Abstract: Purpose: This article explains why early-stage ventures frequently display intense activity yet fail to achieve commercialization in institutionally complex environments. It argues that the problem is not simply resource scarcity or weak infrastructure, but whether product, market, and institutional validation become aligned over time. Design/methodology/approach: The study adopts a longitudinal, abductive, multi-venture design based on 15 early-stage technology ventures operating across African markets over a 12-month period. Drawing on milestone plans, quarterly progress reports, budget allocation records, and advisory or engagement records, the analysis traces how validation processes unfold, interact, and diverge across venture trajectories. Findings: Three recurrent outcome regimes emerge: commercialization, artificial progression, and stagnation. Commercialization occurs when product, market, and institutional validation advance in a coordinated and mutually reinforcing sequence. Artificial progression arises when ventures generate credible activity and visible advancement in one or more domains, yet fail to convert this momentum into commercialization because validation remains cross-domain misaligned. Stagnation occurs when ventures do not accumulate sufficient validation to build cumulative legitimacy. Across cases, sequencing capability, the ability to order validation efforts so that gains in one domain unlock gains in others, appears to be a critical differentiator. Originality/value: The article contributes by reframing venture progress as an alignment-dependent accomplishment rather than an activity count, theorizing artificial progression as a distinct structural condition, and introducing Institutionally Mediated Market Formation (IMMF) as a process-based explanation of commercialization under institutional complexity. The study extends entrepreneurship, legitimacy, and ecosystem research by showing that visible activity is an unreliable proxy for progress unless it becomes commercially convertible through cross-domain validation alignment.

Abstract: Today, the reuse of treated wastewater is considered an important and strategic driver for integrated and sustainable water and soil management in extremely arid desert regions, where significant constraints due to water scarcity, soil salinization, and the fragility of agricultural ecosystems within palm oases place a strain on all sustainable development policies. Through this study, we conducted a comprehensive evaluation of the performance of the treatment, as well as the constraints related to salinity and the implications for the land management of the activated sludge wastewater treatment plant located in the Timimoun desert oasis in southern Algeria. Through monthly monitoring over a 12-month period, we were able to perform an analysis of physicochemical, nutritional and microbiological parameters, as well as a seasonal analysis, in addition to calculating irrigation suitability indicators using first-order kinetic modeling of COD degradation. The results obtained showed high reduction rates for COD (90%), BOD5 (90,5%), and TSS (93.8%), confirming the resilience and effectiveness of biological treatment under very difficult and hostile climatic conditions. Furthermore, the ultraviolet disinfection process ensures microbiological quality that allows for reuse of treated water in agriculture. However, the residual salinity of this water remains a significant limiting factor for sustainable reuse, highlighting the need to integrate soil management strategies, crop selection, and irrigation management into regulatory frameworks for wastewater reuse. Therefore, this study provides us with important and useful scientific data for developing sound and sustainable water and land management policies in the harsh climate of Saharan oases.

Abstract: Community reintegration of formerly incarcerated individuals is one of the most pressing challenges confronting criminal justice systems worldwide. High recidivism rates, fragmented service delivery, stigma, and inadequate coordination among correctional agencies, social service providers, and communities collectively undermine successful reintegration outcomes. Artificial intelligence (AI) offers transformative potential to address these systemic deficiencies through data-driven risk assessment, personalised service matching, and continuous behavioural monitoring. However, no comprehensive, ethically grounded architectural framework currently exists that integrates these capabilities into a unified community reintegration platform. This paper proposes the AI-based Community Reintegration Integration Platform (AI-CRIP), a five-layer architectural framework designed to support the full reintegration lifecycle—from prerelease assessment through post-release community stabilisation. The proposed framework integrates machine learning-based risk classification, natural language processing (NLP) for needs extraction, K-nearest neighbour (KNN) service matching, predictive recidivism analytics, blockchain-based audit trails, and a human-in-the-loop caseworker review mechanism. A formal pseudo-algorithm details the core plan-generation pipeline, demonstrating how structured offender profiles are transformed into personalised, milestone-driven reintegration plans. The framework is evaluated against fifteen representative studies from the existing literature spanning risk assessment models, digital reintegration tools, fairness in algorithmic decision-making, and technology-assisted supervision. The proposed architecture advances the state of the art by synthesising these disparate research threads into a coherent, deployable platform that prioritises fairness, transparency, and individual dignity. Critically, while AI-based tools such as emotive robots, digital avatars, and immersive virtual reality environments have emerged as low-stakes social surrogates for individuals experiencing isolation and withdrawal, they remain limited in their capacity to cultivate genuine human intimacy. Lasting reintegration therefore demands that technological aids be balanced by structural reforms addressing work-life balance, social inclusion, and community belonging, recognising that even highly personalised AI cannot substitute for the human connection that effective rehabilitation ultimately requires. Key technical, ethical, and policy challenges—including algorithmic bias, data privacy, digital inclusion, and stakeholder trust—are also discussed, with directions for future empirical validation. This work contributes a blueprint for practitioners, policymakers, and technology developers seeking to harness AI responsibly in post- carceral rehabilitation.

Abstract: Measurements of charge-changing cross-sections were developed as a method for determining proton radii, particularly for unstable, short-lived nuclei. Such cross-sections must be measured with high precision to determine the precise charge radii. However, there are complexities in the experimental method and leading to uncertainties in determining precise nuclear radii. Therefore, good models describing the complex physics of charged-particle interactions are needed in order to validate the experimental method and to estimate the contribution of systematic uncertainties. GEANT4 is a Monte Carlo simulation code that can describe elementary-particles and heavy-ion interactions in a broad range from typical atomic to cosmic-ray energies. An experiment has been performed to measure charge-changing cross-section of carbon isotopes, namely 10,11,12C nuclei, on different secondary reaction targets using the fragment separator FRS at GSI, Darmstadt. This work presents a comparison between the measured spectra of that experiment and the corresponding GEANT4 simulations.

Abstract: We present GaroNMT, a neural machine translation system for Garo (ISO 639-3: grt), a Tibeto-Burman language spoken by approximately one million speakers in Northeast India, written in a Latin-based orthography with distinctive glottal stop and nasal characters. Despite its significant speaker population, Garo has received virtually no attention in NLP research, with no publicly available machine translation system prior to this work. We fine-tune NLLB-200-distilled-600M by introducing a custom language token (grt_Latn), and conduct a systematic ablation study across six configurations including zero-shot baselines, gold-only fine-tuning, and combined LLM-backtranslation and gold training, with and without continued pretraining (CPT) on Garo monolingual data. Our best model (B1: fresh base, BT + gold) achieves BLEU 14.06 / ChrF++ 51.38 in-domain and BLEU 16.50 / ChrF 54.52 out-of-domain in the English-to-Garo direction, and BLEU 29.50 / ChrF++ 49.23 in-domain and BLEU 45.37 / ChrF 60.15 out-of-domain in the Garo-to-English direction, compared to a zero-shot baseline of BLEU 0.23 / 0.56 respectively. We find that CPT on 55,623 Garo sentences converges to near-zero loss within one epoch and provides no downstream translation benefit. We additionally identify and document a Garo-specific evaluation challenge: Unicode interpunct inconsistency between U+00B7 and U+2219 artificially suppresses automatic metrics and requires normalisation before scoring. We release our 15,441-pair gold parallel corpus and models under CC-BY-4.0 to support future NLP research on Garo and related Northeast Indian languages.

Abstract: We study the SU(N) symmetric model, which describes interaction of gravity with three field multiplets: first, the multiplet of pseudoscalar fields, which is now associated with the multi-component cosmic dark matter; second, the multiplet of vector fields, which represents the so-called color aether, now known as dynamic aether; third, the multiplet of Yang-Mills fields, which provides the SU(N) invariance of the model as a whole. It was previously known that the decay of the color aether in the early Universe could have given rise to emergence of an axionic singlet according to the Peccei-Quinn mechanism; we proposed an extended scheme, according to which the color aether activates an additional internal tool for generating not only a simple axionic singlet, but a whole SU(N) symmetric multiplet of pseudoscalar fields. Late-time evolution of the considered field configuration is analyzed in the framework of Bianchi-I cosmological model, and a hypothesis is proposed that the mentioned pseudoscalar multiplet can be associated with the multi-component cosmic dark matter.

Abstract: We report a statistically significant detection of dihedral D₃ symmetry in the Planck PR3 temperature anisotropy data, validated across all four independent component-separation pipelines (SMICA, NILC, SEVEM, Commander). At a single optimized axis (ℓ, b) = (50.3◦, −64.9◦), the power fraction in the A₂ (reflection-antisymmetric) irreducible representation exceeds isotropic expectations with a two-tier structure: a dense cluster at l ≤ 15 (Fisher PTE = 4.2 × 10⁻³ to 1.2 × 10⁻² across maps), driven by three multipoles significant in all four pipelines, with l = 3 serving as the axis-registration multipole (f_A2 = 0.94, z > 4.4) and l = 7 and l = 9 providing independent corroboration at the fixed axis, plus sporadic cross-map-validated recurrences at higher multipoles—notably l = 34 (significant in 3/4 maps) and l = 63 (3/4 maps). The A₂ excess draws power specifically from the E (rotation-doublet) irrep with anti-correlation r = −0.81, while the A₁ (trivial) irrep is decoupled. Extension to lmax = 150 with NMC = 10,000 simulations shows that the aggregate high-l Fisher PTE is consistent with isotropy (PTE > 0.91), but individual multipoles punctuate this null background. Among the nine strongest cross-map-consistent peaks, none belongs to the l ≡ 2 (mod 3) residue class (p ≈ 0.02 under uniformity), consistent with the C₃ selection rule. Cross-map correlations of f_A2 (l) exceed r = 0.93 for all pipeline pairs (SMICA–NILC: r = 0.997), ruling out component-separation artifacts. A null test on E-mode polarization at the same axis returns Fisher PTE = 0.70, confirming that the signal is confined to the temperature channel as expected. The irrep redistribution is sharply parity-gated: all four maps confine the A2 collecting signal to odd-l multipoles (Fisher p ≤ 2 × 10⁻⁴), with even-l entirely null (p > 0.97). Crossing parity with residue class produces a six-cell grammar dominated by a single cell (odd, l ≡ 0 (mod 3)), with step-function onset at l = 3. Singular-value decomposition reveals that this 2 × 3 grammar admits an approximate rank-1 factorization into a binary parity selector and a D₃ residue routing vector, recovered independently by all four pipelines (rank-1 fraction > 94% in three of four maps). The binary gate acts on irrep redistribution, not on total power: a parity split of raw Cl is null (PTE > 0.61) in every map. The signal morphology—dense at large angular scales with isolated resonances at smaller scales—is consistent with a parity-gated boundary condition on the acoustic eigenvalue problem whose geometry is fully resolved only at l ≲ 15 (θ ≳ 12◦). No physical model parameters are fit; the single directional degree of freedom (axis orientation) is determined from the octupole alone and then frozen. Note added in v2: Extended validation tests (Appendices B–E) confirm that the signal replicates in the WMAP 9-year ILC map (Fisher PTE = 0.0025), is uniquely selected among dihedral groups D₃–D₆, is frequency-independent across Planck HFI channels (100–143 GHz ∆ f_A2 correlation r = 0.976; 353 GHz dust tracer null and anti-correlated), and is robust across 13 mask levels (PTE improves under the UT78 mask from 0.008 to 0.005).

Abstract: Background Long-term stent healing after primary PCI of culprit unprotected left main (ULM) lesions is insufficiently explored. In this setting, large vessel size and bifurcation anatomy may limit angiographic stent optimisation and contribute to persistent strut malapposition and incomplete coverage. Objectives To identify OCT-derived geometric and healing parameters associated with long-term strut coverage and malapposition after angiography-guided primary PCI of culprit ULM lesions. Methods This single-center exploratory study included 30 patients with long-term OCT follow-up after angiography-guided primary PCI of culprit ULM lesions. OCT analysis was performed separately in three prespecified subsegments: the left main (LM), polygon of confluence (POC), and distal main branch (dMB). Five predefined strut-level healing outcomes were analysed: covered struts, malapposed struts, malapposed and uncovered struts, significantly malapposed struts (>400 μm), and significantly malapposed and uncovered struts. Associations between patient-level healing outcomes and OCT-derived measures of lumen geometry, stent dimensions, neointimal response, and an exploratory lumen–stent mismatch variable were assessed using univariable and multivariable linear regression. Results A total of 31,703 struts were analysed. Overall strut coverage was 90.7 ± 6.6%. Compared with the dMB, proximal ULM segments (LM and POC) showed lower strut coverage (82.8% and 84.2% vs. 93.9%, p< 0.001) and higher malapposition rates (17.4% and 14.2% vs. 0.4%, p< 0.001). In regression analysis, larger native lumen dimensions were associated with lower strut coverage and higher malapposition, whereas larger achieved stent area was associated with better strut healing. The exploratory lumen–stent mismatch variable was independently associated with all five healing outcomes in multivariable models (all p < 0.01). Conclusions After angiography-guided primary PCI of culprit unprotected left main lesions, long-term strut healing was significantly influenced by the mismatch between native reference lumen area and the achieved minimum stent area. Whether intravascular imaging–guided optimization of stent sizing and expansion in large-calibre left main anatomy improves strut healing requires further investigation.

Abstract: The effect of periodic temperature variation on short interacting RNA strands is demonstrated using three pairs of competing RNA duplexes as examples. The molecular interaction kinetics are simulated based on the experimental thermodynamic data obtained by M. E. Christiansen et al., as well as on the Eyring theory. The simulated time developments demonstrate the impact of shifting reaction kinetics and a state of continuous non-equilibrium. These factors provide a perpetual energy source for selection processes, creating ideal conditions for an ongoing molecular evolution.

Abstract: Speech in autism spectrum disorder (ASD) carries distinctive acoustic signatures that can offer valuable insight into the nature of autistic communication and its identification. Among these, vowel production remains insufficiently understood, despite its central role in speech intelligibility. This study investigates whether ASD is associated with systematic differences in vowel production. Eighteen Cypriot Greek adults with ASD and 18 peers with neurotypical development (ND), comparable in age, gender, education, nonverbal IQ, and verbal fluency, completed a controlled reading task. Participants produced disyllabic pseudowords embedding the five Greek vowels across four stress-syllable contexts. Acoustic analyses measured vowel-space organization, static spectral properties (F0, F1, F2, F3), dynamic trajectories (ΔF0, ΔF1, ΔF2, ΔF3), duration, and voice-quality indices (jitter, shimmer, harmonic-to-noise ratio [HNR], intensity). Bayesian models were used to evaluate group and vowel-specific differences. The results revealed a larger vowel-space area and greater vowel-space dispersion in the ASD group relative to the ND group, indicating a more expanded and dispersed acoustic vowel system. Group differences in individual acoustic measures were mostly selective rather than global: the clearest effects emerged in vowel-specific patterns of pitch, formants, and some dynamic formant measures. By contrast, duration, jitter, shimmer, and intensity did not show robust vowel-specific group differences. Among voice-quality measures, HNR showed the most consistent group difference, with ASD speakers showing higher HNR across all vowels. These findings challenge the notion of a single, uniform autistic voice, instead demonstrating that autism-related speech differences are multidimensional, vowel-specific, and language-sensitive. They therefore underscore the critical importance of segment-focused, cross-linguistically grounded approaches for advancing theory, assessment, and future speech-based identification in autism research.

Abstract: Urban regeneration has become a central focus in global urban studies, increasingly linked to the dual imperatives of sustainability and urban resilience. Chinese urban villages (Chengzhongcun) and Spanish Suelo Urbano no Consolidado (SUNC, Unconsolidated Urban Land) areas represent two contrasting forms of urban socio-spatial systems engulfed by urban expansion—both characterized by dense, historically rooted morphologies and incomplete infrastructure. While Chinese urban villages retain collective land ownership and self-built structures, SUNC areas preserve working-class housing typologies and community social structures within a sophisticated legal framework. As China shifts from a demolition–reconstruction model toward more sustainable regeneration approaches, this study compares Beijing's Cuigezhuang with Málaga's El Perchel through spatial analysis and stakeholder surveys. The research evaluates how differing planning systems foster or constrain sustainable development alongside social, spatial, and institutional resilience in regeneration processes. Findings demonstrate that Spain's incremental, participatory approach—anchored in Planes Especiales de Reforma Interior (PERI, Special Plans for Inner Urban Renewal) and land readjustment (equidistribution) mechanisms—significantly outperforms China's state-led demolition-based model in supporting long-term sustainability, heritage integrity, community cohesion, and spatial continuity. Spain's legally embedded participation and in situ rehabilitation strategies offer transferable lessons for China's evolving sustainable and resilience-oriented regeneration paradigm.

Abstract: This work introduces the Steepest Descent Evolution Principle (SDEP), a general variational framework that explains how systems evolve by following the path of steepest energy decrease. The principle is formulated in a broad mathematical setting, encompassing normed vector spaces, Banach spaces, and Hilbert spaces, where functional derivatives and gradients provide the foundation for its dynamics. Using Dirichlet energies as test cases, we show that the SDEP naturally recovers classical diffusion laws: the heat equation in the continuum and diffusion equations on graphs governed by the graph Laplacian. These results highlight the unifying power of the principle, offering a simple recipe for deriving dynamical equations across different contexts. Beyond classical physics, the framework opens avenues for applications in data science, network dynamics, and optimization, where energy-based models and steepest descent play a central role.

Abstract: This study examines the impact of data augmentation on machine learning perfor-mance, focusing on how synthetic data influences various neural network architec-tures. Common issues such as limited data, class imbalance, and poor coverage often lead to low model metrics, and data augmentation is frequently used to address these problems. The research aims to identify the optimal proportion of synthetic data, assess its effects across different architectures, and analyze the impact of augmenting only specific classes in a multi-class medical image classification task. Twelve widely used architectures were selected for the experiments, including classical convolutional networks, visual transformers, and the hybrid ConvNeXt model. Results showed that no universal optimal augmentation ratio exists, as model robust-ness to synthetic data varies, even within the same architecture family. Transformer and hybrid models demonstrated greater stability, while convolutional networks exhibited inconsistent behavior, likely due to higher sensitivity to data bias.

Abstract: Severe COVID-19 follows a cliff-edge trajectory: patients appear stable, then deteriorate rapidly and irreversibly. This paper identifies molecular oxygen as the dual control variable governing two previously unconnected biological systems: the DUOX-Lactoperoxidase-Iodine (DLI) airway antiviral defense and HIF-1α, the transcription factor that drives COVID-19 severity. Both share the same oxygen-dependent enzymes (DUOX and PHD, Km approximately 20 μM O₂ corresponding to approximately 94% SpO₂). When SpO₂ falls below this threshold via AT2 cell destruction with surfactant loss, ventilation-perfusion mismatch, and microvascular thrombosis, both systems fail simultaneously, initiating three concurrent cascade arms: (1) collapse of DLI mucosal defense through O₂ substrate depletion; (2) HIF-1α-driven Furin upregulation accelerating viral spike cleavage and entry, with a viral amplification feedback loop; and (3) IL-6-mediated cytokine storm depleting thyroid iodide reserves. These three arms interact multiplicatively, not additively. A Monte Carlo simulation across four populations demonstrates a 40.3% steeper cliff-edge signature than an additive null model. The framework generates three falsifiable clinical predictions and identifies supplemental oxygen initiated before the HIF-1α threshold (SpO₂ 94–95%) as the primary actionable intervention, suppressing all three cascade arms simultaneously.

Abstract: Gravelly soil is widely distributed in the central and western regions of China and serves as a crucial fill material for transportation infrastructure. However, its poor gradation, poor water stability, and low freeze - thaw resistance limit its direct application. To address the problems of high energy consumption and high carbon emissions of existing solidifying agents (such as cement, lime) and achieve the resource utilization of waste foam concrete, this study took waste foam concrete as the raw material, prepared a novel gravel soil stabilizer through crushing, ball milling, and high - temperature calcination, and systematically studied the solidification performance (unconfined compressive strength, water stability, freeze - thaw resistance) of the prepared stabilizer on gravelly soil and its solidification mechanism. The results show that the prepared stabilizer can significantly improve the mechanical properties of gravelly soil. At a dosage of 30%, the unconfined compressive strength reached 6.5 MPa after 28 days, an increase of 333% compared to the control group. The water stability is enhanced with the increase of dosage, and the water stability coefficient is significantly improved at a dosage of 30%. In terms of freeze - thaw resistance, at a dosage of 30%, the mass loss rate was only 2% after 5 freeze - thaw cycles, and the unconfined compressive strength reached 9.56 MPa, an increase of 437% compared to the control group. XRD and SEM analysis indicate that the stabilizer generates cementitious products such as calcium silicate hydrate gel and katoite through hydration reactions, which fill the pores of gravelly soil, cement particles, and optimize the microstructure, thereby improving its mechanical properties, water stability, and freeze - thaw resistance. This study provides a new way for the efficient resource utilization of waste foam concrete and also offers a low - energy and environmentally friendly novel stabilizer for the reinforcement of gravel soil subgrades in cold regions.

Abstract: Hemagglutinin (HA) is the primary surface protein of the influenza A virus, determining its subtype and antigenic properties. Traditional subtype classification methods rely on DNA or amino acid sequence analysis, which does not account for protein spatial folding. In this work, we propose EpitopeGNN — a graph neural network (GNN) that constructs a residue interaction network (RIN) from the 3D structure of HA and classifies the virus subtype. The model was trained on 249 structures from the Protein Data Bank (PDB), containing H1N1, H3N2, H5N1, and other subtypes. By utilizing physicochemical properties of amino acids and topological centrality measures, we achieved 100% classification accuracy on the test set and 97.6% with five-fold cross-validation. A significant correlation was found between the obtained structural embeddings and phylogenetic distances (r = 0.48, p < 0.001), confirming their biological relevance and opening opportunities for structural monitoring of virus evolution, as well as rapid analog searching for novel strains.

Abstract: We study shell kernels for the odd-to-odd Syracuse dynamics generated by uniformly distributed initial windows. For backstepped first-passage shells, we prove short-time localization, derive an exact inverse-affine representation of the fixed-time kernel, and reduce the shell-slice discrepancy to weighted primitive-frequency correlations. We also prove a quantitative boundary-layer estimate and identify a formal renewal model for the corresponding shell mechanism. On the arithmetic side, we obtain an exact block decomposition for the primitive-frequency transfer operator, prove that no naive operator gap is available, and reduce the unresolved step to explicit incomplete principal-unit exponential sums modulo powers of 3. Thus the paper is unconditional up to a final primitive-frequency estimate, which is formulated explicitly.

Abstract: In unmanned pharmacy and home-care medicine management applications, reliable pillbox localization is a prerequisite for automated dispensing and grasping. However, existing detectors still perform poorly in complex environments where dense stacking, occlusion, weak illumination, and high inter-class similarity occur simultaneously. To address this problem, GSPM-YOLO is proposed as an improved detector built on the YOLOv11 framework for complex pillbox recognition, and four novel plug-and-play lightweight modules are developed: GSimConv, a lightweight dual-branch convolution module that incorporates the Attention Weight Calculation Algorithm in HardSAM for edge-preserving feature extraction, PSCAM for position-sensitive coordinate attention, MSAAM, a multi-scale strip-pooling module that integrates the Horizontal Context-Aware Attention weight calculation algorithm to strengthen occluded targets, and LGPFH for bidirectional ghost pyramid fusion. To simulate the complex operating environments of dispensing robots, we construct MBox-Complex, a dataset of 3{,}041 images with 8{,}153 annotations across 25 drug categories. Ablation experiments first validate the effectiveness of the four-module composition, with F1 rising from 0.641 to 0.714, and each module is then individually compared with advanced replacement schemes in dedicated substitution experiments to verify its own effectiveness. The integrated model is then benchmarked against advanced detectors and domain-specific methods on the self-constructed MBox-Complex dataset, achieving 0.727 mAP@50 and 0.427 mAP@50-95 with 3.8M parameters and surpassing YOLOv11 by 7.1 and 4.0 percentage points and YOLOv12 by 4.3 and 3.1 percentage points, respectively. Further cross-dataset evaluation on the VOC and Brain Tumor benchmark datasets verifies the transferability of the proposed model. Grad-CAM is adopted to visualize the detector's attention distribution, and the resulting heatmaps together with detection visualizations confirm that the proposed model focuses more precisely on stacked and occluded regions.

Abstract: This study proposes a novel hybrid prediction model (QGCN-LSTM) that combines Quantum Graph Convolutional Networks (QGCN) with classical Long Short-Term Memory (LSTM). The model takes space-time data as input and achieves quantum information conversion through a quantum encoding layer. Multi-scale features are extracted through the collaborative computation of QGCN and quantum gated loop units, and a quantum attention module is introduced to dynamically screen key information. Finally, the prediction results are generated through quantum measurement and a classical output layer. In the space-time data prediction task of urban traffic flow, a benchmark model system covering classical, cutting-edge, and traditional architectures was constructed. The experimental results show that QGCN-LSTM utilizes quantum entanglement gates to establish non-local road network associations, dynamically allocate feature weights to enhance the impact of critical time steps, and achieves deep compression of lines through quantum line pruning technology, effectively alleviating the common problem of “poor plateau” in quantum neural network training. In terms of prediction accuracy, the average absolute error (MAE) of its key hub nodes is reduced by 34.1% compared to the graph convolution LSTM (GCN-LSTM) model, and the Spatial Correlation Index (SCI) is improved to 0.89. In addition, it also shows excellent performance in dynamic response, edge computing efficiency, and other aspects, meeting the real-time requirements of the traffic signal control system. This study provides an effective paradigm for the application of quantum collaborative architecture in complex spatiotemporal prediction tasks.

Abstract: The nontrivial zeros of the Riemann zeta function are parameterized by the spectral variable \( s\in\mathbb{C} \), and the isomonodromic deformation parameter t of the Painlevé III equation of type \( D_6 \) is connected to s by \( t=s(1-s) \), which maps the critical line \( \Re(s)=\frac12 \) to the positive real ray \( t\in[\frac14,\infty) \). Any de Branges realization of the Riemann Hypothesis within this framework requires four explicit conditions: (C1) geometric feasibility ---the positive lambda-length slice of the \( \mathrm{PIII}_{D_6} \) character variety defines a real form of the wild Stokes and monodromy data; (C2) global positivity---the Riemann--Hilbert jump matrices yield a Herglotz Weyl--Titchmarsh function; (C3) embedding compatibility---the functional equation involution \( s\mapsto 1-\bar{s} \) preserves the positive slice; and (C4) analytic regularity---the tau-function composed with \( t=s(1-s) \) is entire of finite order after gauge removal. We prove all four conditions unconditionally. For (C1), an explicit birational map \( \Phi \) expresses all Stokes multipliers as positive monomials in the lambda-lengths. For (C2), the Painlevé/gauge theory correspondence identifies the \( \mathrm{PIII}_{D_6} \) oper with a Schrödinger operator whose real coefficients force \( \Im m(\lambda,t)>0 \) via a Wronskian argument; isomonodromic uniqueness and Remling's inverse theorem complete the proof. For (C4), integrality of the local exponent \( \alpha\in\mathbb{Z}_{\ge0} \) is the precise criterion, satisfied on an explicit sublocus of the positive slice. With all four conditions established, the Riemann Hypothesis reduces to the Bridge Conjecture alone. We test the direct form of the Bridge Conjecture---the identification \( E_{D_6}(s)=C\,\xi(s) \)---and show it fails for all constant monodromy phases and for all Dirichlet L-functions, because the tau-zero counting \( \mathcal{N}_{D_6}(T)\sim 2T/\pi \) lacks the \( \log T \) factor of the Riemann--von Mangoldt law. This leads to the identification of \( E_{D_6}(s) \) as a new explicit element of the Hermite--Biehler class \( \mathcal{HB}(1/2) \), whose canonical form is the isomonodromic cosine \( F(s)=\cos(2\sqrt{s(1-s)}) \). We prove that \( F\in\mathcal{HB}(1/2) \) is entire of order 1, satisfies \( F(s)=F(1-s) \), has all zeros on \( \Re s=\frac12 \) at \( \gamma_n=\sqrt{(2n-1)^2\pi^2-4}\,/\,4 \), with asymptotic spacing \( \pi/2 \) identified as the WKB semiclassical level spacing of the \( \mathrm{PIII}_{D_6} \) oper arising from the Seiberg--Witten period \( a_{D_6}(t)=2\sqrt{t} \). A four-tier falsifiability diagnostic and the character \( \chi_4 \) scorecard are presented.