Browse Articles

Performance and safety of a fine-tuned small language model for pediatric emergency triage: A benchmark study

Pediatric emergency triage is a safety-critical task, and recent studies have explored whether artificial intelligence, including language models, can support triage decision-making; however, evidence on fine-tuned open-weight language models remains limited. We conducted a retrospective benchmark study using de-identified triage records from a tertiary pediatric emergency department in Korea collected from January 2020 to April 2025. After exclusions, 74,170 encounters were included. Each encounter was reconstructed into a case-level text sequence from triage-time structured variables and nurse-authored narratives. Qwen3-8B-Base was fine-tuned with Low-Rank Adaptation and Group Relative Policy Optimization using a safety-oriented reward design and was compared with a structured-data XGBoost model on a common evaluable test subset of 14,832 encounters. The fine-tuned model achieved an accuracy of 58.60%, a macro-F1 score of 0.417, and a quadratic weighted kappa of 0.535. Within-one-level agreement was 97.13%, and strict under-triage, defined as true Korean Triage and Acuity Scale levels 1 or 2 predicted as levels 4 or 5, occurred in 0.65% of cases. The structured-data comparator showed higher overall performance, with an accuracy of 69.40%, a macro-F1 score of 0.618, and a quadratic weighted kappa of 0.651. However, the fine-tuned model showed fewer extreme errors and lower strict under-triage in selected high-acuity groups, at the cost of higher over-triage. In this real-world pediatric benchmark, the fine-tuned language model did not surpass the structured-data comparator in overall performance but showed a distinct safety-oriented error profile. These findings support its potential role as a decision-support aid for human triage review rather than an autonomous triage system. External and prospective validation will be necessary before clinical implementation.

Generation of anti-bovine CD20 monoclonal antibody with cytotoxic activity for the depletion of bovine B-cell lymphoma cells

Modelling cognitive outcomes in the UK Biobank: Education, noradrenaline and frontoparietal networks

Education is often used as a surrogate measure of so called cognitive reserve (CR) benefiting cognitive functioning in later years. In line with Robertson’s theory we tested here a hypothesis that education acting on the noradrenergic system strengthen the right fronto-parietal networks to facilitate CR and maintain cognition throughout the lifetime. We used machine learning and mediation analysis to model interactions between neurobiological features (genetic variants in noradrenergic signalling, structural and functional fronto-parietal connectivity) and education (proxy of CR) on cognitive outcomes (measured here by general cognitive ability score calculated based on performance across a battery of cognitive tests) in the UK Biobank cohort. We show that: (1) interactions between education and neurobiological variables better explain cognitive outcomes than either factor alone; (2) among the neurobiological features selected using variable importance testing, measures of right fronto-parietal connectivity are the strongest mediators of the association between education and cognitive outcomes. Our findings offer novel insights into neurobiological basis of CR by pointing to between-networks connectivity, representing connections linking the default mode network with the right fronto-parietal network as the key facilitator of CR.

The privacy and safety paradox regarding technical compliance and employer enforcement of home office lighting requirements

Financial development and economic growth in Sub-Saharan Africa using system GMM analysis

The study on financial development and economic growth in Sub-Saharan Africa utilises System GMM analysis to investigate the relationship between financial development indicators and regional economic performance. The research findings reveal significant impacts of various financial indicators on economic growth, such as the positive influence of bank liquid reserves on bank assets ratio (R/A), trade openness, and the broad money to total reserves ratio (M/R) on the economic growth of Sub-Saharan Africa. Additionally, the study highlights the negative impact of Credit extended to the private sector by banks within the country (D_bank) on economic development, emphasising the importance of prudent credit allocation to avoid over-indebtedness and financial crises. These results provide valuable insights for policymakers aiming to foster sustainable economic growth in the region by leveraging financial development effectively.

Stress impairs your brain’s ability to link memories — dampening insight

Multi-objective node placement optimization in multiplex 6G wireless networks using quantum-inspired evolutionary algorithms

Abstract The planning of future sixth-generation (6G) wireless networks needs efficient node placement algorithms that can handle heterogeneous network deployment, high connectivity density, and competing performance requirements. Current node placement algorithms mostly use single-layer graph models and traditional optimization methods, which are inadequate to model the interdependent relationship of future 6G networks. In this paper, we present a quantum-inspired evolutionary learning-based multi-objective optimization framework for node placement in multiplex 6G wireless networks. The network is represented as a multiplex graph to capture the interactions between the capacity, latency, and interference layers. The solutions are probabilistically represented using quantum-inspired representations to facilitate global exploration and prevent premature convergence in the high-dimensional search space. A composite fitness expression is designed to address the optimization of network capacity contribution, incentive-aware participation, and multi-layer node centrality simultaneously. Extensive simulations validate that the proposed framework achieves a balanced fitness score of 3.0355 and a $$75.6\%$$ node cooperation rate. Compared to existing methods, the QIEA framework improves overall deployment fitness by $$61\%$$ over random selection and $$12\%$$ over greedy degree based placement. Furthermore, it converges $$23\%$$ faster than standard Genetic Algorithms (GA), demonstrating superior efficiency and scalability for largescale 6G network optimization.

Exploring the genetic architecture underlying dietary fiber content in Colombian Andean blueberry (Vaccinium meridionale Swartz)

Dietary fiber composition is a major determinant of fruit nutritional quality, yet its genetic basis remains poorly characterized in wild Vaccinium species. Here, we combined extensive phenotyping with a genome-wide association study (GWAS) to dissect the genetic control of dietary fiber traits in Colombian agraz ( Vaccinium meridionale Swartz). Total dietary fiber (TDF), insoluble dietary fiber (IDF), and soluble dietary fiber (SDF), the SDF/IDF ratio, and maturity index (MI) were quantified in fruits from 119 genotypes, representing the most comprehensive evaluation of dietary fiber fractions in fresh Vaccinium fruit to date. Using a GWAS approach based on 16,709 high-quality SNPs across 69 genotypes, we identified 24 QTLs associated with dietary fiber content across 15 chromosomes, revealing a polygenic architecture underlying this trait. A TDF QTL (Chr41:26883013) directly co-localized with VaccDscaff31-augustus-gene-268.33 , a 7-deoxyloganetin glucosyltransferase, embedded within a glycosyltransferase-rich LD block. IDF variation was associated with VaccDscaff33-processed-gene-116.2 (pectin methylesterase 15) while the SDF/IDF ratio co-localized with VaccDscaff55-augustus-gene-9.30 , encoding a xyloglucan endotransglucosylase/hydrolase. Together, the integration of high-resolution phenotyping with QTL mapping connects natural variation in dietary fiber content and composition to specific biosynthetic, remodeling, and regulatory pathways, providing actionable molecular targets for marker-assisted and genomic selection aimed at improving nutritional quality, texture, and processing traits in Vaccinium breeding programs.

AI cracks 80-year-old mathematics challenge — researchers are astonished

Hybrid approximate multiplier architectures for low power DWT image compression

Correction: Effective targeting of breast cancer cells (MCF7) via novel biogenic synthesis of gold nanoparticles using cancer-derived metabolites

Eradication of Mycoplasma contamination in HeLa cells using neomycin resistance gene introduction and aminoglycoside G418 (Geneticin) treatment

Assessing National Health Research System in a resource-limited setting: Insights from Indonesia

High-quality health research is essential for evidence-based policymaking and health system strengthening. Indonesia has undergone major reforms in its health research governance. However, the country’s National Health Research System (NHRS) remains insufficiently mapped. This study aimed to assess the current status of Indonesia’s NHRS using the WHO framework, which includes four domains: governance, financing, creating and sustaining resources, and producing and using research. A qualitative approach was employed, combining an expert panel discussion and key informant interviews with 13 participants consisting of national and subnational stakeholders from government agencies, research organisations, and universities from May to December 2024. Data were analysed using the Framework Method guided by the NHRS framework and supplemented by the WHO Questionnaire on Country Resources for Health Research. This study found that, despite progress in Indonesia’s NHRS, weak governance, particularly poor coordination between national and subnational levels, remains the main challenge. Overall, the system is still fragmented, even though formal structures and institutions are in place. Governance is characterised by a highly centralised, top-down approach to agenda setting, with limited engagement of local institutions. Research financing is mainly programmatic and proposal-based, aligned with national priorities, but is constrained by limited sustainable domestic funding and unequal access to funding. In terms of resources, advanced research infrastructure and growing international collaboration indicate national commitment to health research; however, mismatches between infrastructure and human resource capacity persist. Finally, although research production and utilisation focus on measurable outputs and are supported by mechanisms for evidence use, effective pathways for translating research into policy and practice remain limited. Improving Indonesia’s health research system calls for an integrated national research agenda, well-defined institutional roles, and streamlined coordination between national and regional levels.

Adaptive classification and grading model of cigar wrapper leaf based on improved ResNet algorithm

How does self-control influences college students’ smartphone dependence: The chain mediating effect of mental toughness and social adaptability

The study aimed to investigate the mediating roles of mental toughness and social adaptability in the relationship between self-control and smartphone dependence among college students. In January 2025, through a combination of convenience sampling and quota sampling methods, a total of 600 college students were selected from Hubei Province for a questionnaire survey. Mediation analysis was conducted with the SPSS PROCESS Model 6 to explore how mental toughness and social adaptability influence the connection between self-control and smartphone dependence. Results indicated that self-control, mental toughness, and social adaptability are all significantly negatively correlated with smartphone dependence. Self-control, mental toughness, and social adaptability show a positive correlation between each pair. Self-control can be associated with smartphone dependence through the path of social adaptability. mental toughness and social adaptability had a chain mediating relationship between self-control and smartphone dependence. The research results suggest that self-control ability is a key related factor for college students’ smartphone dependence. It is necessary for colleges and universities to enhance students’ beliefs and abilities to overcome and solve difficulties, enrich their extracurricular lives, help them better integrate into real society, and use smartphones reasonably.

The impact of career calling on medical students’ academic burnout: a chain mediation model

Phenological shifts in mating and lambing timing in response to climate change in Urial wild sheep (Ovis vignei) populations in Iran

Climate change is increasingly disrupting ecological processes across arid and mountainous biomes, with profound implications for the reproductive phenology of large herbivores. These species are especially climate-sensitive, as their breeding cycles are tightly coupled with vegetation dynamics driven by seasonal temperature and precipitation. Yet, in biodiversity-rich regions such as eastern Iran, where climate variability is acute and data are sparse, long-term phenological responses remain poorly understood. Here, we examine how reproductive timing in urial sheep ( Ovis vignei ), a mountain herbivore, responds to climatic variation across six protected areas, as climate-driven mismatches between birth timing and peak forage availability may reduce neonate survival and ultimately affect population viability and connectivity. Climate data (temperature, precipitation, snowfall, and humidity) from the nearest weather station to each study area, along with latitude and mean elevation of each habitat, were integrated using generalized linear mixed models (GLMMs) to assess phenological responses to environmental variables. Our results reveal clear regional differences in mating and lambing time. Mating time was significantly influenced by latitude, summer temperature, and autumn precipitation, with higher latitudes and autumn rainfall delaying mating, while warmer summers advanced it. In contrast, lambing timing was largely dictated by study area-level random effects, which accounted for the majority of variance, whereas fixed effects such as January temperature, snowfall, and latitude contributed only minimally, highlighting the dominant role of spatial differences among study areas in shaping lambing phenology. These findings, over the past decade, underscore the role of climate and latitude in shaping reproductive timing and highlight the urgent need to incorporate phenological data into adaptive wildlife management and habitat-specific climate resilience planning in vulnerable arid mountain ecosystems.

Mapping convergent regulators of melanoma drug resistance by PerturbFate

Opportunistic fungi in tropical beaches: Diversity, WHO fungal priority pathogens, and antifungal resistance case study

Abstract Tropical coastal regions, despite being major tourist destinations, remain underexplored in terms of their role as potential reservoirs of opportunistic fungal pathogens, particularly those listed on the WHO Fungal Priority Pathogens List (FPPL). This case study presents the first integrative molecular and culture-based assessment of fungal communities in beach sand from three sites along the coastline of Prachuap Khiri Khan, Thailand. Using ITS metabarcoding and culture-based methods, we characterized fungal diversity, detected taxa from genera that include FPPL-listed species, and assessed antifungal susceptibility. The results indicated differences in beta diversity among sites; environmental variables showed directional influence on community variation, with temperature significantly associated with the observed patterns, while moisture was not statistically significant. The urban adjacent SON site exhibited a higher prevalence of FUNGuild predicted opportunistic pathogens, while the more remote DSN site showed fewer. Five of the thirteen isolated fungi exhibited elevated minimum inhibitory concentrations (MICs) against azoles and flucytosine, isolates affiliated with Aspergillus section Terrei , which also showed genetic similarity to clinical strains. These findings suggest that recreational beaches may act as potential environmental reservoirs of opportunistic and antifungal-resistant fungi. However, only Fusarium neocosmosporiellum TPS1-10, a high-priority FPPL pathogen, was successfully isolated among taxa detected by ITS metabarcoding, and it did not exhibit elevated MIC values. Overall, these results highlight the need for routine fungal monitoring in recreational beach environments, in accordance with WHO guidelines, particularly in the context of increasing antifungal use, climate change, and growing beach tourism.

Study on plugging law and plugging removal effect of pre filled screen in natural gas hydrate argillaceous silt reservoir

To address the issue of production decline caused by natural gas hydrate blockage in offshore screen pipes, optimization experiments for high-efficiency rotary water jet backflushing technology were conducted. The outer pressure and permeability of the screen pipes were used as criteria to assess backflushing effectiveness. The study examined the effects of various factors, including the type of backflush tool, tool movement speed, screen pipe type, and the presence of perforated pipes. The results revealed that, compared to high-pressure and low-pressure rotary spray guns, the high-pressure cavitation jet rotary flushing tool provided the best backflush performance. Backflushing effectiveness was higher at a tool movement speed of 1 m/min than at 2 m/min. Additionally, 30/40 mesh screen pipes demonstrated significantly better backflush performance than 40/70 mesh pipes. Screen pipes with perforated pipes showed lower backflush effectiveness compared to those without. These findings offer valuable insights for optimizing the backflushing process of offshore natural gas hydrate screen pipes.