McPherson University

Large language models have made substantial progress in mathematical reasoning. However, benchmark development for multilingual evaluation has lagged behind English in both difficulty and recency. Recently, GSM-Symbolic showed a strong evidence of high variance when models are evaluated on different instantiations of the same question; however, the evaluation was conducted only in English. In this paper, we introduce MGSM-Pro, an extension of MGSM dataset with GSM-Symbolic approach. Our dataset provides five instantiations per MGSM question by varying names, digits and irrelevant context. Evaluations across nine languages reveal that many low-resource languages suffer large performance drops when tested on digit instantiations different from those in the original test set. We further find that some proprietary models, notably Gemini 2.5 Flash and GPT-4.1, are less robust to digit instantiation, whereas Claude 4.0 Sonnet is more robust. Among open models, GPT-OSS 120B and DeepSeek V3 show stronger robustness. Based on these findings, we recommend evaluating each problem using at least five digit-varying instantiations to obtain a more robust and realistic assessment of math reasoning.

Deep-learning models for thermal-infrared gas classification have thus far treated thermal images as ordinary photographs, discarding the fact that every pixel contains information on physically calibrated temperature governed by fixed thermodynamic law. This paper presents a Physics-Informed Machine Learning (PIML) system that classifies gas environments from thermal-infrared images using both physics-grounded input features and a physics-penalized training objective. Rather than utilizing the three-channel RGB representation used in all prior work on the MultimodalGasData thermal-image subset, in this work, each image is converted into a six-channel tensor comprising temperature-normalized RGB channels, horizontal and vertical thermal-gradient maps derived from Fourier's law of heat conduction via the Sobel operator, and a local temperature-anomaly map obtained by neighborhood-mean subtraction. A MobileNetV2 backbone with a custom input-projection block is trained in two phases; frozen-backbone feature adaptation followed by fine-tuning of the top thirty backbone layers — under a physics-informed loss function that penalizes confident gas predictions in thermally flat, physically implausible image regions. The resulting model attains 94.80% test accuracy and 96.67% best validation accuracy across four classes (No Gas, Perfume, Smoke, Mixture), with a Mixture-class F1-score of 99.58% and zero false Mixture alarms across 722 non-Mixture test images. Occlusion-based saliency analysis shows model attention concentrated in thermodynamically meaningful regions (the base of rising smoke plumes, diffuse upper-region warmth for aerosol dispersion, and multi-region activation for simultaneous emissions) thus providing the first physical-plausibility validation reported for this dataset. The model improves by 1.80 percentage points over a thermal-only CNN baseline lacking physics-informed features and closes to within 1.2 percentage points of a full multimodal sensor-fusion result, while trailing a purely accuracy-optimized MobileNetV2 baseline on standard preprocessing by 4.96 percentage points. This study provides a quantified interpretable trade-off between raw accuracy and physical grounding which is argued to be favorable for safety-critical deployment.

The rapid growth of digital financial services has significantly increased the volume of online transactions, making fraud detection a critical challenge for financial institutions. Traditional machine learning models often provide strong predictive performance but lack interpretability, limiting trust and practical adoption in financial decision-making. This study proposes an Explainable Artificial Intelligence (XAI)-based fraud detection framework for FinTech transactions using the Kaggle Credit Card Fraud Detection dataset containing 284,807 transactions, including 492 fraudulent cases. To address severe class imbalance, Synthetic Minority Oversampling Technique (SMOTE) was applied, increasing the dataset to 568,630 balanced instances. Data preprocessing involved feature scaling and train–test splitting prior to model training. Three machine learning algorithms—Logistic Regression, Random Forest, and Extreme Gradient Boosting (XGBoost) were developed and evaluated using accuracy, precision, recall, F1-score, and ROC-AUC metrics. The experimental results demonstrate strong predictive performance across all models. Logistic Regression achieved 94.50% accuracy, 97.32% precision, 91.51% recall, 94.33% F1-score, and a ROC-AUC of 94.50%. Random Forest produced the highest overall performance with 99.99% accuracy, 99.98% precision, 100.00% recall, 99.99% F1-score, and 99.99% ROC-AUC. XGBoost also achieved excellent results with 99.97% accuracy, 99.94% precision, 100.00% recall, 99.97% F1-score, and 99.97% ROC-AUC. To improve model transparency, SHapley Additive exPlanations (SHAP) and Local Interpretable Model-Agnostic Explanations (LIME) were integrated with the XGBoost model to provide both global and local interpretability. SHAP analysis identified transaction amount and several transformed principal component features as the most influential predictors of fraudulent behavior, while LIME provided instance-level explanations for individual fraud predictions. Feature importance analysis from Random Forest and XGBoost further validated the consistency of the most influential variables. The findings demonstrate that combining high-performing machine learning models with explainable AI techniques can significantly enhance fraud detection accuracy while maintaining transparency and interpretability. The proposed framework offers a reliable and practical approach for intelligent fraud prevention in financial technology systems and supports trustworthy decision-making in real-world financial environments.

Background Research is crucial to evidence-based practice. Nurses need to build their capacity to conduct research and other scholarly endeavors. Little is known about the research needs and the capacity of Nurses in Lagos State, Nigeria, which prompted this study. Method A cross-sectional study was conducted among 475 nurses in nursing departments of healthcare institutions in Lagos. Data were collected via an online questionnaire and were analyzed using descriptive and inferential statistics. Hypotheses were tested using Pearson’s Chi-Square at 0.05 level of significance. Results The perceived research needs of nurses were predominantly high (77.3%) and moderate (20.4%), spanning all phases of research, from conceptualization to dissemination. However, their perceived individual research capacity was generally low (35.4%) to moderate (43.2%). In contrast, the perceived research capacity of nursing departments was perceived to be moderate (66.3%) and high (21.3%). A statistically significant association was observed between hospital setting (x2 = 20.56p =<0.001), years of experience (x2 = 20.066p = 0.010) and the perceived research needs of nurses. Furthermore, significant associations were found between hospital setting and both the individual perceived research capacity of nurses and the overall perceived research capacity of their departments (p < 0.05). Conclusion Nurses in Lagos State demonstrated high research needs but low to moderate individual perceived research capacity, despite moderate institutional support. This gap between research requirements and existing competencies highlights the need for targeted capacity-building initiatives.