Search Results (3,381)

Against the background of climate change intensifying extreme hail events, the photovoltaic module industry faces a critical trade-off between improving hail resistance and maintaining environmental sustainability. This study establishes an emergy–life cycle coupling assessment framework to systematically evaluate the environmental sustainability of six typical hail resistance enhancement designs across four hail risk scenarios in China. Five hierarchical hypotheses are proposed and validated through quantitative analysis. The optimal design point occurs at 30 mm hail resistance using 3.2 mm tempered glass, achieving a minimum unit environmental impact per impact resistance UEIC of 9.63 × 10¹² sej/mm. The ranking divergence index SDR between coupled emergy–LCA and conventional LCA methods is 0.267, with ecosystem service dependence ESD reaching 0.241 for composite backsheet designs, revealing natural capital overlooked by traditional methods. A complete ranking reversal occurs at a threshold hail frequency of 1.3 events per year, above which the 3.2 mm glass design outperforms standard modules with life cycle emergy input LCEA of 3.20 × 10¹⁴ sej versus 3.41 × 10¹⁴ sej under high-risk scenarios. Material type dominates environmental impact over structural parameters by a factor of 2.32, with recycled aluminum frames reducing ELCI by 12.4%. The dual-optimum design is identified as the 3.2 mm tempered glass scheme, achieving a combined sustainability score CSS of 0.782 and emergy yield ratio EYR of 3.86, outperforming the industry average of 3.61. Multi-objective optimization using NSGA-II yields a Pareto front of 12 non-dominated solutions, with the 3.2 mm glass design maintaining Pareto optimal status in 72% of Monte Carlo iterations. This research provides a quantitative decision-making framework recommending standard modules for regions below one annual hail event, the 3.2 mm glass design for regions between one and four annual events, and steel frame combinations above four annual events, demonstrating that moderate enhancement achieves the optimal balance between hail protection and environmental sustainability. Full article

Cohabiting companion dogs establish hierarchy among themselves. It is hypothesized that the owner represents the main and undividable resource, thus primary access to this is a main organizing factor of rank-related behaviors of dogs. Here we tested high- and low-ranking cohabiting companion dogs’ (N = 70) reactions to their owner’s absence in a 3 min separation test. Rank scores have been assigned with a validated questionnaire (DRA-Q). We predicted that dominant dogs would show stronger reactions to being separated from their owner. Indeed, we found that higher-ranking dogs showed more intense activity and sooner arising attempts to leave the room (rearing at the wall, scratching the door, moving around, barking) than lower-ranking dogs did. These reactions may show also the intention to reestablish their connection with the absent owner. The associations between dogs’ rank and the behavioral responses were modified by the dogs’ age (negatively), the number of cohabiting dogs (positively), and we found that subcategories of the dog’s dominant status (such as ‘agonistic’ and ‘leadership’ subscales) were also associated with finer details of the outcome. These are the first results indicating that presence of the owner may provide more reassurance to higher-ranking dogs against stress than it does to lower-ranking dogs. Full article

Background and Objectives: Elderly patients with head and neck tumours frequently present with multiple comorbidities and a potentially difficult airway, making general anaesthesia high-risk. Dexmedetomidine, a selective alpha-2 adrenoceptor agonist, provides conscious sedation without clinically significant respiratory depression, offering a compelling locoregional alternative. This study evaluated the haemodynamic profile, sedation kinetics, and satisfaction outcomes of a standardised dexmedetomidine-based protocol for head and neck surgery under local infiltration anaesthesia. Materials and Methods: A prospective, single-centre observational study was conducted at the University Hospital Centre Zagreb. Twenty-three consecutive adult patients received a continuous dexmedetomidine infusion at 0.5 μg/kg/h, initiated preoperatively in the post-anaesthesia care unit without a loading dose. Haemodynamic parameters, sedation-to-incision interval, cumulative dose, and postoperative patient and surgeon satisfaction (NRS 1–10) were recorded. Spearman rank-order correlation and the Mann–Whitney U test were used for statistical analysis. Results: The primary outcome of haemodynamic stability—defined as the absence of vasoactive or inotropic rescue—was achieved in all 23 patients (100%). The median cumulative dexmedetomidine dose was 52 μg (IQR 44–68 μg). Controlled hypotension was achieved in all patients, with a median nadir systolic blood pressure of 98 mmHg. Supplemental oxygen was required in only 2 of 23 patients (8.7%). Patient and surgeon satisfaction reached a median NRS score of 10 in both groups. The sedation-to-incision interval correlated with total drug dose (ρ = 0.74, p < 0.001), consistent with fixed-rate infusion pharmacokinetics. Hypertensive patients exhibited a greater reduction in systolic blood pressure (median 45 vs. 28 mmHg; p = 0.015). Conclusions: A fixed-rate dexmedetomidine infusion initiated in the post-anaesthesia care unit provides a feasible and potentially effective conscious sedation strategy for head and neck surgery under local infiltration anaesthesia in selected elderly and comorbid patients. In this pilot series, the protocol was associated with haemodynamic stability in all cases, low supplemental oxygen requirements, and high procedural satisfaction among both patients and surgeons. These findings are preliminary and require confirmation in larger, controlled studies. Full article

Background/Objectives: Triple therapy combining targeted therapy (lenvatinib/bevacizumab), immune checkpoint inhibitors (ICIs), and interventional therapy (TACE/HAIC) has shown promising efficacy, but clinical outcomes differ among patients. We developed and tested a prognostic model based on sex, ALBI grade, and ALR to estimate survival in patients with intermediate-to-advanced HCC receiving triple therapy. Methods: This single center retrospective study included 184 intermediate-to-advanced HCC patients between November 2017 and December 2024. The patients enrolled received lenvatinib (n = 88) or bevacizumab (n = 96) plus PD-1/PD-L1 inhibitors and interventional therapy. The risk scoring model was derived from univariate Cox regression, LASSO Cox regression, and multivariate Cox regression analyses that were screened for independent prognostic factors. The median risk score defined the cutoff for separating patients into two risk subgroups (high- and low-risk). Overall survival (OS) across subgroups was evaluated and compared by Kaplan–Meier analysis and log-rank test. Model performance was evaluated using the C-index, time-dependent AUC at 6, 12, and 24 months, calibration curves, the Brier score, and decision curve analysis, with internal validation performed via Bootstrap resampling. Results: Multivariate analysis identified male sex, ALBI grade 3, and a high ALR level as independent risk factors of poorer OS. The resulting risk model showed a C-index of 0.62. Moreover, the median OS differed significantly between the two risk groups (p < 0.001). The model displayed moderate discrimination, with AUCs of 0.66, 0.66, and 0.74 at 6, 12, and 24 months. Calibration and the Brier score showed reasonable agreement and acceptable prediction errors. No interaction between risk factors and treatment type was observed (p > 0.05), indicating model applicability to both lenvatinib and bevacizumab-based regimens. Conclusions: A prognostic model integrating sex, ALBI grade, and ALR can offer some capacity to stratify survival risk in intermediate-to-advanced HCC patients. However, its overall discriminative performance is limited, and further validation in larger and external cohorts is needed to confirm its clinical value. Full article

Fault-injection campaigns are widely used to evaluate silent data corruption (SDC) in AI hardware, but exhaustive campaigns over workloads, dataflows, processing elements, and datapath roles are expensive. This paper presents an architecture-level risk-guided fault-injection prioritization method for systolic AI accelerators. The method ranks candidate transient functional perturbations before downstream validation, with the goal of enriching the discovery of candidates that produce a thresholded relative-output-error outcome under a limited validation budget. The evaluation uses a fixed candidate fault pool: all ranking policies score the same 21,000 candidate faults across 30 workload/dataflow/array configurations, corresponding to five GEMM-derived workloads, three array sizes, and two dataflows. Fault magnitudes are sampled once per candidate and are independent of all ranking scores. Candidate faults are modeled as transient architecture-level perturbations in MAC, accumulator, or forwarding paths. The proposed full-risk score combines activity, composite spatial stress, tensor sensitivity, and a path-class weight. In the proposed architecture-level simulation environment and under the fixed-pool protocol, the proposed method achieves the highest mean top-10% SDC-proxy lift, AUPRC, NDCG@10%, and rank correlation with relative output error among the evaluated principle-based ranking policies. At the calibrated threshold, it achieves a mean top-10% lift of 5.65× [4.91, 6.38], compared with 4.61× for AVF-like exposure and 4.33× for output sensitivity. Paired configuration-level tests, threshold sensitivity, and outcome-model sensitivity analyses characterize the result while showing that the proposed score is not universally dominant under every synthetic outcome assumption. The method is intended as a front-end architecture-level screening tool for validation prioritization, not as a replacement for RTL, gate-level, FPGA, or silicon reliability signoff. Full article

Background: Dynamic risk stratification is fundamental to the modern management of pulmonary arterial hypertension (PAH). However, data on the impact of sequential add-on therapy in patients with Group 1.4 PAH—particularly in Latin American populations—remains limited. This study evaluated changes in risk classification using COMPERA 2.0 and REVEAL Lite 2 scores in patients treated with endothelin receptor antagonist (ERA) and phosphodiesterase type 5 inhibitor (PDE5i) combination therapy (macitentan + sildenafil) at a referral center in Mexico. Methods: In this single-center, retrospective cohort study, 25 patients with a confirmed diagnosis of PAH between 1st January 2022 and 31st December 2024 were evaluated at baseline and after 24 weeks of treatment. Clinical, functional, and biochemical parameters were recorded. Within-patient changes were analyzed using the Wilcoxon signed-rank test, and agreement between risk assessment tools was assessed using Spearman’s correlation coefficient. Results: At 24 weeks, patients demonstrated significant improvement in World Health Organization functional class (p = 0.002) and a significant reduction in brain natriuretic peptide levels (p = 0.003). Both COMPERA 2.0 and REVEAL Lite 2 scores showed a consistent shift toward lower-risk categories. A strong concordance between the two tools was observed. Conclusions: Sequential add-on ERA + PDE5i therapy was associated with meaningful improvement in risk stratification among patients with Group 1.4 PAH. These findings support the clinical utility of simplified, noninvasive risk assessment tools in real-world settings, particularly in resource-constrained environments. Full article

Background/Objectives: Sepsis detection relies on clinical variables and scoring systems assumed to perform uniformly across hospital settings. However, sepsis phenotype distributions shift between clinical environments, suggesting that variable importance may be setting dependent. This study aimed to quantify service-specific variability in the discriminatory capacity of clinical variables for sepsis detection and to evaluate whether this heterogeneity translates into differential performance of machine learning models compared to traditional clinical scoring systems. Methods: This stratified sub-analysis of the BIAlert Sepsis cohort (203,755 patients; 11,864 sepsis episodes, 2014–2018) evaluated 61 structured quantitative variables across nine hospital services (≥90 sepsis episodes each). Within each service, the Mann–Whitney–Wilcoxon test (p < 0.01, Holm-corrected) assessed differences between septic and non-septic episodes. Five machine learning models (Random Forest/BIAlert, XGBoost, CatBoost, SVM, Neural Network) and three clinical rules (NEWS, SIRS, qSOFA) were evaluated globally and stratified across four clinical environments. Results: The proportion of significant variables ranged from 95.1% in the Emergency Department (58/61) to 37.7% in the Intensive Care Unit (23/61). Lactate was the only universally significant variable (9/9 services). Clinical scoring systems collapsed in Critical Care (qSOFA and NEWS AUC 0.459). BIAlert maintained the highest AUC across all environments (0.975–0.857). The Friedman test confirmed significant differences (χ² = 28.00, p < 0.001), with BIAlert achieving a mean rank of 1.0. Conclusions: The discriminatory capacity of clinical variables for sepsis detection is not uniform across hospital services. ML models, particularly BIAlert, maintained robust performance where fixed-rule scoring systems failed. Full article

Background/Objectives: Following heart transplantation (HT), a subset of patients will require an early or late permanent pacemaker (PPM). We explored risk factors and outcomes associated with PPM implantation in this population. Methods: Using the United Network for Organ Sharing (UNOS) database, we identified all adult patients undergoing HT from 2013 to 2023 who received a PPM early (prior to discharge) or late (>6 months post transplant). Propensity score matching (PSM) was used for control cohorts was. Primary outcomes included recipient survival at 30 days and 1 and 5 years. Predictors of early and late PPM, as well as post-PPM mortality, were assessed using Cox and logistic regression models. Kaplan–Meier survival curves were compared using a log-rank test. Results: Following PSM, the early PPM cohort included 354 patients, and the late PPM cohort included 554 patients. Early PPM patients showed similar 30-day and 1- and 5-year survival (p = 0.582, 0.421, and 0.2844 respectively) but lower rates of graft failure (1.1% vs. 4%, p = 0.017) and primary graft dysfunction (PGD) (1.7% vs. 4.2%, p = 0.046). Late PPM patients had reduced survival at 30 days and 1 year but not at 5 years (p < 0.001, p = 0.0023, 0.050 respectively). Neither early nor late PPM was independently associated with increased risk of mortality after HT. Donation after Circulatory Death (DCD) organs were associated with a lower risk of early PPM (aOR = 0.409, p = 0.020). Late PPM patients showed higher rates of PGD (2.5% vs. 0.5%, p = 0.007). Conclusions: Early or late PPM is not an independent risk factor for mortality after HT, but differing short-term morbidity and mortality are observed. Full article

Significant progress in legal natural language processing (NLP) has enabled advancements in tasks such as legal judgment prediction, case retrieval, and question answering. However, the development of analogous technologies for Arabic legal texts remains severely constrained by the scarcity of large-scale, publicly available benchmarks for summarisation and classification. This paper addresses this gap by introducing a novel, comprehensive dataset of 9699 Arabic legal cases sourced from the Saudi Board of Grievances. This corpus is unique in pairing full-length court decisions with expertly human-crafted abstractive summaries and multi-class category labels (Administrative, Commercial, and Criminal), establishing a dedicated benchmark for Arabic legal NLP. The dataset was constructed via a robust, reproducible pipeline that ensures high textual fidelity, incorporating specialised optical character recognition (OCR) via Google Document AI and precise structural segmentation into facts, reasons, and summaries. To establish robust baselines, we conduct an extensive empirical evaluation of seven summarisation models—encompassing four extractive algorithms (TextRank, LexRank, Latent Semantic Analysis, and Luhn) and three transformer-based abstractive architectures (AraT5v2, AraBART, and mBART)—each evaluated in both base and fine-tuned configurations. Results across ROUGE, BERTScore, BLEU metrics and human evaluation demonstrate substantial performance gains achieved through domain-specific fine-tuning, with the fine-tuned AraBART model achieving the strongest performance among all evaluated models. Furthermore, we present a novel analysis of the downstream utility of generated summaries by evaluating their performance on legal category classification using five machine learning models. This investigation reveals a strong positive correlation between summarisation quality and classification accuracy, empirically demonstrating that domain-adapted abstractive summarisation not only enhances intrinsic evaluation scores but also significantly boosts extrinsic task performance. By providing this essential dataset and comprehensive benchmarking, our work contributes a much-needed resource to the field, facilitating future research and innovations in Arabic legal text analysis. Full article

Online apparel size recommendation remains difficult because consumers cannot physically assess garment fit before purchase. It is a multi-dimensional fit evaluation problem, particularly for complex garments such as jackets, where multiple body areas jointly influence perceived fit. Existing methods often rely on limited anthropometric measures, heuristic rules, or behavioral data, restricting both accuracy and interpretability. To address this issue, this study proposes an interpretable multi-dimensional fit evaluation framework based on garment ease theory. The framework defines ideal ease as the target fit condition and quantifies deviations through a segment-based weighting mechanism. Section-level mappings between body and garment measurements are established, and differentiated penalties are assigned according to the semantic fit interval of each body area. Section-specific evaluations are aggregated into an overall fit score (OFS) for candidate size ranking and Top-K recommendation, while also providing detailed fit feedback. Experiments involving 270 female participants and two jacket styles show high recommendation accuracy, achieving Top-3 accuracies of 99.6% for the regular-fit jacket and 98.9% for the tight-fit jacket. Compared with traditional heuristic methods, the proposed approach demonstrates clear advantages in both performance and interpretability, offering a practical solution that balances accuracy, transparency, and deployability. Full article

Background/Objectives: First-generation CDK4/6 inhibitors (palbociclib, ribociclib, abemaciclib) target the conserved ATP-binding pocket of CDK4 and, despite clinical success, are limited by acquired resistance and insufficient exploration of alternative regulatory sites. This study aimed to identify a putative allosteric small-molecule candidate at the CDK4 αE-helix–Cyclin D1 α1-helix protein–protein interaction (PPI) interface within the CDK4/Cyclin D1/p21 ternary complex using RapidFunnel-AI, a decision-interpretable virtual-screening pipeline. Methods: Starting from 50,000 ChEMBL 33 molecules, the pipeline sequentially applied a Q-Fold/RapidFunnel topological Tanimoto scan based on clinical CDK4/6 inhibitor motifs, fragment-level electronic-property enrichment, ADMET/PAINS filtering, dry Vina-GPU docking, hydration-mediated AutoDock-GPU (Version 1.6) docking, explicit-solvent molecular dynamics, contact-retention analysis, and MM-GBSA energy decomposition. The Q-Fold Thermo-Core surrogate model provided fragment-level enrichment, predicting the HOMO–LUMO gap (R² = 0.93) and isotropic polarizability (R² = 0.98) on QM9. Candidate selection did not rely on the lowest docking or MM-GBSA score alone, but on pose persistence, contact continuity, and energy-component consistency. Results: The workflow reduced the initial library to 43 topologically prioritized candidates, 25 ADMET/PAINS-filtered ligands, and 9 docking-derived complexes for MD validation. Ligand_020 emerged as the only candidate that preserved a persistent binding mode at Site 2 during a 500 ns simulation—an interface engagement reproduced across three independent 500 ns replicates with no full dissociation in any replicate—with a protein Cα RMSD of 2.88 ± 0.32 Å, a ligand heavy-atom RMSD of 3.56 ± 0.28 Å, and a van der Waals-dominated MM-GBSA profile (ΔGbind = −28.23 ± 3.57 kcal/mol). In contrast, palbociclib and ribociclib, forcibly placed at Site 2 as negative controls, lost most initial contacts within 5 ns and tended to detach despite more favorable MM-GBSA values. Conclusions: These results suggest that single-score docking or MM-GBSA ranking can generate false positives at shallow PPI interfaces. By integrating AI-assisted prioritization, multipocket docking, explicit-solvent MD, contact-retention analysis, and energy-component consistency, RapidFunnel-AI nominated Ligand_020 as an experimentally testable putative allosteric hit targeting the CDK4/Cyclin D1 interface, offering a reusable platform for PPI-focused oncological drug discovery. Full article

Intrusion detection systems (IDSs) are commonly evaluated through aggregate classification metrics, although operational workflows require detected flows to be interpreted, prioritized, and transformed into actionable evidence. This study proposes a detection-to-triage framework for multiclass intrusion detection using a CSE-CIC-IDS2018-derived experimental subset containing 213,463 records across one benign class and fourteen attack classes. The framework combines supervised multiclass classification, SHAP-style post hoc explanation, class-specific false positive analysis, and a Suspicious Flow Priority Score (SFPS) for analyst-oriented suspicious flow ranking. The practical role of SFPS is to reorder suspicious flows by combining model confidence, explanation strength, predefined attack severity, and validation-based false positive control, thereby producing a transparent triage list rather than a probability-only alert queue. Three detection backbones were evaluated under a shared preprocessing protocol: Random Forest, XGBoost, and a lightweight multilayer perceptron baseline. To assess stability, experiments were repeated across five random seeds. XGBoost achieved the strongest mean performance across most aggregate indicators, with an accuracy of 0.9494 ± 0.0011, a macro F1-score of 0.8366 ± 0.0193, a weighted F1-score of 0.9494 ± 0.0011, and a Matthews Correlation Coefficient of 0.9429 ± 0.0012. Random Forest produced closely comparable results, while the lightweight MLP remained lower on aggregate and macro-level indicators. False positive analysis showed that the alert burden was concentrated in selected classes and differed across models, confirming that aggregate performance alone is insufficient for assessing IDS usefulness. SHAP-style analysis identified stable flow-level contributors to XGBoost discrimination, while SFPS substantially changed the post-detection ordering of suspicious flows compared with probability-only ranking. The study does not claim universal state-of-the-art superiority, causal explanation, or deployment validation; instead, it demonstrates how multiclass IDS outputs can be extended into explainable, false positive-aware, and triage-oriented rankings for analyst review. Full article

Early stage mineral exploration requires the reliable identification of anomalous gold zones from drillhole geochemistry in data-limited environments. This study applies a machine learning (ML) classification framework to detect anomalous gold zones (Au ≥ 0.68 ppm; 90th percentile) from bulk XRF multielement drillhole geochemistry in a Paleoproterozoic Birimian greenstone belt sulphide-hosted orogenic gold system, West African Craton. A total of 53,126 one-metre diamond core samples from 301 drillholes were preprocessed within a compositional data analysis (CoDA) framework, with Au being explicitly excluded from the centred log-ratio (CLR) transformation to eliminate target–predictor circularity. After Minimum Covariance Determinant (MCD) outlier filtering, 40,385 samples were retained to construct a 19-feature matrix of 10 CLR-transformed elements, 1 rock-type feature, and 8 sulphide–lithology interaction features. Drillhole-based block cross-validation (DH-block CV), validated by an experimental along-hole variogram (practical autocorrelation range ≈ 20 m), ensured spatially honest performance estimates. Four nonlinear classifiers—Random Forest (RF), XGBoost, LightGBM, and Multi-Layer Perceptron (MLP)—were benchmarked against a Logistic Regression (LR) linear baseline. All nonlinear classifiers achieved validation AUC of 0.936–0.938, outperforming LR (AUC = 0.931) with F1-score improvements of +0.09 to +0.11 and precision gains of up to +35 percentage points—directly reducing wasted drill holes in applied exploration. MLP recorded the highest F1-score (0.666) and precision (0.765), and XGBoost the highest recall (0.787). Permutation importance identified S-Ti (ΔAUC = 0.028), S-Fe (0.021), and S-Al (0.013) as the top-ranked features, confirming that sulphide enrichment relative to lithological background is the primary discriminating signal. Partial dependence analysis revealed a threshold-driven non-monotonic Fe dependence at CLR(Fe) ≈ 3, marking the transition from lithological dilutant to sulphide co-indicator—a nonlinear pattern inaccessible to linear classifiers. Full article

Background and Objectives: This prospective cohort study aimed to assess patellar and tibial donor-site bone defect volume and regeneration on MRI at 4 weeks and 12 months after bone–patellar tendon–bone anterior cruciate ligament reconstruction and to determine their association with knee function and donor-site morbidity at 12 months. Materials and Methods: This single-center prospective observational cohort study included 30 patients who underwent ACL reconstruction with a BTB autograft. Donor-site bone defect volume was estimated on MRI using a triangular prism approximation at 4 weeks and 12 months by two independent evaluators blinded to patient-reported outcome scores. Clinical outcomes were assessed at 12 months using the International Knee Documentation Committee (IKDC) subjective knee form and the Donor Site Morbidity Questionnaire (DSMQ). Associations between MRI-derived parameters and patient-reported outcomes were analyzed using Spearman’s rank correlation coefficient. Results: At 4 weeks, mean donor-site bone defect volume was 2602.4 ± 684.7 mm³ in the patella and 2993.9 ± 714.3 mm³ in the tibia. At 12 months, defect volume decreased to 628.0 ± 279.7 mm³ and 980.8 ± 488.2 mm³, respectively. Tibial defects were significantly larger than patellar defects at both time points, while regeneration was significantly greater in the patella than in the tibia (74.8 ± 11.5% vs. 67.2 ± 15.1%; p = 0.0264). Regeneration was not significantly associated with IKDC or DSMQ scores (all p > 0.05). Larger patellar defect volume at 4 weeks was associated with worse subjective outcomes (both p = 0.0107). Conclusions: After BTB ACL reconstruction, tibial donor-site bone defects were larger, whereas patellar defects showed greater regeneration over time. Larger patellar defect volume at 4 weeks, but not regeneration percentage, was associated with worse subjective outcomes at 12 months. Full article

Engaging students in sustainability challenges is often easier in theory than in practice. This study examines first-year French undergraduates’ patterns of engagement with the TEDS module (Transition Ecologique pour un Développement Soutenable), a nationwide programme developed in France to promote ecological transition and sustainable development. Data were collected through an online questionnaire comprising 24 closed- and open-ended questions exploring students’ self-reported familiarity with, understanding of, concern about, and self-reported intentions to engage in sustainability-related actions, as well as perceived learning needs and background characteristics. Only 18 questions (143 items) were included in the present analysis, covering all dimensions except those related to the evaluation of the training programme. Results indicate that environmental concern is the factor most strongly associated with self-reported engagement intention, despite persistent gaps in conceptual understanding, particularly regarding the Anthropocene and alternative socio-economic models. Knowledge score and concern are structured hierarchically according to issue visibility, with climate change ranking highest. Engagement depends not only on concern but also on perceived opportunities for action, yet students struggle to identify concrete pathways. The absence of significant differences across gender and disciplines points to a strong generational convergence that reshapes the determinants of environmental engagement. Overall, the key challenge for sustainability education is linking systemic knowledge to concrete contexts of learning and everyday life. Full article