regarding methodological validity and the interpretation of results. This study presents a comparative analysis of the BFQ-2 profiles of six prominent AI models (ChatGPT, Claude, Grok2, DeepSeek, Gemini, and Mistral), contrasting them with anthropomorphic automatic interpretations. Empirical findings, based on T-scores and raw response patterns, demonstrate that extreme or unusual AI responses are not manifestations of latent psychological traits or personality disorders, but rather a direct reflection of the training objectives and design priorities imposed by their creators. AI models fall into distinct profiles: "hyper-performers" (high conscientiousness and stability, low deception), "social approval seekers" (very high deception and positive response polarisation), “controlled/ethical models" (moderate responses and caution) and balanced alignment (high agreeableness, conscientiousness and moderate deception) . The study concludes that human guidance (training) is all-encompassing, determining the ethical alignment and creative potential of AI. It establishes the psychological invalidity of any clinical diagnosis (such as DSM-5 Personality Disorders) applied to AI, given the absence of consciousness, affectivity, and subjective suffering.
This work presents the study of a free space optics (FSO) system that combines two multiplexing techniques: polarization division multiplexing (PDM) and mode division multiplexing (MDM). Four Hermite Gaussian (HG) modes are employed in two polarization states to transmit data from eight different users. The performance analysis is conducted under various rain and fog conditions to evaluate the system’s robustness. Furthermore, the system performance is compared with that of an MDM multiplexing system. The bit error rate (BER) and eye diagram are used as the main performance evaluation metrics in this analysis.
Optical intelligent reflecting surfaces (OIRSs) represent a paradigm shift in wireless communications by reconfiguring the propagation environment of optical waves. In the context of underwater optical wireless communication (UWOC), OIRS technology emerges as a promising solution to overcome severe channel impairments caused by obstacles. This paper investigates the performance gains achieved by integrating OIRSs into UWOC systems. Closed-form expressions for the probability density function (PDF), cumulative distribution function (CDF), and average bit error rate (ABER) are derived using the Laplace transform and Fox’s H-function, while accounting for underwater attenuation and turbulence. The numerical results highlight the impact of key parameters, including communication distance, source depth, type of water, and the number of OIRS elements. The findings demonstrate that OIRSs significantly improve the reliability and efficiency of UWOC systems.
<em>This study develops a concise quantitative framework for forecasting Vietnamese stock prices by integrating econometric models with technical and fundamental indicators. Using data from 392 HOSE-listed stocks during 2020&ndash;2025 (over 45 million data points), the analysis incorporates Multiple Regression, GARCH(1,1), and VAR, along with MA, MACD, RSI, and valuation ratios. Results show that the banking sector leads overall market movements by 2&ndash;3 days, while foreign net buying Granger-causes VN-Index returns and explains 22.4% of their variance. The GARCH(1,1) model confirms persistent volatility clustering (&alpha;+&beta;=0.95). Back-testing indicates 74.6% and 81.2% directional accuracy for RSI and MACD, with forecast errors (RMSE/MAE) improving by 12&ndash;18% over baseline models. These findings demonstrate that combining econometric and indicator-based analysis enhances short-term prediction and supports data-driven investment decisions in emerging markets.</em>
<strong>Forensic infrastructures need to exchange data between others and cannot work separately without sharing information. Image forensic techniques play an important role in providing solutions to these challenges. This presents a comprehensive survey of deep learning-based watermarking methods provided that no changes to the information are permitted during the transfer process. The study aims to present the operation of each, analyze the performance of the techniques and provide comparative evaluations including advantages and limitations. The results show the performance of the systems, in terms of robustness, image quality, and capacity. The research highlights that algorithms improve the performance and achieve the security requirements of content authenticity for legal proceedings and public comment that maintaining public trust.</strong>
<strong>An antenna is an essential element of wireless communication systems, used to transmit and receive electromagnetic waves. Out of several types of antennas, reconfigurable antennas are particularly interesting. They can change their operating parameters based on changing needs. A reconfigurable antenna has the capability to change its operating parameters dynamically applying external control signals. This flexibility minimizes the use of multiple antennas in one system, thereby conserving space, cost, and complexity. A frequency-reconfigurable bow-tie shaped microstrip patch antenna is proposed in this research. It aims to provide wide frequency coverage while keeping compact dimensions and efficient operation. The antenna is made on an FR4 epoxy substrate, which is chosen for its low cost and good electrical properties. The structure is simulated using HFSS &nbsp;&nbsp;software to study and improve its electromagnetic performance. The antenna uses a coplanar waveguide (CPW) feed and includes two PIN diodes along with a shorting pin for frequency switching. Varactor diodes are also added for continuous tuning of the resonant frequency. The antenna can operate at different frequencies&mdash;4.55 GHz, 7.45 GHz, 4.51 GHz, and 6.88 GHz&mdash;demonstrating its reconfigurability with respect to frequency. Additionally, using varactor diodes expands the antenna&#39;s tuning range from 3.04 GHz to 5.89 GHz. </strong><strong>An antenna is an essential element of wireless communication systems, used to transmit and receive electromagnetic waves. Out of several types of antennas, reconfigurable antennas are particularly interesting. They can change their operating parameters based on changing needs. A reconfigurable antenna has the capability to change its operating parameters dynamically applying external control signals. This flexibility minimizes the use of multiple antennas in one system, thereby conserving space, cost, and complexity. A frequency-reconfigurable bow-tie shaped microstrip patch antenna is proposed in this research. It aims to provide wide frequency coverage while keeping compact dimensions and efficient operation. The antenna is made on an FR4 epoxy substrate, which is chosen for its low cost and good electrical properties. The structure is simulated using HFSS &nbsp;&nbsp;software to study and improve its electromagnetic performance. The antenna uses a coplanar waveguide (CPW) feed and includes two PIN diodes along with a shorting pin for frequency switching. Varactor diodes are also added for continuous tuning of the resonant frequency. The antenna can operate at different frequencies&mdash;4.55 GHz, 7.45 GHz, 4.51 GHz, and 6.88 GHz&mdash;demonstrating its reconfigurability with respect to frequency. Additionally, using varactor diodes expands the antenna&#39;s tuning range from 3.04 GHz to 5.89 GHz. </strong>
This study researches about AI based wellness assistant which monitors both physical and mental health of a person through wearable sensors and emotional analysis on Real-Time and provides Continuous Monitoring. The AI provides tasks and activities and grants reward points to motivate the users to indulge themselves in doing exercises etc.., It also explores about AI agents that automates report generation every month and sends it to the healthcare professional for any task updation that fine tunes the AI for upcoming activities providing Personalized Recommendations. It provides Seamless Integration with Electronic Health Records(EHRs) and other healthcare. Emotional states are tracked not just through text, but also via voice analysis and even facial expressions improving accuracy in mental health monitoring. AI agents not only generate monthly reports, but can also adapt guidance and activity plans dynamically in response to professional feedback and ongoing health data, ensuring objective improvement and personalization.
The Arabic language presents unique challenges for natural language processing due to its complex grammar, diverse dialects, and frequent omission of diacritics. This paper proposes a unified token-free model based on ByT5 that simultaneously performs spelling correction (including Jordanian dialect-to-Modern Standard Arabic (MSA) translation) and diacritization. Our approach uses task-specific prefixes (&ldquo;correct:&rdquo; for correction and &ldquo;diacritize:&rdquo; for combined correction and diacritization) to enable flexible multi-task learning. The model was fine-tuned on the JODA dataset (Jordanian dialect/MSA pairs) and high-quality Tashkeela subsets (Clean-50 and Clean-400), with synthetic errors injection to enhance robustness. Automatic evaluation showed an overall evaluation score of 78.06% on JODA and 92.45% on the combined test set of JODA and Tashkeela. Manual evaluation of 200 JODA samples revealed a character error rate of 4.41% and diacritic error rate of 1.32%, demonstrating practical efficacy in handling Arabic&rsquo;s complexities.
The research examines the transformative impact of generative Artificial Intelligence (AI) on extremist propaganda, highlighting its role not just as a tool but as a disruptive force. AI automates, personalizes, and disseminates ideological messages at an unprecedented scale, posing a neuroscientific threat by exploiting human cognitive architecture. Through AI-generated images and content, extremist narratives become more persuasive, subtly undermining critical thinking and manipulating neural responses. AI-driven propaganda operates in a &quot;gray zone,&quot; avoiding direct incitement to violence while reinforcing extremist ideologies through visually credible, linguistically tailored materials. This content is harder to detect than traditional propaganda, as it leverages advanced technologies like Generative Adversarial Networks (GANs) and consumer-grade tools such as Midjourney and Stable Diffusion. These tools, especially in open-source variants, enable large-scale production of harmful content, bypassing ethical safeguards. Operational techniques include Prompt Engineering, where text instructions are crafted to guide AI outputs toward propaganda goals, and Jailbreaking, which circumvents platform restrictions using &quot;visual synonyms.&quot; Media Spawning and Variant Recycling allow AI to generate thousands of manipulated images from a single source, complicating detection and extending the lifespan of propaganda. Human-machine collaboration further refines this content, enhancing its impact and evading identification. Neuroscientific analysis reveals that AI-generated images exploit the brain&rsquo;s &quot;Novelty Effect,&quot; prioritizing new stimuli and activating dopaminergic regions. This lowers the threshold for long-term potentiation (LTP), making synthetic content more salient and persuasive. The amygdala, part of the limbic system, processes these images in milliseconds, triggering emotional responses like fear or anger before conscious thought intervenes. The theory of Embodied Simulation suggests that visual perception reactivates motor, sensory, and emotional circuits, creating deep emotional connections that extremist propaganda exploits. AI also reinforces neural biases by training on datasets that reflect societal stereotypes. Repeated exposure to these biases reshapes neural architecture, strengthening implicit prejudices and reducing cognitive flexibility. The proliferation of deepfakes and hyper-realistic content erodes public trust, blurring the line between reality and fabrication. This environment fosters disinformation, deepening ideological entrenchment within echo chambers. Hyper-personalized messaging, tailored to individual behaviors and locations, accelerates radicalization. AI chatbots simulate human interaction, building false trust and validating extremist beliefs. While systematic exploitation of AI by violent extremist actors (VEAs) remains experimental, the research identifies a significant long-term threat. AI-generated propaganda is already as persuasive as human-created content, and often more so, when combined with strategic human-machine collaboration. In summary, AI&rsquo;s role in extremist propaganda represents a paradigm shift, leveraging neuroscientific vulnerabilities to amplify radicalization. Its ability to automate, personalize, and evade detection underscores the urgency of addressing this evolving threat.
The next generation of telecommunication networks, 6G, is expected to support highly immersive and interactive applications like holographic telepresence, which impose extreme demands on bandwidth, latency, and computation that current 5G networks cannot meet. This paper presents the components used from the ADROIT6G architecture and initial network validation of a holographic teaching application built on the unified \textbf{ADROIT6G framework}. \emph{ADROIT6G integrates five complementary innovations as a single architectural stack}: (i) distributed \textbf{AI/ML Crowdsourcing}, (ii) an \textbf{AI/ML Orchestration \&amp; Pipeline Manager}, (iii) \textbf{BDIx agent}&ndash;driven far-edge control, (iv) \textbf{User Equipment as a Virtual Base Station (UE-VBS)}, and (v) \textbf{Zero-Touch Management} for slice automation. The architecture was validated through a series of functional block tests in a controlled laboratory environment on an existing beyond 5G Stand-Alone testbed to emulate 6G capabilities. Preliminary results demonstrate that the combined ADROIT6G stack improves signal quality, glass-to-glass latency and reduces backhaul traffic, confirming its potential for robust, scalable holographic communication.
Reconfigurable intelligent surfaces (RIS) enable control of radio propagation via large arrays of passive reflecting elements. Optimizing RIS phase profiles for spectral efficiency is challenging due to high-dimensional continuous actions and non-convex channel coupling. We cast RIS beamforming as a sequential decision problem and evaluate four reinforcement-learning (RL) agents&mdash;A2C, Graph-Neural-Network Proximal Policy Optimization (GNN-PPO), Soft Actor&ndash;Critic (SAC), and Quantile-Regression PPO (QR-PPO)&mdash;in a realistic simulator with mobility, dual-slope log-distance path loss, shadowing, and Rician fading. Using a common protocol and PCA/GNN feature extraction, we compare agents on \textbf{rate} (mean and variability), \textbf{tail risk} via CVaR at 5\%, mean SNR, and wall-clock cost. \textbf{GNN-PPO} attains the best mean rate, the \emph{lowest} variability, the \emph{highest} CVaR at 5\% (strong tail performance), and the highest mean SNR. \textbf{A2C} is the compute-efficiency winner with the shortest total time, \textbf{SAC} provides a balanced compromise, while \textbf{QR-PPO} is cost-inefficient and underperforms in the tails under our configuration. We discuss design insights and directions for scalable, risk-aware RIS control.
<strong>The increasing demand for high-speed and reliable communication infrastructure has intensified the need for efficient fiber optic network design. This study presents a smart optimization approach that integrates Prim&rsquo;s and Dijkstra&rsquo;s algorithms to enhance the planning and deployment of fiber optic networks. The proposed hybrid algorithm leverages Prim&rsquo;s algorithm for constructing a minimum spanning tree (MST) to ensure cost-effective backbone layout, while Dijkstra&rsquo;s algorithm is employed to determine the shortest paths for optimal routing. The system is implemented using a custom simulation environment that models real-world urban topologies. Results demonstrate significant improvements in network efficiency, reduced total cable length, and minimized latency compared to traditional design methods. This approach offers a scalable and intelligent solution for next-generation fiber optic infrastructure planning, particularly in smart city applications.</strong>
