Recent advances in multimodal large language models (LLMs) have expanded their potential for solving complex mathematical problems; however, their reasoning capabilities in real, curriculum-aligned examinations remain underexplored. This study evaluates the mathematical reasoning performance of OpenAI’s o4-mini model on the mathematics section of the Korean College Scholastic Ability Test (KCSAT), using official 2023–2025 exams and mock tests as benchmark data. The o4-mini is a multimodal LLM capable of processing both textual and diagrammatic information, enabling assessment on problems that require integrated visual reasoning. Quantitative analyses reveal that o4-mini demonstrates strong performance on standard problems and moderate success on high-difficulty items, while qualitative analyses uncover a distinctive “Reasoning–Verification–Reanalysis (RVR) loop” in its reasoning process. This iterative pattern—where the model reasons, self-verifies intermediate steps, and reanalyzes discrepancies—reflects a human-like metacognitive behavior that often improves problem-solving accuracy. These findings provide new insights into the cognitive dynamics of multimodal LLMs, highlighting both their potential and limitations in high-stakes mathematical assessments, and suggesting pathways for enhancing AI reliability through structured reasoning and self-verification mechanisms.

Alzheimer’s disease (AD) is a progressive neurodegenerative disorder that severely affects memory and cognitive function, underscoring the need for accurate and early diagnostic tools. Deep learning has shown strong potential in automated AD detection from brain MRI scans; however, many existing approaches either overlook the full volumetric context of MRI data or rely on limited feature fusion strategies that weaken spatial coherence. To address these challenges, this study proposes BiFPN3DViT, a hybrid deep learning framework that integrates a hierarchical 3D convolutional backbone as the foundational component of a Bidirectional Feature Pyramid Network (BiFPN), coupled with a Vision Transformer (ViT) for global contextual learning. This study evaluated under leakage-safe subject-level splits on the ADNI dataset. The unified 3D CNN–BiFPN module jointly extracts and fuses multi-scale volumetric features through both bottom-up and top-down pathways, forming a rich spatial representation that is subsequently refined by the transformer encoder. Evaluated on 4,706 MRI scans from the Alzheimer’s Disease Neuroimaging Initiative ((ADNI) dataset, the proposed model achieved over 92% classification accuracy across AD, mild cognitive impairment (MCI), and cognitively normal (CN) subjects. Comparative and ablation studies confirm the complementary strengths of convolutional and transformer-based modelling, while attention visualisation highlights clinically relevant brain regions. These results demonstrate that BiFPN3DViT offers a scalable, interpretable, and computationally efficient framework for MRI-based Alzheimer’s disease diagnosis, advancing the integration of multi-scale feature learning and attention-driven analysis in neuroimaging.

The Vehicle-to-Grid (V2G) system has been a significant solution for enhancing the power grid’s stability and supporting renewable energy. However, the primary barrier to the practical application of V2G technology has been a fundamental economic conflict: the accelerated battery degradation from aggressive battery cycling required to provide grid services creates a direct trade-off between generating revenue from grid services and preserving the battery asset’s life. Therefore, the objective of this research is to develop an intelligent control framework to optimize both the profit generated from grid services provided and the longevity of the batteries. The proposed Physics-Informed Deep Reinforcement Learning (PI-DRL) framework utilizes a Digital Twin of the electrochemical behavior of the batteries to generate a real-time physics-based cost signal of degradation, which guides the learning of the policy by a deep reinforcement learning agent. Comprehensive VPP simulation results demonstrate that the proposed PI-DRL framework outperforms all benchmark approaches, achieving significant increases in net profitability and drastic reductions in fleet-wide capacity fade. The agent learned sophisticated control strategies, including making proactive deviations from the optimal control trajectory to avoid acute mechanical stress on the battery and optimizing operations across a heterogeneous fleet by using robust chemistries for high-intensity grid service tasks. A key implication of this research is that there is now a viable blueprint for the economically sustainable and equitable provision of V2G services, with an asset-preserving strategy being the most profitable method.

This systematic literature review investigates how evolvable knowledge graphs (KGs) enhance manufacturing. Evolvable KGs adapt to changing knowledge by leveraging machine learning algorithms and human expertise to improve decision-making, operational efficiency, and predictive maintenance capabilities beyond the capabilities of prevailing solutions enabled by dynamic KGs. This review maps the existing literature based on the stages of the KG development process. The main results are an updated model of the creation and maintenance of evolvable KGs in manufacturing and a literature-based synthesis of learning approaches for enabling KG evolution through knowledge processing and knowledge utilization. The results contribute to updating the KG development process model and synthesizing an understanding of evolvable KGs that utilize the observed learning approaches on knowledge processing to change KG relations and nodes, and three learning approaches through knowledge utilization (human-guided, machine-driven, and human-machine collaborative knowledge updates). Despite emerging advances, challenges persist in quality assurance, process planning, and integration of human expertise. The findings advocate addressing these issues to promote greater adoption and optimization of KGs in manufacturing. By deepening the understanding of how KGs can evolve through learning, this review sets a conceptual basis for future research to develop more dynamic and intelligent KG-based manufacturing systems.

A present challenge is the planning of a resilient power distribution system (PDS) as to energize critical loads after weather events. To that end, this paper sets forth: (i) A formulation for planning of resilient PDS that considers the formation of microgrids when isolating faulted lines. (ii) An approach to train a constrained-policy reinforcement learning (CPRL) planner that uses only two deep neural networks in lieu of one per constraint. These novelties are showcased using a 17-node and a 60-node PDS while contrasting against proximal policy optimization (PPO), deep Q-network (DQN), and greedy (GRD) algorithms. The CPRL planner achieved the lowest average unserved demand across stochastic scenarios and the lowest constraint violation when contrasted against PPO, DQN, and GRD methods.

Polyp segmentation is essential in reducing the incidence of colorectal cancer by enabling the timely detection and removal of polyps, which are the precursors to most colorectal cancers. Accurate segmentation provides crucial information that aids in the early diagnosis and treatment of the disease. However, this task is challenging due to the considerable variation in polyp texture, color, and size. Additionally, some polyps closely resemble the surrounding healthy tissue in color, making it difficult to distinguish the boundaries between the polyp and its surrounding area. To address these challenges, we propose a deep learning model based on an encoder-decoder architecture. In the encoder section, we utilize a Pyramid Vision Transformer (PVT) to extract robust global feature representations across multiple stages. We introduce a multi-scale fusion module that combines features from different scales, allowing the model to capture diverse information. A custom convolution refinement attention module is employed to further refine the features, and it is integrated with the output of an efficient simple residual path, which preserves essential polyp features. Experiments are conducted on five publicly available polyp datasets. CVC-Clinic and Kvasir-SEG datasets are used for training, while CVC-ColonDB, Etis LaribDB, and CVC-300 datasets are used to evaluate generalization. The proposed approach achieves state-of-the-art performance, with a mean Dice coefficient of 93.9% and a mean Intersection over Union of 89% on CVC-Clinic, and 91.0% and 85.9% on Kvasir-SEG, respectively. Furthermore, our approach demonstrates the ability to adapt to new, unseen polyp features, achieving optimal results on the CVC-ColonDB, Etis LaribDB, and CVC-300 datasets. This adaptability makes it suitable for real clinical scenarios.

Industrial digitalization and the adoption of smart technologies have transformed asset management in capital-intensive sectors. Despite these advancements, little is known about how organizations translate their strategic orientation into tangible operational outcomes. This study examines how Entrepreneurial Orientation (EO) drives Operational Performance (OP) through Smart Maintenance (SM), considering Environmental Dynamism (ED) as a moderating factor. Drawing on the Dynamic Capabilities Theory (DCT) and the Contingency Theory (CT), a mediated and moderated model integrating strategic and operational domains is proposed. A cross-sectional quantitative approach using partial least squares structural equation modeling (PLS-SEM) was applied to a sample of 204 industrial firms in the manufacturing and mining sectors in Peru. The results showed that EO significantly influenced SM (β = 0.51; p < 0.001) and that SM improved OP (β = 0.35; p < 0.001), providing evidence of the partial mediation of SM between EO and OP. ED showed direct positive effects on both SM and OP, although its moderating influence on EO-related paths was not significant. The model achieved R² = 0.37 for OP and R² = 0.36 for SM, with satisfactory predictive validity (Q² > 0). These findings enhance the understanding of SM as a translation mechanism between strategy and performance, reinforcing the importance of aligning entrepreneurial capabilities, technology, and environmental dynamism to improve operational competitiveness in emerging economies.