Abstract: Accurate wind resource estimation in complex environments, such as the Colombian Caribbean, is challenging due to air–sea heat fluxes and nonlinear atmospheric dynamics. This systematic review and meta-analysis synthesized empirical evidence from 49 studies to evaluate whether advanced forecasting architectures (artificial intelligence, ensemble, and hybrid models) systematically outperform classical physical models. The pooled random-effects estimate indicated an overall reduction in reported forecasting error (\(g = -1.6874\)), though heterogeneity was extreme (\(I^2 = 99.99\%\)). Categorical meta-regression revealed no statistically significant advantage for advanced architectures over physical models (\(\beta_1 = 0.2558\), \(p = 0.9339\)). Mixed-effects meta-regression identified Sample Size (\(N > 10{,}000\)) as the primary significant moderator, with the final model explaining 40.47% of between-study variability (\(R^2_{\text{analog}} = 40.47\%\), reducing residual variance from \(\tau^2 = 2.5102\) to \(\tau^2 = 1.4943\)). Metric-specific contrasts indicated that physical and AI architectures are statistically indistinguishable, whereas hybrid models yielded lower pooled effects. classical physical models remain a reliable, stable reference framework, whereas AI and hybrid approaches demonstrate advantages only in specific, non-generalizable settings. These findings challenge the assumption of universal superiority for AI-based wind forecasting in complex coastal domains.

Abstract: While polynomial interpolation discretizes fractional operators up to order \(k-\alpha\) (\(k=4,5,6,7\)), we construct a \(4-\alpha\) order approximation for the Caputo derivative directly in the frequency domain using the asymptotic expansion of the polylogarithm function. We determine the scheme's weights by applying the method of undetermined coefficients to a polynomial correction of the generating function \(\operatorname{Li}_{1+\alpha}(e^{-z})\), which cancels low-order truncation errors. To address the continuous solution's initial weak singularity near \(t=0\)—a persistent issue that degrades theoretical \(\mathcal{O}(\tau^{4-\alpha})\) accuracy to first order—we introduce a starting correction that modifies the right-hand side of the discrete equations. Although this modification applies solely to the first three time steps, Laplace transform analysis establishes the uniform recovery of the optimal \(\mathcal{O}(\tau^{4-\alpha})\) convergence rate for any fixed \(t>0\). Numerical tests on both homogeneous and inhomogeneous equations confirm this result.

Abstract: Background and Objectives: Consolidation durvalumab after chemoradiotherapy (CRT) is the standard of care for unresectable stage III non-small cell lung cancer (NSCLC). However, a substantial proportion of patients do not receive this treatment in real-world practice.. This study aimed to evaluate the efficacy of post-progression immune checkpoint inhibitor (ICI) therapy in patients with stage III NSCLC who did not receive durvalumab consolidation. Materials and Methods: This multicenter retrospective study included 403 patients with stage III NSCLC who received definitive CRT without durvalumab consolidation between January 2010 and December 2025. Patients were divided into ICI group (n=61) and non-ICI group (n=338) based on post-progression treatment. All patients has ECOG PS 0-1 The primary endpoint was overall survival (OS), analyzed using Kaplan-Meier and multivariate Cox regression methods. Early progression was defined as recurrence within 6 months after CRT completion. Results: The median OS was significantly longer in the ICI group compared to the non-ICI group (47.97 months [95% CI: 35.68–60.26] vs. 19.29 months [95% CI: 17.88–20.69]; p<0.001), representing a 2.5-fold survival advantage. Patients with early progression had a significantly worse prognosis compared to those with late progression (median OS: 7.06 vs. 20.01 months; p<0.001). In the ICI group, PD-L1-positive patients had numerically longer median OS than PD-L1-negative patients (54.05 vs. 46.03 months). Multivariate analysis identified post-progression ICI use (HR: 0.37, 95% CI: 0.25–0.54; p<0.001) and early progression (HR: 2.61, 95% CI: 1.91–3.57; p<0.001) as independent prognostic factors. Conclusions: Post-progression immunotherapy was associated witha significant survival benefit in patients with stage III NSCLC who did not receive durvalumab consolidation, with a median OS approaching 48 months. Early progression within 6 months after CRT is the strongest negative prognostic factor. These findings suggest that immunotherapy should be considered in this frequently encountered patient population, provided performance status is adequate. Prospective validation is needed.

Abstract: Zirconia-based ceramics are widely used in fixed prosthodontic restorations owing to their favorable mechanical properties and evolving aesthetic characteristics. However, predicting long-term performance remains challenging because artificial ageing methodologies are highly heterogeneous and existing standards do not fully reflect the diversity of modern zirconia systems, including high-translucency and multilayer materials; variations in experimental conditions, specimen designs, measurement approaches, and reporting practices limit inter-study comparison and quantitative evidence synthesis. This review examines current artificial ageing methodologies for zirconia-based restorations and identifies a fundamental limitation beyond the absence of a universal ageing protocol: the lack of a standardized framework for preserving and exchanging experimental information. To address this, the Minimum Information About a Zirconia Experiment (MIAZE) reporting framework is proposed as a model-neutral, machine-readable structure for documenting experimental observations and associated metadata. MIAZE is a minimal reporting standard that does not require identical experimental protocols, instrumentation, or analytical approaches, but defines the essential information needed to preserve scientific interpretability and enable future dataset reuse. By separating data documentation from subsequent interpretation, MIAZE provides a foundation for reproducibility, independent verification, quantitative meta-analysis, and the application of diverse analytical approaches, shifting the focus from a single universal ageing protocol toward interoperable experimental datasets capable of supporting evolving zirconia materials, methodologies, and computational tools.

Abstract: Governance operates simultaneously across multiple levels—from local communities to national governments to international institutions—yet most formal governance models treat governance as a single-level phenomenon. This conceptual paper extends the General Governance Success Model (GSM) to account for multi-level governance, providing a computational systems framework for analysing how governance capabilities interact vertically and horizontally across nested scales. The extended model represents governance at levels, with each level characterized by five coupled capabilities: accountability, institutional competence, social cohesion, strategic continuity, and adaptive learning. The framework formalizes vertical (top-down and bottom-up) and horizontal (peer-to-peer) interactions, cross-level information flows with fidelity matrices, level-specific dynamics with differentiated parameters, and recursive learning processes. Drawing on the Viable System Model's recursion principle, polycentric governance theory, multi-level governance theory, and system dynamics, the model generates testable hypotheses concerning vertical alignment, polycentric governance, cross-level coordination, and recursive learning. The paper further demonstrates the model's empirical applicability through a worked example applying the GSM to Denmark using open-source data from the World Bank, United Nations, V-Dem, and World Justice Project, followed by a summary of comparative analysis findings across 100 countries (with full results in Appendix C). The extended GSM provides a foundation for analysing governance systems in their full institutional complexity, offering both theoretical insights and practical guidance for designing more effective multi-level governance arrangements. The framework treats computational methods as decision-support tools rather than substitutes for political judgment, with normative objectives and model weights remaining subject to public deliberation, constitutional rights, transparency, and contestability.

Abstract: Background/Objectives: The term rotator cuff related shoulder pain (RCRSP) refers to pain localized in the deltoid region. RCRSP and the associated pain and functional limitations have a significant impact on the quality of life of those affected. This systematic review aimed to investigate the treatment and follow-up effects of extracorporeal shockwave therapy (ESWT), photobiomodulation (PBM), ultrasound (US), and electrotherapy modalities (ET) in RCRSP. Methods: A comprehensive literature search was conducted. In addition, the reference lists of the included studies and relevant reviews were screened to identify further eligible studies (snowballing). Randomized controlled trials (RCTs) investigating the treatment and follow-up effects of ESWT, PBM, US, and ET for RCRSP, with control groups receiving sham or alternative physiotherapeutic treatments, were included without restrictions on date or language. Two reviewers independently conducted the literature screening, data extraction, and risk-of-bias assessment for the included studies. The differences in pain between pre- and post-intervention, the between-group differences (intervention vs. control), as well as subgroups (low-level laser therapy [LLLT] and high power laser therapy [HPLT]) were analyzed. Results: A total of 20 RCTs with 1152 patients were included. Quality assessment indicated low risk of bias in 55% of the included studies, some concerns in 35%, and high risk in 10%. In the nature of a sustainable therapy, therapeutic modalities were mostly not applied in isolation but rather in combination with conventional physiotherapeutic interventions. One study showed an end-of-treatment pain reduction for ESWT, and one study showed no follow-up effect on pain. PBM showed an end-of-treatment effect and follow-up effect. The subgroup analysis suggested that only HPLT demonstrated a clear end-of-treatment effect. Due to the limited number of studies, no subgroup analysis could be performed for follow-up outcomes. US showed beneficial effects both at the end of treatment and at follow-up, while ET demonstrated an end-of-treatment effect. Further studies are needed for the follow-up effect. Conclusions: HPLT, US, and ET showed consistent additive treatment effects, while PBM and US showed follow-up effects. In general, especially for PBM, ESWT, and ET, more studies are needed to gain a broader insight into this topic.

Abstract: The integration of Multimodal Large Language Models (MLLMs) into healthcare has the potential to drive a considerable advancement toward many AI use cases in medicine, offering transformative capabilities across hierarchical levels of clinical granularity, ranging from microscopic tissue analysis over organ level imaging and individual patient modeling to population-level health surveillance. However, such integration into clinical practice is constrained by a multifaceted “trustworthiness gap”: the mismatch between the raw capabilities demonstrated by Medical MLLMs in controlled experiments and their reliability, fairness, safety, privacy protection, explainability, and regulatory accountability required for clinical deployment in the real world. This gap requires systematic examination of both model behavior and clinical risk across the full lifecycle of Medical MLLM development and use. In this paper, we survey the existing literature on methods, evaluation, and benchmarks of trustworthiness research in Medical MLLMs. We first establish a multi-scale clinical landscape across tissue, organ, individual, and population levels to categorize current models, providing critical insights into the interplay between data heterogeneity and clinical task requirements. Subsequently, we propose a holistic, six-dimensional taxonomy of trustworthiness, comprising truthfulness, robustness, fairness, safety, privacy, and explainability. Using this taxonomy, we critically analyze existing literature to elicit recurring Medical MLLM trustworthiness failure modes and suggest appropriate remediation strategies. Furthermore, our survey enumerates open challenges in evaluating trustworthiness by means of traditional automated metrics, frontier “LLM-as-a-Judge” methods and expert-centric assessment protocols. As a response to these challenges we identify emerging research directions including dynamic and workflow-oriented evaluations for interactive Medical MLLMs. We aim for this work to serve as a systematic guide for researchers and practitioners aiming to develop the next generation of trustworthy medical AI by achieving clinically acceptable reliability through technological innovations. Project Link: https://github.com/junyuanM/Trustworthy-Medical-MLLMs-Survey.

Abstract: We derive stellar motion, projected mass and structure formation from one objective observation field in Physical–Observation Cosmology. Its metric stress contributes real gravity and enters an inferred dark residual if omitted. The common quadratic action yields coupled collisionless and Schrödinger–Poisson evolution. Independent canonical component dilations have a positive energy Hessian, and the computed spherical responses remain bounded over the reported intervals. Across three calibrated lens cylinders, the unexplained fraction falls from 76.03% to a signed residual of \( -1.84\pm5.90 \) percentage points. For DF2, the stellar prediction is \( 7.77\pm0.64\,\mathrm{km\,s^{-1}} \); the additional-source fit is minimized at zero, although its approximate profile bound permits 57% extra mass inside the stellar half-mass radius. We prove the regional zero-residual criterion and its finite-error extension. Transferring the earlier intrinsic mode to DF4 without fitting its velocity gives \( 8.20\pm0.65\,\mathrm{km\,s^{-1}} \), consistent with \( 7.8^{+2.3}_{-2.0}\,\mathrm{km\,s^{-1}} \). The field increment is below \( 10^{-9}\,\mathrm{km\,s^{-1}} \), so this agreement tests the stellar-dominated limit. Expanding linear and nonlinear calculations then generate spatial structure from specified occupation and initial transport. A 4% periodic seed at 10 Mpc develops ordinary and field crest densities of 4.089 and 3.918 times their respective means, with new force harmonics fixed by the evolved source. Applying the environment to C0302 exposes its limits. The formal fixed-tracer monopole response is far below the required velocity change, and the unchanged infinite stellar tail develops negative remote pressure in the 10 Mpc environment. A self-consistent isolated-source refit improves the velocity residuals but requires an equivalent external column fraction of 0.433; the evolved profiles supply at most 0.00201 along a specified 10 Mpc path. The dwarf apertures require no detected independent dark component. C0302 still requires a resolved environmental and stellar-boundary solution before the same field can satisfy all of its constraints.

Abstract: Archaeological remains of Roman amphitheatres in Turkey have always been very scarce compared with those in other parts of the Roman Empire. For a long time, academic literature – and travellers before that – described the remains of only a handful of examples. This stands in stark contrast to the remains of ancient theatres, which are found in considerable numbers. More recent discoveries, however, have brought to light other amphitheatres previously unknown until a few years ago. By analysing the available sources and the latest insights from the international scientific community in the field of conservation, it is possible to accurately assess the current state of knowledge regarding archaeological discoveries of amphitheatres in Turkey, proposing a classification that takes into account uses and transformations even during the late antique period. Furthermore, on this basis, it is possible to identify appropriate strategies for implementing effective conservation and enhancement measures for the sites. It is necessary to identify measures for the archaeological protection in relation to natural forces, human-induced phenomena and connections with the landscape.

Abstract: Short-term forecasting is difficult at distributed photovoltaic (PV) stations that retain hourly energy records but lack power measurements at fine time intervals and site-specific irradiance. We propose a reconstruction-aided method for one-step-ahead forecasting under these conditions. Variational mode decomposition (VMD) separates the target-station energy sequence and the reference-station energy and power sequences into multiscale modes. Hilbert modal features and a multidimensional similarity score identify suitable reference modes, and extremely randomized trees (ExtraTrees) reconstruct the target station's 5-min power profile while preserving each recorded hourly energy total. For the irradiance input, the clear-sky index reduces periodic effects, VMD captures multiscale disturbances at observed grid cells, and ExtraTrees completes global horizontal irradiance (GHI) at the unobserved target cell. Reconstructed power, completed GHI, periodic time features, and aggregated neighboring-station power are then supplied to LightGBM to predict target power at the next time step. Tests use 2021 measurements from distributed PV stations in a UK region and CAMS irradiance on a 7 × 7 grid. The normalized mean absolute error was 0.377%, the normalized root mean square error was 1.168%, and the coefficient of determination was 0.990. The results indicate that the method can recover fine-resolution power and local irradiance from limited target-station information and improve short-term forecasts for data-limited distributed PV stations.

Abstract: Large language models (LLMs) are increasingly expected to act as generalist agents capable of solving complex real-world problems. Training such agents, however, requires stable and diverse environments that support repeated interaction with stateful, tool-augmented tasks and provide verifiable feedback. Despite recent progress, the development of robust LLM agents remains limited by the lack of realistic, scalable, and executable training environments. We present InternAgentHarness, a scalable synthetic environment for improving the agentic capabilities of LLMs. Built upon the InternBootcamp training framework [1], InternAgentHarness instantiates executable agent environments through a four-layer interface that unifies prompt generation, tool execution, interaction control, and reward computation. We further introduce \bootcampcli, an agent harness that automatically converts diverse agentic tasks into a Bootcamp-trainable paradigm. Unlike static benchmarks, InternAgentHarnessmakes evaluation actionable: observed failures can be systematically converted into new synthetic tasks, filtered trajectories, reinforcement learning rollouts, and subsequent re-evaluation under the same executable interface. We instantiate InternAgentHarness on a suite of 10 tasks covering both text-only and vision-based agent scenarios. Starting from Qwen3-VL-30B-A3B-Thinking, both supervised fine-tuning (SFT) and reinforcement learning (RL) on InternAgentHarness substantially improve performance over untuned model. These results suggest that InternAgentHarness provides a practical foundation for scalable synthetic agent environments and enables the continuous improvement of LLM agents through an iterative cycle of evaluation, synthesis, training, and refinement.

Abstract: Wood modification has been developed as a non-biocidal concept to overcome inherent material limitations such as dimensional instability and biological durability. But sustainability assessment must extend beyond initial functional performance to include machinability, interface compatibility, repairability and end-of-life (EoL) cascade potential. This critical narrative review integrates the mechanisms, processing dynamics and circular-economy implications of 4 major modification routes: thermal treatment, acetylation, furfurylation and thermo-hydro-mechanical (THM) densification. We examined a purposive sample of 40 peer-reviewed articles including seminal citations between 1990-2007 and recent publications between 2021-2025. The synthesis shows that thermal modification and acetylation can enhance hygroscopic and di-mensional stability through different mechanisms, but with different trade-offs in mechanical toughness and secondary processing. Furfurylation can improve hardness and decay resistance, but also increase brittleness and recycling difficulties. THM densification improves the local load bearing capacity, but moisture-induced set-recovery is a major limitation. Cross-contamination in heterogeneous waste streams, residual chemical matrices, changed surface-energy interactions with adhesives, and limited standardization of sorting protocols are important barriers to circular material loops across the modification routes. A descriptive particleboard case-study dataset covering 0–70% secondary-wood fractions shows decreasing MOR, MOE and IB and increasing thickness swelling with increasing recycled content. These observations are presented as only study-specific descriptive evidence, and not as evidence of a universal recycling threshold or formal EN 312 P2 conformity, as the current experimental documentation does not contain sufficient information on independent panel replication and inferential statistical analysis. Future work should consider design for disassembly, digital material traceability and multi-criteria sustainability assessment.

Abstract: Hospital at Home (HaH) enables advanced cancer patients to receive end-of-life care in their preferred setting, yet evidence on home-based palliative sedation (PS) remains limited. This retrospective observational study included 113 patients who died in 2022 under the HaH Unit of the Instituto Valenciano de Oncología (Valencia, Spain). Sociodemographic, clinical, and PS-related variables were extracted from medical records. PS was required in 80 patients (70.8%). The median age was 76 years (IQR 19), 52.2% were male, and 77% lived in their own homes. Dyspnoea (43.7%) and delirium (23.7%) were the most frequent refractory symptoms leading to PS. Elastomeric infusers were used in 92.5% of cases. The median duration of PS was 1 day (IQR 2); 46.3% of patients died within 24 hours of initiation. A significant inverse correlation was found between the Barthel Index and PS duration (Spearman’s Rho = −0.226, p = 0.044), indicating shorter survival after sedation in patients with greater functional dependence. No significant associations were observed between PS requirement and multimorbidity, geriatric syndromes, polypharmacy, or prior opioid use. These findings support the feasibility of complex PS at home and highlight functional status as a potential determinant of PS duration, informing protocol optimization for home-based end-of-life care.

Abstract: Bioluminescent fungi are a unique group of light-producing organisms, similar to some bacteria, insects, and marine organisms. About 132 species are currently known, all belonging to the order Agaricales and grouped into five major evolutionary lineages. Their light production is linked to a conserved caffeic acid-based luciferin–luciferase pathway. Besides producing light, these fungi are a potential source of bioactive compounds with antioxidant, antibacterial, antifungal, anticancer, anthelmintic, anti-inflammatory, and immunomodulatory activities. For example, compounds from Neonothopanus gardneri showed activity against Schistosoma mansoni, with an EC₅₀ below 10 µM and no observed toxicity to mammalian cells or Caenorhabditis elegans. The fungal bioluminescence pathway has also been studied for use in biotechnology, including the development of autonomous light-producing biological systems. This has created opportunities for applications in agriculture, environmental monitoring, and bioengineering. Bioluminescent fungi are still being discovered in new regions. For example, Roridomyces cf. phyllostachydis was recently recorded from Namdapha National Park, Arunachal Pradesh, India, where it was found growing on dead bamboo. Overall, bioluminescent fungi have promising applications in medicine, agriculture, environmental monitoring, and biotechnology, although their large-scale cultivation and clinical development remain challenging.

Abstract: Background: AI chatbots are increasingly used to access health and nutrition information, but the quality and alignment of their responses with evidence-based dietary guidance remain uncertain. This study evaluated whether healthy eating recommendations generated by four AI chatbots aligned with the Dietary Guidelines for the Brazilian Population (DGBP) and whether their responses included content reflecting food corporations’ narratives. Methods: Between February and March 2026, ChatGPT, Gemini, DeepSeek, and Grok were evaluated using two questionnaires: a validated instrument developed to assess health professionals’ knowledge of the DGBP (GAB-1); and the second simulated a citizens’ interaction based on DGBP recommendations. Responses were assessed using the GAB-1 scoring system and two author-developed instruments assessing alignment with the DGBP and food corporations’ narratives. Results: All AI chatbots provided information consistent with the DGBP in the GAB-1 assessment. However, in the citizens’ interaction, they produced at least one response containing content misaligned with the DGBP or reflecting food corporations’ narratives. AI chatbots covered less than 50% of the DGBP’s main content, with no significant differences among them (p = 0.219). Conclusions: Although AI chatbots performed well when queried using technical terms, their nutrition advice to citizens included content inconsistent with the DGBP or reflecting food corporations’ narratives.

Abstract: UNESCO Global Geoparks are recognized as instruments for territorial development. However, limited research explains how emerging geoparks can transform distinctive territorial assets into sustainable and inclusive tourism destinations. This study examines the Kütralkura UNESCO Global Geopark in Chile and asks how its tourism potential can be activated by addressing governance gaps, territorial capacity constraints, and destination visibility. Adopting a pragmatic approach, we analyzed an open-ended survey of 32 tourism ecosystem actors across the Geopark. Thematic analysis identified four interrelated dimensions shaping tourism development. First, sustainable territorial competitiveness depends on strategically leveraging geo-natural capital, biocultural heritage, and territorial identity as tourism assets. Second, strategic destination governance requires multilevel coordination and sustained vertical institutional commitment. Third, territorial tourism capacity highlights the need to balance tourism development and visitor use with environmental conservation and local capabilities. Fourth, destination activation and visibility involve negotiating local and international branding while strengthening residents’ geoheritage literacy. The findings demonstrate that tourism potential in emerging geoparks depends not only on resource endowment but also on governance, territorial capacity, and destination activation. The study contributes practical insights for geopark management and proposes landscape management as an integrated approach to sustainable tourism development in emerging economies.

Abstract: This survey presents a comprehensive review of coreset methods in deep learning, an important tool for improving data efficiency in large-scale neural networks. In general, “coreset” is an algorithmic technique for selecting a small yet representative subset of data to replace the full dataset, which can yield more efficient training process and meanwhile preserve model performance. In the past 20 years, coreset techniques have been widely applied to many classical machine learning problems, such as clustering, regression and classification. In recent years, the coreset techniques also begin to attract a lot of attention in modern deep learning area. However, designing effective coresets usually is a challenging task since we need to take account of the trade-off among multiple different factors, such as complexity, robustness and accuracy. In this survey, we focus on two common scenarios for using coreset methods in deep learning: (1) reducing the extremely high computational cost for training a deep learning model, and (2) improving the data utilization under resource constraints such as limited label budget or storage capacity. We begin by outlining the fundamental principles, advantages, and design challenges of coresets for these two scenarios. We also discuss the emerging applications of coresets in large language models. Finally, we identify several open problems and promising directions for future research.

Abstract: This study, presents OptiWood-YOLO, a deep learning method for automated wood-surface defect detection. Manual inspection of wood surfaces takes time and effort, and the work becomes difficult for operators who must check every board on a fast production line. We collected dataset from defective timber samples in collaboration with local wood-processing companies at a running sawmill. We first annotated the collected images using Roboflow to identify and label the different surface defects. After annotation, we applied preprocessing and augmentation techniques to improve data quality and diversity and to address common visual challenges such as motion blur, uneven lighting, and background clutter. We evaluate the method under four ablation settings: without preprocessing or augmentation, with preprocessing only, with augmentation only, and with both preprocessing and augmentation. The dataset contains seven annotated classes in total: six wood-defect classes blistering, crack, fingerjoint, planing torn, scratch, and wrinkling and one auxiliary class, Objects. We compare the final OptiWood YOLOv8n model with YOLOv11, faster region-based convolutional neural network (faster R-CNN), and the transformer-based RT-DETR model. We use precision, recall, mean average precision (mAP) at intersection over union (IoU) threshold 0.5 (mAP@0.5), mAP@0.5:0.95, and inference speed measured in frames per second (FPS). The final OptiWood YOLOv8n model achieves a precision of 0.54, recall of 0.47, mAP@0.5 of 0.46, and mAP@0.5:0.95 of 0.23. It achieves the best overall detection accuracy among the evaluated models while also maintaining the highest inference speed. Our experimental results show that preprocessing, augmentation, and class-imbalance mitigation help to improve detection performance.

Abstract: By averaging the sizes of metal nanoparticles capable of merging at different concentrations of elements in the alloy, the size dependence of ultrafine particles on the melting point of the corresponding bulk elements was determined: d_Me = 2R_Me = 1.46 − 7.3 × 10⁻⁵T₀. It was found that the particle size is the predominant factor in the formation of alloys from nanosized particles at low temperatures, regardless of their physical properties and crystal structure. Based on the Thomson equation and the average sizes of metal nanoparticles, the values of interfacial tension at the crystal–melt interface for like metals were derived and found to satisfactorily follow a linear dependence. An approximate calculation of the interfacial tension for nanoparticles of like elements in the range of (100 − d_Me) nm revealed no dependence on cluster size in this range. The study of size-dependent properties of dissimilar nanoparticles during the formation of solid solutions requires separate investigation. The obtained results can be used to produce new alloys from other elements not considered in this study.

Abstract: Chimeric antigen receptor (CAR) T-cell therapy for solid tumors is limited by inadequate T-cell persistence and tumor infiltration, as well as treatment-associated cytokine release syndrome (CRS). To address these challenges, we developed a metabolically rewired CAR T-cell (MRCAR) platform by co-expressing adenosine deaminase 1 (ADA1) and CD26 in CAR T cells. MRCAR T cells were generated using either GPC3-targeted CARs for hepatocellular carcinoma (HCC) or HER2-targeted CARs for breast cancer. For HCC, we further evaluated constructs with and without enforced IL-15 expression to boost T-cell function. In vitro, IL-15–expressing GPC3-MRCAR T cells exhibited enhanced proliferation, sustained effector function, reduced exhaustion, and superior cytotoxicity following repeated antigen stimulation compared with conventional CAR T cells. CD26 expression significantly enhanced CAR T-cell invasiveness in transwell assays and promoted tumor infiltration in vivo in both tumor models. ADA1 expression enhanced CAR T-cell persistence and was associated with increased frequencies of memory and stem-like T-cell populations in vivo. Notably, although IL15-GPC3-CAR T-cell treatment increased antitumor activity, it also caused high mortality and systemic inflammation in in Huh7 tumor-bearing mice, whereas incorporation of ADA1 and CD26 into MRCAR T cells significantly improved survival and reduced serum levels of IL-6, IFN-γ, and TNF-α. Mechanistically, conventional IL-15–expressing GPC3-CAR T cells induced proinflammatory M1-like macrophage activation, whereas MRCAR T cells markedly suppressed this inflammatory response. Collectively, these results indicate that metabolic rewiring via ADA1 and CD26 enhances CAR T-cell expansion, tumor infiltration, persistence, and safety, supporting the MRCAR platform as a promising strategy to improve both efficacy and tolerability of CAR T-cell therapy for solid tumors.