Search Results (2,949)

Retrieval-augmented generation has emerged as a core technological paradigm for addressing the bottlenecks of hallucinations and knowledge lag in large language models. However, many existing reviews focus on a single technical branch or vertical application scenario, making only scattered references to hardware, evaluation methods, and cross-industry empirical evidence, and lacking a systematic, end-to-end integration. This paper conducts research based on a total of 115 papers, comprising foundational literature from 1998–2019 and core RAG literature from 2020–2026, systematically cataloging end-to-end technologies and supporting solutions, establishing a quantitative hardware comparison table and comparing 11 categories of open-source and commercial APIs, constructing a two-tier, four-level standardized evaluation framework, and compiling empirical evidence and implementation challenges across eight industries from 2024 to 2026. Based on this, the paper identifies four major structural contradictions—the retrieval–creation trade-off, the geometric–semantic misalignment as a symmetry problem between representation space and semantic structure, the autonomy–reliability paradox, and evaluation blind spots—as a unified analytical framework for the five major technological strands. Finally, this paper proposes four research directions for practical implementation—differentiable joint optimization, hybrid geometric space learning, interpretable causal reasoning, and multidimensional diagnostic evaluation—providing a systematic reference for both theoretical research on RAG and its deployment in the private sector. Full article

Old urban areas face severe spatial supply–demand contradictions as aboveground space nears saturation while public service demand grows. Developing underground public spaces offers a key solution, yet fragmented development limits overall benefits, necessitating coordinated aboveground–underground development to achieve urban renewal and sustainability. Taking Shanghai’s old urban areas as a case, this study constructs an evaluation system with 17 aboveground indicators across 5 dimensions and 9 underground indicators across 3 dimensions. Using the combination of AHP–entropy weight method for weighting, the coupling coordination degree model, and the obstacle degree model, this study identifies the temporal evolution trends in the development levels of the two systems, the characteristics of their coupling coordination stages, and the main constraining factors from 1995 to 2025. The results show: (1) Both systems have shown continuous growth, with underground public space accelerating its development after 2010, and by 2015, it had nearly caught up with the aboveground system in the time-series projection results; (2) The D value of coupling coordination has increased from 0.2431 to 0.9532, experiencing three stages of low coupling coordination, general coupling coordination, and high coupling coordination; (3) The obstacle factors have shown a dynamic evolution path from scale shortage to morphological complexity, and then to the synergy of the aboveground and underground morphologies. In the higher coupling coordination stage, the length of the aboveground bus lines and the landscape shape index of the underground became the dominant obstacles. This study provides a quantitative basis for coordinated planning and decision-making in the renewal of old urban areas. Full article

Generative artificial intelligence (GenAI) is being absorbed into education as a new infrastructural layer, promising individualized tutoring, instant feedback, translation, and accessibility support at marginal cost. Sustainable Development Goal 4 (SDG 4) calls for inclusive and equitable quality education for all, yet the evidence on whether GenAI advances or undermines that goal points firmly in both directions at once. At the task level, GenAI and intelligent tutoring systems repeatedly compress performance distributions, with the largest gains accruing to lower-skilled and lower-baseline participants. At the system level, a parallel literature on access, AI literacy, language, disability, teacher capacity, and over-reliance finds that benefits are conditioned by resources that track prior advantage. These two literatures are usually read as being in tension, and the tension is usually resolved by privileging one of them. This article argues that both are correct and that the appearance of contradiction is an artifact of conflating distinct stages of a single pathway. We develop the divide cascade: a four-stage filter model—access, effective use, benefit realization, and durable learning—in which each stage has a pass rate that correlates with prior advantage. Because pass rates compound multiplicatively across stages while compression acts additively within a stage, a technology can compress outcomes among those who clear every filter and still stratify outcomes across the population as a whole. We formalize this structure, derive the condition under which the stratifying force dominates the equalizing one, and state three predictions that distinguish the cascade from an access-centred account: that access-only interventions should attenuate rather than close benefit gaps, that a single intervention can narrow one gap while widening another simultaneously, and that measured equity gains should decay as the evaluation horizon lengthens. We also specify what would falsify the model. The framework is then used to identify the conditions that set the sign of GenAI’s distributional effect, to map those conditions onto SDG 4 targets, and to derive a testable research and policy agenda. A recurring corollary is methodological: the strongest evidence for compression comes from workplace-productivity studies that measure produced artifacts rather than durable learning, so its transfer to education is an open question that the cascade locates precisely rather than assumes. GenAI, we conclude, is neither inherently an equalizer nor an amplifier; it is a multiplier whose sign is set by how completely the cascade is engineered for the learners who start behind. Full article

When a system must detect several related conditions over the same inputs, its designer can train a specialized classifier for each condition or one pooled model for all of them. Published comparisons of these designs contradict one another, and certain dataset properties, class imbalance above all, usually take the blame. We re-examine the comparison on code-smell detection, a domain where the contradiction is well documented, testing three ways of building the pooled baseline across five code-smell datasets, four generic benchmarks, four classical classifier families, and a neural network, with matched cross-validation folds and threshold-independent scoring. Much of the disagreement turns out to be manufactured by the comparison itself. The conventional pooled baseline appends a task indicator to the features, which quietly forces linear models, and neural networks conditioned only through the input, to share one set of feature weights across tasks; tree ensembles and softmax classifiers, which carry per-task structure of their own, are unaffected. Fixed-threshold F1 scoring accounts for the rest. Once the pooled model receives per-task weights, the gap closes on every dataset, and controlled positive-rate sweeps show that class imbalance moderates nothing. For tabular task families of this kind, the choice between the designs is an engineering decision, not an accuracy one; we close with a short protocol for making such comparisons fairly. Full article

Accelerating aridity and desertification driven by global climate change pose growing threats to dryland forest plantations, making it essential to understand transpiration dynamics across developmental stages for both forest-hydrology theory and the sustainable management of protective shelterbelts. This study compared transpiration characteristics, stomatal conductance (g_s), and water-regulation strategies between a 41-year-old mature stand and a 13-year-old young stand of Mongolian pine (Pinus sylvestris L. var. mongolica Litv.) on the southern margin of Horqin Sandy Land during the 2024 growing season, using thermal-dissipation sap-flow measurements combined with meteorological monitoring and water-potential sampling. Mean individual-tree daily transpiration in the mature stand (2.36 mm·d⁻¹) was approximately 2.6 times that of the young stand (0.90 mm·d⁻¹), yet the young stand showed a disproportionately stronger sap-flow response to small rainfall events. Sap flow in both stands was jointly driven by vapor pressure deficit (VPD) and photosynthetically active radiation, with a near-synchronous lag of ±10 min and tight canopy-atmosphere coupling (decoupling coefficients: 0.177 and 0.256, respectively), indicating transpiration was predominantly governed by stomatal regulation. Stomatal conductance declined with rising VPD in both stands, with a steeper decline in the mature stand. Water-potential analysis revealed pronounced anisohydric behavior in the mature stand (σ = 2.643, R² = 0.765, p < 0.01), and near-strict anisohydric regulation in the young stand (σ = 0.919, R² = 0.105, p > 0.05), indicating high tree-level hydraulic variability and precluding a definitive classification along the iso/anisohydric continuum for this developmental stage. Both g_s and transpiration increased with tree size, contradicting the hydraulic limitation hypothesis. These findings elucidate distinct water-use strategies between the young and mature Mongolian pine stands in this paired design and provide a physiological basis for stage-differentiated, precision water management of dryland shelterbelt plantations. Full article

Production data transparency in smart manufacturing simultaneously enhances and impairs employee performance across organizational contexts. Existing research has not resolved this theoretical contradiction. Drawing on socio-technical systems theory, cognitive appraisal theory, and conservation of resources theory, this study develops a dual-pathway model. Data transparency influences adaptive performance through a bright empowerment pathway and a dark surveillance pathway mediated by EPM-induced strain. Procedural justice of data governance and digital self-efficacy operate as a perceived-institutional boundary condition and an individual-capability boundary condition, respectively. Latent moderated structural equations were applied to survey data from 412 employees in Chinese smart manufacturing enterprises. Results support both pathways and reveal a theoretically consequential asymmetry between these boundary conditions. Johnson–Neyman analysis indicates that institutional justice attenuates the strain pathway to non-significance within the observed distribution of responses. Conversely, digital self-efficacy requires near-ceiling levels to achieve the same pattern. This asymmetry indicates that continuous moderation produces sharply different practical outcomes across the observed data range. Institutional and individual remedies therefore address the visibility paradox on different practical scales. Governance adequacy represents a more attainable managerial lever than individual capability development. These findings advance socio-technical systems theory by detailing the asymmetric buffering capacities of different organizational resources. Full article

The value-creation capacity of active asset management remains one of the most debated issues within modern portfolio theory. While the relationship between costs and performance is typically negative in developed markets, in smaller and less liquid markets—where information asymmetry is more pronounced—higher fees may also be interpreted as signals of managerial ability. This study investigates this apparent contradiction in the context of the Hungarian equity mutual fund market, using a panel dataset covering 79 funds over the period 2017–2024, with a particular focus on identifying non-linear effects among performance determinants. The methodological framework combines fixed-effects panel regression with Driscoll-Kraay robust standard errors, complemented by quantile regression estimates to examine different segments of the return and alpha distributions. The results indicate that growth dynamics (NAV_change) and cumulative historical performance (Yield_from_start) consistently enhance fund performance, while the negative effect of past returns suggests the dominance of mean reversion. The impact of the total expense ratio (TER) proves to be non-linear and specification-dependent—a finding con-firmed by an extensive battery of robustness checks—thereby rejecting the cost-signalling hypothesis with respect to risk-adjusted excess returns (Jensen’s alpha). Quantile estimates further reveal that the effects of economies of scale and cost structure differ significantly between underperforming and top-performing funds, confirming that analyses based on average effects obscure the heterogeneity of market dynamics. By jointly modelling returns and risk-adjusted performance across the full conditional distribution, the study contributes a distribution-sensitive theoretical account of active management’s limitations in a small, less liquid market, showing that these limitations are conditional on fund size and cost structure rather than uniform across the fund population. Full article

To address the contradiction between high resolution and structural complexity in traditional laser self-mixing measurement systems, the co-axial dual-wavelength solid-state laser self-mixing technology was proposed and studied. An LD pumped the Nd:YVO4 crystal and doubled the frequency to generate two wavelengths laser at 1064 nm and 532 nm, and a co-axial dual-wavelength laser self-mixing measurement system was constructed. A 90° phase difference between the two wavelengths was produced by adjusting the angle of incidence of the parallel glass plate. As a result, a set of orthogonal signals was built to distinguish the direction of the displacement for the target. The experiments showed that the measuring system can adapt to different vibration waveforms and that the frequency measurement upper limit can be reached at 7 kHz. The system exhibited high precision in displacement measurement, with an RMS displacement noise of 8.64 nm and a cumulative error of 33.66 nm at a peak-to-peak amplitude of 5000 nm, along with a short-term resolution better than 2 nm. Full article

Graphite ore grade sorting is a key step toward intelligent mineral processing; however, it faces three representational contradictions: ambiguous classification posteriors at grade boundaries, asymmetric multi-scale feature interaction, and the mismatch between class-agnostic self-distillation assignment and sample-level difficulty. Targeting these, this paper adopts D-FINE as the baseline and introduces three decoupled improvements at its decoder, encoder, and criterion layers. (1) Boundary-Grade-aware Distribution Refinement (BG-FDR) online identifies boundary samples via the Top-2 classification score gap and modulates regression-distribution refinement, yielding +2.69 percentage points in mAP@0.5 with zero additional trainable parameters. (2) Bidirectional Feature Pyramid with Global–Local Spatial Attention (BiFPN-GLSA) builds a learnable weighted bidirectional multi-scale fusion path. (3) Difficulty-Aware Decoupled Distillation with Wise-Inner-Shape-IoU (DADD+Wise-IoU) imposes class- and sample-level difficulty-aware constraints. In the integrated full model, this increases Precision from 66.21% to 71.43% (+5.22 pp), F1 from 73.57% to 77.57%, and mean IoU from 97.81% to 98.35%, while false positives drop by 19.6%; the only parameter overhead (+3.84M) comes from BiFPN-GLSA, with BG-FDR and DADD adding effectively no network weights. Ablation on a self-built 3800-image dataset reveals a non-monotonic AP–Precision relationship: the mAP-optimal configuration (BG-FDR+BiFPN-GLSA, 94.17%) and the Precision-optimal one (DADD+Wise-IoU, 77.54%) do not coincide, providing a quantitative basis for objective-driven module selection in industrial sorting. Full article

Normative AI assistants are increasingly used in domains governed by duties, permissions, prohibitions, exceptions, priorities, and institutional policies. Existing retrieval-augmented generation (RAG) and legal AI benchmarks evaluate answer accuracy, retrieval quality, citation grounding, natural-language inference, clause extraction, or general legal reasoning ability. These dimensions are necessary but insufficient when supplied evidence is incomplete, mutually inconsistent, or defeasible. The objective of this study is to introduce ParaTraceBench, a paraconsistent trace-based benchmarking framework for post-retrieval normative reasoning over fixed evidence packages. Each scenario contains a query, evidence fragments, extracted facts, defeasible rules, typed attack edges, priority relations, an expected conclusion status, and a gold diagnostic trace. The formalism uses evidence-grounded arguments, a single edge-based attack representation, explicit attack-licensing rules, acyclic priority bases with a transitive closure, grounded argument labeling, trace-normal-form alignment, and deterministic scoring. The operational NER metric is explicitly interpreted as inconsistency-conditioned unsupported-conclusion avoidance rather than proof of logical non-explosion. We evaluated the framework using 140 scenarios, external validation on 567 anonymized Russian-language cases from Russian Federation and EAEU-related materials, reasoning-oriented baseline adaptations, five-run prompt-fairness and stability controls, and a deterministic component-dependency audit. On the full external set, the trace-based configuration reached 85.7% answer-status accuracy, 85.5% contradiction-localization accuracy, 94.2% operational NER, 84.1% priority-handling accuracy, and 83.7% belief-revision accuracy. These results indicate that contradiction-aware trace evaluation provides diagnostic information beyond final-answer accuracy under the evaluated fixed-evidence conditions, while not establishing causal architectural superiority, logical non-triviality, or end-to-end RAG performance. Full article

This paper investigates a segmented reconfigurable intelligent surface (RIS)-enabled backscatter communication riding over ambient non-orthogonal multiple-access (NOMA) signals. For a practical hardware, a simultaneous adjustment of the reflection coefficients of the RIS elements in phase and continuously in amplitude is physically not feasible, contradicting the conventional symbiotic approach of continuously adjusting the amplitude of the reflection coefficient from 0 to one. To address this bottleneck, a novel on–off mechanism of the RIS’s reflecting elements for symbiotic backscatter NOMA systems is proposed. To begin with, the coexistence outage probability and the ergodic capacity of the proposed system are analyzed for diverse dispersed end users and the corresponding performance boundaries in the high signal-to-noise-ratio (SNR) regime are subsequently characterized. In addition, Monte Carlo simulations are provided to verify the correctness of the proposed analytical framework. Finally, the numerical results show that the transmission effectiveness of the proposed on–off mechanism approaches that of the ideal continuous one with an increase in the number of RIS elements. The findings also reveal that the proposed on–off mechanism offers the advantage of reduced reflection coefficient control vis-à-vis the deployment of the underlying RIS-enabled symbiotic backscatter system without the need to adjust the reflection coefficient in amplitude and in phase simultaneously. Full article

Since the 17th century, when illustrations began to appear in children’s schoolbooks, the role of illustrations has undergone a development in concept. It has shifted from a narrow view of illustrations demonstrating the text, as well as lightening up the textual monotony, to a broader view of illustrations as a way of perceiving an additional interpretational layer of the meanings inherent in the text. Concurrently, the hierarchy indicating the supremacy of the written text over the illustrations has been undermined, opening the way to viewing children’s stories as a literary work in which the text and the illustrations conduct a complex relationship, together affecting the addressees’ experiencing of the meaning. After five decades of analyzing picturebooks, there are still some gaps between the research on children’s literature and the research on illustrations, whereby the study of children’s literature mainly deals with the textual aspects of works. The present research offers a model examining the functions of illustrations in children’s literature. The model includes a spectrum of layers located on an axis, which may be termed “the axis of influence”. This spectrum begins with illustrations that are limited to a simple description and adhere to the text, and it reaches illustrations that contrast with the text, accompanying it as a “counterpoint”. The illustrations’ modes of function are divided into three groups. The first group is that of illustrations describing the text faithfully. The second group includes illustrations which expand the meaning by providing layers that do not exist in the text: illustrations describing the space in which the protagonists are located; illustrations portraying the inner world of the protagonists; illustrations explicating complex concepts or themes; inclusive illustrations expanding the cultural and social variety of the characters in the story; and illustrations which promote social or national values. The third group includes illustrations subversive to the text, often even contradicting it or promoting a message that does not necessarily adhere to the contents of the text, thereby providing an independent narrative. The existence of an independent narrative leads to a poetic gap, by which there are two narratives dwelling in the same story. The research examines four types of subversive illustrations: illustrations creating gender correction; the implicative illustrations containing a visual subtext; illustrations casting a new light upon the story’s ending; and illustrations voicing a counterpoint to the content delivered in the text. Full article

Predicting CRISPR-Cas9 guide RNA efficiency and off-target activity is a precondition for precise genome editing. Computational models have progressively incorporated chromatin accessibility and epigenetic descriptors into their feature sets, yet synthesising findings from independently published studies—especially when those studies contradict one another—remains an unresolved methodological gap. Large Language Models (LLMs) have been proposed as a route to automate cross-study synthesis, but their utility depends on a constraint that receives less attention than model architecture: how much of the source text actually reaches the model at inference time. Cloud-based models process 48,000-token corpora without hardware limitations, but at the cost of data leaving the local environment and with limited reproducibility across API versions. Local RAG systems avoid the cloud dependency while fragmenting the input, discarding the global context needed to link biological arguments that are distributed across separate papers. We benchmark these strategies using a corpus of four CRISPR-Cas9 efficiency prediction studies and apply the Reduced Interaction Sampling (RIS) engine—a local sparse attention method—to retain the full sequence within the memory envelope of a laboratory server. Preserving that context uncovers three latent inconsistencies. The static epigenetic markers used in DeepCRISPR (CTCF, DNase I) show near-zero Spearman correlations with off-target cleavage ($\rho \le 0.07$), while nucleosome positioning scores from the Block Decomposition Method reach $\rho =0.388$–$0.423$. The sequence-only Apindel model was published in June 2022 without incorporating nucleosome descriptors reported in the concurrent literature. The benchmark review by Konstantakos et al. attributed 10–20% of rank correlation to epigenetics—a figure that reflects the weak feature subset evaluated, not a ceiling on chromatin influence. These discrepancies are invisible when papers are read individually or retrieved as chunks; they become traceable only when the full corpus is processed as a single context window. An independent empirical analysis of 2000 CRISPR-Cas9 off-target cleavage events provides evidence consistent with this pattern: static epigenetic markers yield $\left|\rho \right|\le 0.11$, whereas computed NuPoP Affinity descriptors reach $r=-0.622$ ($p<{10}^{-210}$). On a 30-question cross-study synthesis benchmark (5 independent seeds), baseline accuracy is 53.33%, RAG 60.00%, and RIS (30 seeds, 3% density) 70.00% ($p<0.0001$, t-test vs. RAG, $\sigma =0.00\%$ for all configurations). Full article

The contradiction between high compliance and low load-bearing capacity of flexible manipulators limits their engineering applications. Meanwhile, the theoretical modeling of the deformation and variable stiffness characteristics of flexible joints still faces considerable challenges. This paper proposes a positive-pressure double-airbag gap-constrained particle-jamming variable stiffness method and develops a novel actively variable stiffness pneumatic flexible bending joint with an integrated configuration of actuator, variable stiffness device (VSD), and primary structure. Based on classical elasticity theory and Coulomb–Amontons’ law of friction, theoretical models for the bending angle and tangential stiffness are established and verified through prototype experiments. With VSD activation, the joint reaches a bending angle of 56.35° at 0.4 MPa. At 40° forward bending, VSD activation increases the tangential stiffness from 0.167 N/mm to 0.832 N/mm, with the stiffness ratio between 40° and 0° increasing from 1.56 without VSD to 4.80 with VSD activation. Model predictions agree well with experimental data, yielding mean relative errors of 6.77% for the bending-angle model with VSD and 6.76% for the forward tangential-stiffness model with VSD activation. A coupling effect between bending deformation and stiffness is observed. The results demonstrate that the proposed joint achieves substantial stiffness regulation, providing a basis for its application in flexible robotic systems. Full article

Shoulder pain ranks among the leading causes of functional decline in older adults, affecting approximately one in five community-dwelling individuals and contributing to loss of independence and reduced quality of life. Despite its prevalence, shoulder pain in this population is frequently underreported and poorly understood beyond structural pathology, limiting early identification and targeted intervention. This narrative review synthesizes current evidence on the associations between four interrelated physical domains, body composition, grip and shoulder muscle strength, cervical–shoulder range of motion (ROM), and postural alignment, with particular emphasis on forward head posture and shoulder pain and disability in community-dwelling adults aged 60 years and older. A structured search of major electronic databases, including PubMed and Web of Science, was conducted using predefined eligibility criteria and keyword combinations linking shoulder pain and disability with each physical domain. Additional primary studies were identified through manual reference screening of key articles and citation tracking. Evidence was synthesized thematically, highlighting patterns of association, methodological inconsistencies, and contradictions across studies. Findings indicate that higher adiposity, reduced lean mass, shoulder and grip strength deficits, restricted glenohumeral and cervical ROM, and postural deviations are each associated with greater shoulder pain and disability. However, the evidence is not uniformly consistent: associations between Body Mass Index and shoulder ROM are largely absent at moderate adiposity levels, resting postural deviation does not reliably distinguish individuals with shoulder pain from those without, and the causal direction between strength deficits and postural deterioration remains debated. Despite these contradictions, emerging evidence suggests that these domains may interact with one another, and the literature collectively supports an integrated, multifactor assessment framework over single-variable approaches for the evaluation and management of shoulder dysfunction in community-dwelling older adults. Evidence quality was not formally assessed, and evidence for forward head posture specifically in older adults remains limited and largely indirect, drawn mainly from younger populations. Full article