Muhammad Asad

The extensive use of large language models (LLMs), particularly in the finance sector, raises concerns about the authenticity and reliability of generated text. Developing a robust method for distinguishing between human-written and AI-generated financial content is therefore essential. This study addressed this challenge by constructing a dataset based on financial tweets, where original financial tweet texts were regenerated using six LLMs, resulting in seven distinct classes: human-authored text, LLaMA3.2, Phi3.5, Gemma2, Qwen2.5, Mistral, and LLaVA. A context-aware representation-learning-based model, namely DeBERTa, was extensively fine-tuned for this task. Its performance was compared to that of other transformer variants (DistilBERT, BERT Base Uncased, ELECTRA, and ALBERT Base V1) as well as traditional machine learning models (logistic regression, naive Bayes, random forest, decision trees, XGBoost, AdaBoost, and voting (AdaBoost, GradientBoosting, XGBoost)) using Word2Vec embeddings. The proposed DeBERTa-based model achieved an impressive test accuracy, precision, recall, and F1-score, all reaching 94%. In contrast, competing transformer models achieved test accuracies ranging from 0.78 to 0.80, while traditional machine learning models yielded a significantly lower performance (0.39–0.80). These results highlight the effectiveness of context-aware representation learning in distinguishing between human-written and AI-generated financial text, with significant implications for text authentication, authorship verification, and financial information security.

Cyberbullying has become a pervasive threat on social media, impacting the safety and wellbeing of users worldwide. Most existing studies focus on monolingual content, limiting their applicability to online environments. This study aims to develop an approach that accurately detects abusive content in bilingual settings. Given the large volume of online content in English and Arabic, we propose a bilingual cyberbullying detection approach designed to deliver efficient, scalable, and robust performance. Several datasets were combined, processed, and augmented before proposing a cyberbullying identification approach. The proposed model (FusionBullyNet) is based on fine-tuning of two transformer models (RoBERTa-base + bert-base-arabertv02-twitter), attention-based fusion, gradually unfreezing the layers, and label smoothing to enhance generalization. The test accuracy of 0.86, F1 scores of 0.83 for bullying and 0.88 for no bullying, and an overall ROC-AUC of 0.929 were achieved with the proposed approach. To assess the robustness of the proposed models, several multilingual models, such as XLM-RoBERTa-Base, Microsoft/mdeberta-v3-base, and google-bert/bert-base-multilingual-cased, were also trained in this study, and all achieved a test accuracy of 0.84. Furthermore, several machine learning models were trained in this study, and Logistic Regression, XGBoost Classifier, and Light GBM Classifier achieved the highest accuracy of 0.82. These results demonstrate that the proposed approach provides a reliable, high-performance solution for cyberbullying detection, contributing to safer online communication environments.

PxKu70 and PxKu80 are highly conserved non-homologous end joining (NHEJ) components with strong expression in adult reproductive tissues of Plutella xylostella. CRISPR/Cas9-generated ΔPxKu70 and ΔPx...

Foot-and-Mouth Disease (FMD) is caused by the FMD virus. Indirect Sandwich Enzyme-Linked Immunosorbent Assay (IS-ELISA) was standardized to characterize the FMD serotype “O” virus. Total protein content in the guinea pig serum (whole serum), ammonium sulfate precipitated guinea pig serum (ASPGPS) protein and ion-exchange-based purified guinea pig serum (IEGPS) protein was measured as 52 µg/mL, 24 µg/mL and 10 µg/mL respectively. The whole serum of guinea pigs and rabbits showed the 1:32 and 1:64 anti-FMD serotype “O” virus neutralizing antibody titers, while the anti-FMD serotype “O” virus neutralizing antibody titer was 1:128 in the IEGPS proteins. IEGPS protein with 1:128 neutralizing antibody titers were used as capture/trapping antibodies in the standardization of the assay. The IEGPS protein 1:1000 diluted with 10 µg/mL of protein content was found to be optimum for capture/trapping antibodies. To cover residual blank spaces, different available blocking buffers were evaluated and Skimmed Milk Solution 5% in Phosphate-Buffered Saline (PBS5%) proved best amongst blocking buffers. Coating of 1:1000 diluted IEGPS at 37 °C for 1 h followed by storage at 4 °C for overnight was best for incubation time. FMD serotype “O” virus 1:100 diluted was optimum in IS-ELISA. Similarly rabbit anti-FMD serotype “O” virus specific immune serum 1:10,000 diluted and goat anti-rabbit IgG horseradish peroxidase conjugate 1:4000 diluted were found to be optimum during the standardization of the assay. Lastly ELISA plates proved to be best amongst the available plates for assay. In each experiment, the plateau region, test background and plate background were recorded. Lastly it became possible for the establishment of an optimized and potentially cost-effective IS-ELISA requiring further diagnostic validation in research and diagnostic laboratories in the country.

Stainless steel 316 L (SS316L) is widely used in precision engineering because of its corrosion resistance and mechanical reliability. This study presented a comprehensive investigation on the performance of a novel micro-electric discharge machining (µ-EDM) set up employing a graphene nanoplatelet (GNP)-enriched dielectric fluid. Adding GNP to dielectric fluid served to enhance the performance of µ-EDM on stress-relief heat treated SS316L. Experiments were conducted using copper (Cu), brass, and aluminium (Al) electrodes under different settings of EDM process parameters such as pulse ratio (PR), peak current (IP), and GNP concentration (CP). Machinability was evaluated in terms of accuracy index (AI, %), material removal rate (MRR, mm³/min), and surface integrity through microscopic and 3D topographic analysis. The topographic evaluation revealed that setting EDM process parameters at their lower bounds (i.e. PR = 0.5, IP = 10 A, CP = 0.5 g/L) minimized surface defects and craters, yielding the highest AI(%) and lowest MRR. Conversely, higher energy inputs (i.e. PR = 1.5, IP = 20 A) increased MRR but degraded surface quality. Parametric sensitivity analysis assessing the mean effects of EDM process parameters on the response measures AI (%) and MRR indicated that setting PR and IP at their lower bounds and CP at its upper bound (i.e. 1.5 g/L) always yield the highest values of AI (%) for all electrodes; the Cu electrode outperformed the others with respect to both AI (%) and MRR. Experimental data were fitted by means of artificial neural networks (ANN) modelling. Multi-objective desirability optimizations identified the optimal settings − PR=1.5, IP=10 A, CP = 1.5 g/L − achieving AI (%) = 97.16% and MRR = 3.45 mm³/min for the Cu electrode (desirability score: 0.7390). Multi-objective optimization employing the non-dominated sorting genetic algorithm NSGA-II further refined these outcomes, delivering MRR values of 3.08, 2.20, and 4.45 mm³/min, and AI (%) values of 94.72%, 90.26%, and 96.01% for Al, brass, and Cu electrodes, respectively, with the optimal settings PR = 1.49, IP = 19.95 A, and CP = 0.58 g/L. Compared to DOE baseline solution, NSGA-II showed significant gains in MRR (up to 72.81%) and marginal improvements in AI (%) (< 0.35%). The results of this study fully demonstrated the efficacy of GNP-enriched dielectrics in delivering high-precision, high-performance µ-EDM of stress-relief heat treated SS316L with enhanced surface finish and machining efficiency; Cu electrode was the most suited for performing µ-EDM operations on SS316L workpieces.

Metaheuristic methods are extensively applied to tackle difficult continuous optimization problems in science and engineering. But many traditional methods still face challenges in achieving a balance between exploration and exploitation, which may result in premature convergence and local optima. This paper introduces the Father-Inspired Metaheuristic (FIM), a human-based hybrid metaheuristic inspired by a father’s guidance to children through discipline, learning, reward, and correction. The proposed FIM is organized into three main phases. The first phase is the neighborhood-repulsion-based discipline phase, which identifies weak local neighborhoods and avoids solutions going to weak nearby solutions, while still being guided by the father solution. The second phase is a hybrid of self-learning and elite guidance, where a solution either performs an adaptive perturbation around its current position or follows an elite solution guided by peer-difference information. In the third phase, reward-correction is applied, where a light Quasi-Newton refinement improves promising solutions, and a periodic correction mechanism redirects weak solutions to restore diversity. The algorithm is evaluated on the IEEE CEC2017 benchmark suite in 10D, 30D, and 50D, classical benchmark functions, and engineering design problems. Comparative results against five well-established metaheuristic algorithms show that FIM achieves strong accuracy, stable convergence, and robust performance across different continuous optimization landscapes.