Search Results (8,321)

In large-scale state-owned enterprise audit and compliance tasks, ensuring that similar violations receive consistent disciplinary decisions is essential for procedural fairness and institutional credibility. However, existing retrieval methods face three major challenges: lexical matching methods fail to recognize semantically equivalent violation descriptions, general-purpose semantic encoders lack knowledge of inspection-specific terminology and regulatory distinctions, and retrieved precedents are often not directly transformed into actionable disciplinary references. To address these problems, this paper proposes InspectCL, a domain-enhanced contrastive learning and Retrieval-Augmented Generation framework for similar case retrieval, validated on audit data from a provincial power grid company. First, to provide task-specific supervision that is unavailable in existing benchmarks, we construct InspectCase, a de-identified dataset of 4200 audit and compliance cases across 12 violation categories, with expert-validated positive pairs and hard negative pairs. Second, to overcome the weak domain awareness of generic encoders, we design a domain-enhanced contrastive learning model. Specifically, terminology-masking augmentation improves robustness to specialized inspection expressions, regulatory semantic injection incorporates disciplinary rules to distinguish factually similar but legally different cases, and hierarchical contrastive optimization strengthens both case-level similarity learning and category-level boundary separation. Third, to convert retrieved precedents into practical decision support, the Top-K similar cases are used as evidence for a large language model to generate structured disciplinary recommendation summaries, including violation classification, penalty references, applicable regulations, and rectification measures. Experimental results on InspectCase show that InspectCL substantially outperforms BM25, BERT-base, SimCSE, and Legal-BERT baselines, achieving 56.9% ± 0.7% Recall@5 and an 87.6% ± 0.4% Penalty Consistency Score (PCS). These results demonstrate that the proposed problem-driven modules jointly improve semantic retrieval accuracy and disciplinary decision consistency, offering a practical reference for similar power-grid audit scenarios, with broader applicability to be validated in future cross-domain studies. Full article

Gob-side roadways driven along the floor while retaining top coal in thick soft coal seams are prone to instability under strong mining-induced dynamic loading. To clarify the instability mechanism and develop an effective control method, the 1609 return airway of Jiulishan Mine was investigated using field survey, borehole imaging, FLAC3D numerical simulation, industrial testing, and field monitoring. The results show that, under the combined effects of large mining height, insufficient filling of the gob by the caved immediate roof, weak retained top coal, and low coal strength, shear failure planes tend to develop within the narrow coal pillar and extend from the gob-side roof toward the floor. Once the dominant shear plane cuts through the pillar, the overall bearing structure is destroyed, leading to shear slip, asymmetric rib deformation, roof subsidence toward the coal-pillar side, and rib–roof coupled instability. Based on a multi-index evaluation of pillar load-bearing capacity, plastic zone development, stress concentration, roadway deformation, and coal recovery, a 3 m coal pillar was determined as the rational width. A coordinated “narrow coal pillar + cross-rib anchorage” scheme was proposed, and field verification confirmed its effectiveness in controlling roof separation, roadway surface displacement, and internal surrounding-rock damage. Full article

Sustainable agricultural development in rural sub-Saharan Africa increasingly depends on coordinated governance of energy and water resources. Despite the growing deployment of solar photovoltaic water pumping systems (SPVWPS), little is known about how the institutional framework shapes SPVWPS effectiveness for productive agricultural use in rural Tanzania. Drawing on systems thinking concepts, specifically hierarchy, interaction, and interconnectedness, this study analyses the institutional frameworks governing energy and water provision for irrigation and livestock keeping across three rural Tanzanian communities. A mixed-methods design was employed, with qualitative inquiry as the primary mode; 65 household surveys, nine semi-structured interviews with community leaders, SPV developers, and local officials, and seven focus group discussions with farmers and livestock keepers were conducted across the three study areas. National energy and water policy documents, reports, and strategic plans were also reviewed to contextualise the institutional frameworks governing energy and water delivery in rural areas. Findings reveal limited coordination among stakeholders, particularly between NGOs, government agencies (REA, RUWASA, and NIRC), and local communities in the planning and implementation of SPVWP projects. Top-down delivery mechanisms marginalised community feedback, undermining local ownership and limiting the productive use potential of installed systems. This study proposes an integrated institutional framework that combines systems thinking with bottom-up and top-down approaches, explicitly embedding structured feedback mechanisms and aligning stakeholder roles across all governance levels. The framework was validated through interviews with experts in the rural energy and governance field, confirming its practical relevance and applicability to rural energy–water governance. The framework offers actionable guidance for policymakers and development practitioners seeking to strengthen institutional coordination in rural energy–water–agriculture governance, contributing to progress towards SDG 7 and SDG 2 across sub-Saharan Africa. Full article

Fine-grained aircraft recognition in optical remote sensing imagery remains highly challenging under cloud occlusion, as visibility degradation causes structural information loss and weakens discriminative representation learning. To address this issue, we propose DGR-MAE, a teacher–student masked image modeling framework based on posterior semantic correction for robust representation learning under incomplete observations. Unlike existing semantic-guided masking methods that modify token visibility during input construction, DGR-MAE preserves high-ratio stochastic masking in the student branch and introduces semantic correction after visibility degradation through teacher-guided differential reconstruction. Specifically, a semantic-aware teacher branch estimates patch-level importance to partition masked regions into semantic-critical and non-critical subsets, enabling region-dependent reconstruction prioritization. A collaborative feature refinement mechanism is further incorporated to enhance contextual consistency and structural reasoning during pretraining. To support controlled evaluation, we construct the ASRAir benchmark with hierarchical cloud occlusion levels. Experimental results show that DGR-MAE achieves 74.28% Top-1 accuracy on ASRAir-Occ and achieves the best Top-1 performance while maintaining competitive Top-5 accuracy compared with representative self-supervised baselines. In particular, it demonstrates substantially improved robustness under moderate-to-severe cloud occlusion, validating the effectiveness of posterior semantic correction for remote sensing representation learning under visibility degradation. Full article

Individual tree classification has a long history of diverse development, with recent trends focusing on the adoption of machine learning and deep learning approaches. It is a simple and powerful approach that allows the model to auto-pilot while reducing the need for physical characteristic understanding. Over more than a decade of research, we have focused on establishing a direct representation of individual trees that bridges 2D top-down imagery and true 3D models. In this study, we investigated the fundamental question of the influence of the input data on these ML/DL models. In 2024, we introduced a novel data transformation method, the Pseudo Tree Crown (PTC), which provides a pseudo-3D pixel-value perspective that enhances the informational richness of images and significantly improves classification performance. Our original implementation was successfully tested on urban and deciduous trees in 2024 and was later extended to Canadian natural conifer species under snow conditions in 2025. However, the original PTC relied on the green band, limiting its applicability to green-leaf species. In this study, we analyzed and compared the performance of different data variations and transformations, such as the Green–Red Vegetation Index (GRVI) and principal component analysis (PCA), as direct input and used their PTC forms. Classifications were conducted using Random Forest (RF), ResNet50, YOLOv10 and Segment Anything (SA). The results confirmed the effectiveness of the PTC, which consistently improves the classification accuracy by at least 5% without introducing additional computational time or complexity. Furthermore, PTC exhibits robust, consistent behavior across all data forms, demonstrating its strong resilience and reliability. Full article

The field of pose estimation is a major problem in computer vision, enabling the direct transformation of an input image into a hierarchical representation of the human skeleton for application in the fields of virtual/augmented reality and human–machine interaction tasks. Research in this field has exploded between 2018 and 2025, with traditional taxonomies such as 2D versus 3D or top-down versus bottom-up no longer sufficient to capture the essence of the evolution of ideas. To solve this problem, we propose a conceptual review in the field of pose estimation, focusing on the intellectual evolution of methods and architecture rather than the standard flat classifications of papers. We divide recent advances into five structural pillars: Representation, which traces the evolution from pixel coordinate regression to heatmaps and probabilistic representation; Architecture, which analyzes the transition from multi-stage CNNs to transformers and state space models (SSMs); Ambiguity and Generalization, which analyzes how self-supervised, uncertainty-aware, and diffusion models address 3D depth ambiguity, occlusion, and domain gaps by modeling multiple plausible poses and reducing dependence on fully supervised in-the-wild 3D labels; Context Extension, which covers temporal dynamics, multi-view fusion, and potential sensors; and Applications, which links algorithms to efficiency, privacy, and foundation models. By providing an in-depth detailing of these pillars, we provide a unified view of the evolution of research paradigms that define human pose estimation and enable the identification of future problems and solutions in pose estimation and human-centered tasks. Full article

Paper-based logistics documents remain widely used in multi-enterprise supply chains, where heterogeneous layouts, noisy document images, non-standard material descriptions, and limited edge-computing resources make structured extraction and material matching difficult. This paper proposes RRA-Logis, a lightweight multimodal large-language-model framework for logistics document understanding and material entity alignment. Instead of treating logistics document processing as a conventional field-extraction task, RRA-Logis formulates it as a document-to-entity alignment problem under resource constraints. The framework combines schema-constrained image-to-JSON extraction, LoRA/QLoRA instruction tuning, vector-based candidate recall, LLM-based semantic verification, confidence-gated decision making, and human-in-the-loop data evolution. Its methodological contribution lies in organizing these components into a resource-aware decision mechanism that determines whether a material-matching result should be automatically accepted or routed to human verification according to confidence and ambiguity margins. Experiments under a 24 GB VRAM constraint show that the fine-tuned Qwen2.5-VL-7B model achieves 85.4% document-level extraction accuracy and 100% JSON compliance on L-Doc-2K, while the proposed two-stage material-alignment method achieves 92.8% Top-1 accuracy. Ablation results indicate that LLM-based re-ranking, fused scoring, and confidence-gated verification each contribute to improved alignment reliability. Additional evaluation on the public DocILE benchmark and a desensitized real-document subset further examines cross-domain extraction transfer and the gap between Sim-to-Real data and operational logistics documents. The results suggest that RRA-Logis provides a practical framework for logistics document automation under constrained computing resources, while larger-scale real-world validation and broader benchmarking against specialized document-intelligence systems remain necessary. Full article

With the expanding application of digital technologies in urban renewal, more effective ways of incorporating dispersed public experience and needs into the renewal process still require further exploration. To address this issue, this research innovatively proposes an AI-assisted renewal method for historic districts driven by urban memory, constructing a continuous methodological chain from the identification of public evaluations to problem translation, to scheme generation and feedback validation. This research integrates the concept of interessement devices from Actor-Network Theory (ANT) with generative AI technologies for case application and validation. Taking Hanzheng Street as a case study, this research extracts the public’s urban memory of the historic district from online comments and identifies renewal demands. These demands were further associated with urban image elements to clarify their spatial carriers and support the subsequent generation of scene-based renewal schemes. On this basis, AI-generated images are further used to present renewed scenarios, and public evaluations of the renewal effects are collected. The results show that urban memory of Hanzheng Street can be summarized into five themes, which were further translated into five obligatory passage points (OPPs), one core issue, and corresponding renewal demands for scene units. The renewal schemes generated through this method achieved a relatively high level of public recognition overall, with mean evaluation scores ranging from 4.10 to 4.27, an overall satisfaction mean of 4.19, and a Top-2 proportion of 82.8%. By incorporating public experience into the formation of renewal schemes, this research provides a people-oriented and effective pathway for participation and feedback in the renewal of historic districts, while also offering methodological reference for the renewal of similar historic districts. Full article

Following the global financial crisis, the transition to IFRS 9’s forward-looking Expected Credit Loss (ECL) model has introduced significant implementation complexity, particularly in emerging markets facing data limitations. This study investigates the heterogeneous ECL compliance strategies adopted within the Cambodian banking sector during a period of heightened credit stress, marked by a system-wide non-performing loan ratio of 8.6%. Utilizing a multiple-case study design and replication logic, a qualitative content analysis was conducted on the 2024 audited financial statements of 13 representative institutions, ranging from market leaders to international subsidiaries. The findings reveal a pronounced technical divide: market leaders utilize advanced internal statistical methods, such as cohort analysis, whereas international subsidiaries rely on top-down parent-group proxy models to bridge local data gaps. A “macro-correlation paradox” was identified, where certain institutions prioritize faithful representation by excluding macroeconomic variables when statistical links to historical defaults remain weak. Furthermore, a significant transparency gap exists, where granular disclosures are consistent with a signaling interpretation regarding institutional safety. These results suggest that ECL compliance in data-limited environments may be interpreted as a strategic management choice rather than a standardized technical exercise, highlighting the need for regulatory standardization of modeling assumptions to improve inter-bank comparability. Full article

With increasingly fierce competition in the global semiconductor industry, equipment manufacturing, as a core link in the industrial chain, directly determines the overall development level of the semiconductor industry through its technical capabilities and industrial layout. This study takes the semiconductor equipment industrial chain as the research object and selects two leading international enterprises, AMAT and ASML, and two representative Chinese enterprises, NAURA and AMEC, for study. Based on corporate annual report data, a knowledge graph method is adopted to construct industrial chain knowledge graphs for the four enterprises. Through ontology design, knowledge extraction, and graph construction, while incorporating indicators such as network density, betweenness centrality, and average path length, a comparison and analysis of the industrial chain structures of the four enterprises is carried out. Quantitative analysis results indicate that the network densities of AMAT and ASML are 0.01456 and 0.01430, with average path lengths of 2.2953 and 2.1743. In combination with their network scale and the fact that their key nodes with high betweenness centrality are core equipment and top enterprise clients, their industrial chain networks exhibit strong connectivity. The network densities of NAURA and AMEC are 0.01221 and 0.01736, with average path lengths of 1.9725 and 1.8738, respectively. Specifically, NAURA is characterized by wide industrial chain coverage, while AMEC features a high connection density among industrial chain links. The comparative analysis reveals the differences between Chinese and international semiconductor equipment enterprises in terms of industrial chain network connectivity and control over core nodes. Full article

This paper presents a passive residue-coupled discrete-time delta–sigma ($\Delta \Sigma$) modulator for low-power narrowband sensing applications. Instead of adding a fourth active integrator, the proposed architecture keeps a third-order switched-capacitor main loop and reuses the intrinsic top-plate residue of an 8-bit asynchronous successive-approximation-register (SAR) quantizer. The retained capacitive digital-to-analog converter (CDAC) residue is passively reinjected through a charge-redistribution path, introducing an additional high-pass error-propagation factor in the effective noise transfer function (NTF). Under a bounded effective coupling coefficient, the proposed loop approaches fourth-order-like in-band noise suppression while retaining third-order active-loop complexity. Behavioral simulations show that the Enhanced mode improves the peak signal-to-noise-and-distortion ratio (SNDR) by 16.9 dB over the Baseline third-order mode at an oversampling ratio (OSR) of 128. Circuit-level corner verification of the standalone SAR confirms correct bit cycling and a settled residue-retention window under typical–typical (TT), slow–slow (SS), and fast–fast (FF) conditions: with the slowest conversion window of about 21.4 ns at the SS corner and a sampling period of 39.06 ns at $f_s=25.6$ MHz, roughly 17.66 ns of timing margin remains for residue holding, passive reinjection, and clock non-overlap. The proposed method provides an architecture-level route for improving in-band noise shaping without increasing the number of active integrator stages, and is particularly attractive for low-power, narrowband, and sensor-oriented analog-to-digital converter (ADC) applications. Full article

For the sake of investigating the theoretical design optimization of high-rise plane frames, an optimization model was established by taking the minimum top-story lateral displacement as the objective function and treating material strength, story height, and span length as design variables. The design parameters of the frame were optimized using an Assimilation–Competition Quantum-behaved Particle Swarm Optimization (ACQPSO) algorithm. First, the accuracy and computational efficiency of the ACQPSO algorithm were evaluated using four benchmark functions. Then, a five-span, seven-story reinforced-concrete plane frame with a total span of 24 m and a total height of 34 m was taken as a case study. The cross-sectional dimensions of the beams and columns were determined according to relevant design specifications, and the top-story lateral displacement calculated by the D-value method was verified using the Finite Element Method (FEM), confirming its accuracy and effectiveness. Finally, a parametric analysis was carried out to investigate the effects of material strength, story height, span length, and member cross-sectional dimensions on the objective function. The results indicate that story height and column concrete strength have a greater influence on the top-story lateral displacement, whereas the effect of span length is relatively small. In addition, the cross-sectional dimensions of beams and columns affect the top-story lateral displacement more significantly than beam strength. Full article

Engineering consultants and design firms are central to the success of construction projects. However, the systematic evaluation of their performance in the Saudi Arabian context remains methodologically fragmented and empirically underdeveloped. Existing prequalification frameworks rely predominantly on administrative criteria and single-method ranking approaches that cannot adequately differentiate between high- and low-performing firms. To address this gap, the study proceeds in two distinct parts. Part I—Literature Review: A PRISMA-compliant systematic literature review across five major academic databases was conducted to map the existing evidence base, identify three substantive gaps in the Saudi and GCC engineering firm evaluation literature, and derive a consensus-based set of 29 performance criteria grouped into seven dimensions. This review constitutes an independent contribution: it establishes the gap that motivates the empirical work and provides the criterion framework on which that work is built. Part II—Practical Application: A structured questionnaire was administered to 288 construction professionals in Saudi Arabia (Cronbach’s α = 0.936), and the collected data were analyzed through a hybrid RII–Shannon Entropy Weighting (EWM)–TOPSIS pipeline that produced a Composite Priority Index (CPI) for each criterion, enabling a stable and discriminating ranking that integrates subjective expert consensus with objective distributional information. The main finding revealed that five criteria attained Very High Priority status (CPI > 0.70): Supervisory Experience (CPI = 0.740), Engineers’ Capability Index (CPI = 0.717), License Class (CPI = 0.709), Client Satisfaction Index (CPI = 0.708), and Average Delay Time (CPI = 0.705). These top-ranked criteria collectively center on technical leadership, regulatory standing, client-reported outcomes, and schedule reliability, indicating that procurement decisions should prioritize demonstrable competence over structural size or geographic footprint. The consistently lower importance of physical branch networks and headquarters location further suggests that remote management capabilities and digital coordination tools are reshaping performance expectations under Saudi Vision 2030. The Quality Indicators dimension achieved the highest mean CPI across all seven dimensions. The findings provide actionable evidence for procurement authorities, regulatory bodies, and engineering firms seeking to strengthen performance-evaluation practices in the Saudi construction sector. Full article

Against the backdrop of the automotive industry’s transition to low-carbon operations, assessing the sustainability of pure electric vehicle products remains crucial. Existing multi-criteria evaluation methods often follow a compensatory logic, allowing high carbon emissions to be offset by other advantages. This contradicts the core principle that sustainability must be non-negotiable. To address this issue, we propose a two-stage non-compensatory evaluation framework. First, we apply a carbon footprint threshold based on life cycle assessment: any candidate vehicle exceeding this threshold is eliminated. Second, the remaining models are evaluated across ten indicators (economic, social, and technical), and a comprehensive ranking is generated using entropy weighting, fuzzy analytic hierarchy process (FAHP), and the TOPSIS method. This framework has been validated on seven mainstream BEV midsize sedans. The results show that the non-compensatory screening mechanism eliminated two high-carbon-emission models, confirming that environmental criteria must be considered independently. The top-ranked model was not the one with the lowest carbon emissions but rather the one demonstrating balanced performance, indicating that environmental performance and overall competitiveness can be enhanced synergistically. The ranking results remained relatively robust even under a combination of objective and subjective weightings. This study provides a more logically consistent tool for evaluating pure electric vehicles at the product level. Full article

Remote sensing change detection is crucial for urban expansion monitoring and ecological assessment. Recently, methods based on Convolutional Neural Networks (CNNs) and Transformers have advanced significantly. However, state-of-the-art models relying primarily on pure spatial-domain modeling and absolute feature differences struggle to balance global semantics with high-frequency boundary details. This paradigm loses physical change directionality and amplifies pseudo-change noise in complex backgrounds. To overcome this, we propose a Frequency-Aware Cross-Layer Change Detection Network (FACDNet) that leverages frequency-spatial synergy to enhance feature discriminability. Specifically, a Wavelet Interaction Block (WIB) decouples bitemporal features using Haar wavelets, employing heterogeneous attention to targetedly reinforce macroscopic semantics and edge textures. Furthermore, to mitigate noise in shallow features, a Cross-Layer Frequency Context Aggregator (CLFCA) injects deep global semantics top-down, purifying multi-scale spatial gating signals. Finally, a Context-guided Difference Fusion Module (CDFM) extracts direction-aware bidirectional difference features, utilizing the purified gating to accurately suppress pseudo-changes. Extensive experiments on the LEVIR-CD and highly challenging SHCD datasets demonstrate FACDNet’s remarkable robustness. It achieves change-class F1-scores of 92.04% and 83.64%, and Intersection over Union (IoU) scores of 85.26% and 71.89%, respectively, achieving highly competitive performance compared with existing mainstream methods. Full article