Abstract: Artificial intelligence (AI) is transforming organizational processes and workforce capabilities across multiple sectors, generating important implications for sustainable organizational performance. In educational institutions—an underexplored organizational context—administrative staff represent a critical workforce segment whose competencies, adaptability, productivity, and decision-making capacity directly shape institutional sustainability. Yet empirical evidence on how AI adoption affects these outcomes in emerging economy educational settings remains limited. Addressing this gap, the present study examines the predictive relationships between AI adoption and four organizational sustainability indicators: job competencies (CL), resistance to change (RC), administrative productivity (PA), and decision-making autonomy (ATD) among administrative personnel in educational institutions in Chimbote, Peru. A quantitative, cross-sectional, non-experimental design was employed, using Partial Least Squares Structural Equation Modeling (PLS-SEM) with SmartPLS 4.0. Data were collected from 98 administrative staff members across 54 educational institutions. The measurement model confirmed adequate reliability, convergent validity, and discriminant validity across three constructs; however, the Resistance to Change construct exhibited insufficient internal consistency reliability (Cronbach’s alpha below .70) and weak indicator loadings, failing to meet recommended PLS-SEM thresholds [77,81] and precluding its inclusion in the structural model. The structural results indicate that AI adoption exerts a positive and statistically significant predictive association with job competencies (β = 0.627, t = 11.55, p < 0.001), administrative productivity (β = 0.589, t = 9.885, p < 0.001), and decision-making autonomy (β = 0.398, t = 5.267, p < 0.001). The three empirically testable hypotheses (H1, H2, H3) are supported; H4 (Resistance to Change) could not be tested due to measurement reliability constraints. These findings position AI as a substantive driver of sustainable organizational performance in resource-constrained educational contexts, offering empirical evidence from a Latin American emerging economy perspective in alignment with Sustainable Development Goals 4, 8, and 9.

Abstract: To analyse impact of levee axis adjustment on flow variation in the Xinsha Island which is located in the middle segment of the Fuchun river waterway in Fuyang, Hangzhou, a two-dimensional river flow model was constructed. In the model steady flow with different return periods and unsteady flow in 20-year period were simulated. Consistent outcomes were obtained under steady and unsteady flow. Results indicated that after the levee axis is adjusted, the longer the return periods, the higher the water level in the southern waterway, with a maximum increase of 0.183 m. Conversely, the northern waterway exhibits a more pronounced water level decrease, with a maxi-mum reduction of 0.128 m. The flow velocity of the southern waterway slows down, and the flow velocity of the northern waterway increases. After the levee axis is ad-justed, the flow diversion capacity of the north waterway is effectively enhanced, thereby benefiting flood regulation. These findings provide a sound theoretical basis and well-founded recommendations for adjusting levee axis position and enhancing flood resilience in the Xinsha Island area of the Fuchun River.

Abstract: Background: In Canada, racialized communities, including African, Caribbean, and Black (ACB) people, are disproportionately affected by HIV and COVID-19. Experiencing multiple forms of discrimination in healthcare settings compromises care engagement and health outcomes. The objective of this study was to assess the various forms of discrimination experienced by ACB people during the COVID-19 pandemic, changes in the levels of discrimination experienced before and during the pandemic and the demographic factors associated with increased experience of discrimination among ACB people when accessing healthcare services during the pandemic. Methods: Data were collected via an online survey co-led by the Public Health Agency of Canada, University of Ottawa, ACB community leaders and researchers across Canada. Participants were recruited via email contact. Eligibility criteria included living in Canada at the time of the survey, aged 18 years or older, ability to read English or French, and self-identifying as African, Caribbean or Black. The survey queried access to health services, and experiences of multiple forms of discrimination when accessing healthcare services before and during the COVID-19 pandemic. Multivariable logistic regression was used to identify factors associated with discrimination. Results: Of 1,556 participants, 39.6% were aged 25-39, 42.7% were resident in Ontario, and 63.2% were of African origin. Prior to the COVID-19 pandemic, 75% experienced at least one form of discrimination in a healthcare setting. During the COVID-19 pandemic, over 66% experienced at least a form of discrimination, with 25% reporting a perceived increase in the frequency with which they experienced discrimination. The perceived increase in the frequency of discrimination was 10.8%, 15.3%, 15.9%, 17.0%, 18.1%, 18.7%, and 31.2% among participants who experienced sexual orientation, gender, substance use, disability, age-based, economic status, and race-based discrimination, respectively. In the multivariate logistic regression, the odds of experiencing discrimination in participants aged 50 and above was 0.38 times (95%CI:0.21, 0.69) that in participants who were 31-40 years of age. Conclusion: The proportion of participants with perceived increased experience of discrimination when accessing healthcare services during the COVID-19 pandemic was high. Although there is variation in levels of experienced discrimination, the different forms of discrimination participants experienced (race, gender, sexual orientation, substance use, economic status, disability and age-based discrimination) are alarming. This underscores the need for concerted efforts to address multiple forms of discrimination in healthcare settings to improve care engagement and health equity among ACB communities. There was a significant association between perceived increased experience of discrimination and only one sociodemographic factor - older age (50 and above), other factors that could possibly contribute to participants perceived increased experience of discrimination when accessing healthcare services needs to be explored.

Abstract: We introduce a theory in which gravity emerges as a thermodynamic phenomenon governed by a scalar field that sets the local rate of quantum evolution. Building on Jacobson’s thermodynamic derivation of Einstein’s equations and Verlinde’s entropic gravity, this framework extends these ideas into a unified theory of spacetime thermodynamics. In the strong-field limit it reproduces General Relativity, while in weak-field and low-density environments it predicts modified gravitational dynamics that account for galaxy rotation curves and galaxy cluster mass discrepancies without invoking particle dark matter. On cosmological scales, the theory predicts an early epoch of emergent inflation without an inflaton field and a late-time evolving accelerating expansion driven by the gradual depletion of vacuum thermodynamic capacity, implying cyclic cosmic evolution. From first principles, the framework yields parameter-free predictions for the Hubble constant and the present matter density consistent with observations. We confront the theory with Pantheon+, Cosmic Chronometers, DESI DR2 BAO, and the CMB angular scale \( \theta_\ast \), and find that it provides a statistically preferred description of the data relative to ΛCDM, with ΔBIC = -18.5, and suggests a resolution to the Hubble Tension as an artifact of thermodynamic evolution. These results indicate that a thermodynamic origin of gravity and spacetime offers a coherent explanation of gravitational and cosmological phenomena.

Abstract: Rationale: Reproductive tract infections (RTIs), including pelvic inflammatory disease (PID), are significant public health concerns among young women, especially in sub-Saharan Africa. The university environment, characterized by communal living and varying access to personal healthcare, provides a unique setting to investigate these infections. Understanding the prevalence, microbial patterns, and antimicrobial resistance in university settings is critical to developing effective health interventions. Objectives: This study aimed to assess the prevalence of RTIs, identify the microbial pathogens responsible for pelvic inflammatory disease, and determine their antimicrobial susceptibility patterns among female students residing in hostels at Niger Delta University, Bayelsa State, Nigeria. The study also sought to explore the age distribution of affected individuals and the microbial burden in the university hostel environment. Methods: A descriptive cross-sectional study was conducted at Niger Delta University in Amassoma, Nigeria. Fifty female students within the reproductive age group residing in the university’s hostels participated. Data were collected using high vaginal swabs and midstream urine samples, which were cultured for microbial growth. Antimicrobial susceptibility testing was performed using the Kirby–Bauer disk diffusion method. Descriptive statistical analysis was employed to present the findings. Results: The study found that 52% of participants were in the 18–21 age group, while 48% were in the 22–25 age group. Candida species were the most commonly isolated pathogens (70%), followed by Escherichia coli (30%). The growth rates on Sabouraud dextrose agar revealed a predominance of fungal infections. Antimicrobial susceptibility testing showed varying levels of resistance, with Ciprofloxacin and Levofloxacin exhibiting the highest susceptibility, while higher resistance rates were observed for commonly used antibiotics such as Amoxicillin and Augmentin. Conclusion: The findings suggest that fungal infections, particularly those caused by Candida species, are a significant concern among young female university students. The presence of antimicrobial resistance highlights the need for alternative treatment strategies and enhanced infection control measures. Recommendations: Implement hygiene education and improved sanitation in hostel facilities, introduce routine screening for RTIs and provide access to effective antimicrobial treatments and integrate reproductive health education and regular medical check-ups into the university’s healthcare services. Health Significance: This study underscores the importance of addressing RTIs among young women in university settings to prevent long-term reproductive health issues. The findings contribute to the understanding of microbial resistance patterns, which is essential for the development of effective public health policies and interventions targeting PID and associated complications such as infertility.

Abstract: Spatially explicit knowledge of soil nutrient heterogeneity in arid irrigated agroecosystems remains limited, constraining precision fertilization. In Wuwei City (Hexi Corridor, northwestern China), nutrient management has largely relied on coarse regional averages, while validated geostatistical characterization of cultivated soils is lacking. This study aimed to quantify the variability, interrelationships, and spatial dependence of four key plough-layer nutrients, soil organic matter (SOM), total nitrogen (TN), available phosphorus (AP), and available potassium (AK), across 638 cultivated-land sites sampled in 2022. Descriptive statistics, Pearson correlation, semivariogram modelling, and ordinary kriging with independent validation were conducted to characterize nutrient structure and predictive performance. All nutrients exhibited substantial variability (CV 44.8-97.1%), with AK showing the highest dispersion and weakest spatial continuity. SOM and TN were strongly correlated (r = 0.954), indicating near-collinearity and shared regulation of organic matter. Nugget-to-sill ratios (0.559-0.734) indicated predominantly moderate spatial dependence, while AP exhibited a correlation range of approximately 90 km, reflecting regional-scale gradients superimposed on local management effects. These results demonstrate nutrient-specific spatial structures within the same agroecosystem and underscore the limitations of uniform fertilization practices. Spatially differentiated nutrient management, particularly for K, is recommended, and integration of environmental covariates is needed to enhance predictive precision.

Abstract: Sustainable housing finance has emerged as a critical tool for achieving inclusive, resilient, and environmentally responsible urbanization in developing economies, yet access to affordable, climate-resilient housing remains limited. Rapid urbanization, weak institutional frameworks, high borrowing costs, and underdeveloped mortgage markets exacerbate housing deficits, particularly for low-income populations. Recent developments in financial deepening, including the expansion of banking services, fintech innovations, and microfinance programs, provide new opportunities to address these challenges, but integration with sustainability objectives remains uneven. This review synthesizes existing literature and practical experiences to examine innovative mechanisms that can enhance sustainable housing finance, including green mortgages, ESG-linked lending, climate risk-adjusted finance, blended financial instruments, and digital financial technologies. The analysis identifies persistent gaps in the alignment of affordability, environmental sustainability, and financial viability, highlighting the need for context-specific solutions that mobilize both domestic and international capital. Policy frameworks that incentivize sustainable practices, capacity building for financial institutions and developers, and the adoption of data-driven and technology-enabled solutions are emphasized as essential for scaling impact. The review argues that sustainable housing finance should be understood as a strategic nexus of finance, social equity, and environmental resilience capable of accelerating progress toward SDG 11 while stimulating local economic growth.

Abstract: Staten Island is less developed than the other boroughs of New York city, however outcrops of rock and surface sediment are limited, making interpretation of its geologic history challenging. When small areas of sediment are exposed, they can be used to improve our understanding of changes in sediment erosion and deposition over time. In this study of two small temporary outcrops, the beds of sediment were logged in the field and samples were collected for textural and compositional analyses. The results were interpreted in the context of previous work on similar exposures nearby. The sediments were found to be sands and gravels of fluvioglacial origin, containing reworked sediments of both the Pliocene Pensauken Formation, and older Triassic rocks of the Newark Basin. It is likely that they were deposited on an outwash plain during the Illinoian glaciation. They were deposited in a topographic low, directly overlying Cretaceous sedimentary rocks, but adjacent to sediments of the Pensauken Formation which had in turn been deposited as an earlier valley fill. This interpretation solves an apparent disagreement between previous studies, by illustrating how both the Pensauken Formation and later fluvioglacial sediments can be exposed over a small area.

Abstract: Environmental contaminants pose threats to various organisms and negatively impact the nervous, cardiovascular, immune, and reproductive systems. Per- and polyfluoroalkyl substances (PFAS) are synthetic chemicals that are ubiquitous in the environment. Given that mixtures of environmental contaminants have the potential to exacerbate toxicity, we reviewed current literature on pesticides, microplastics, or metal exposure in combination with PFAS on vertebrates and invertebrates. The objectives were to evaluate the toxicological effects of mixtures of different pollutants (microplastics, pesticides and metal ions) with PFAS on aquatic organisms to better understand biological responses in animals. Based on our review, an increase in toxicity is observed in mixtures of pollutants, including enhanced oxidative stress, developmental abnormalities, impaired reproduction, metabolic disruption, altered gene expression, and changes in enzymatic activity; however, some antagonistic interactions were also reported, underscoring the complexity of mixture effects in real environments. A computational assessment demonstrates that PFOS can engage in intermolecular interactions with pesticides, microplastic monomers, and metals, suggesting chemical-level mechanisms that could modify toxicity or bioavailability. Future studies should focus on elucidating the mechanisms underlying these complex interactions, investigating effects at different trophic levels and in a broader range of species, including mammalian models, and considering chronic exposures and environmentally relevant mixtures.

Abstract: Background/Objectives: Fully robotic pancreatic resections using the Hugo™ RAS platform haven’t yet been described in the literature. Methods: A 72-year-old male with a cystic lesion in the pancreatic tail underwent a fully robotic distal pancreatectomy and splenectomy using the Hugo RAS and the newly introduced robotic vessel sealer LigaSure RAS. The proposed configurational setup and technical details are described. Results: The procedure was completed safely without complications, blood loss was &lt;50 ml, total duration of surgery was 305 minutes, console time 195 minutes. The postoperative period was uneventful, and the patient discharged on postoperative day 7. Conclusions: Distal pancreatectomy with the Hugo RAS may be feasible and safe in selected cases.

Abstract: The strategy developed to carry out a scaled test program aimed at reproducing the behavior of skin heat exchangers to alleviate the heat dissipation requirements in future hybrid electric propulsion regional aircraft is presented. The test program is intended to reproduce, as best possible, the conditions faced by the skin heat exchanger on a predefined nominal cruise flight operation, while conducting the tests in a wind tunnel operating at low velocities and near standard atmospheric conditions. For that purpose, dimensional analysis is used to establish the best geometrical scale and approach flow conditions in the wind tunnel test program. The validation of the strategy is achieved by comparing dimensionless parameters characterizing the turbulent heat transfer process taking place at the skin heat exchanger/airflow interface surface in the flight and wind tunnel environments, by using CFD analysis based on RANS turbulence modeling. The comparison reveals that the adopted wind tunnel strategy is indeed capable of reproducing the heat transfer process taking place in the flight environment, thus paving the way to achieve mid TLR validation of the skin heat exchanger technology.

Abstract: Spinocerebellar Ataxia type 2 (SCA2) and Amyotrophic Lateral Sclerosis type 13 (ALS13) are triggered by polyglutamine expansion in Ataxin-2 (ATXN2). To understand these neurodegenerative disorders at the molecular level, the brains of 10-month-old Atxn2-CAG100-knockin mice were analyzed as microglial, astroglial and neuronal fractions via global RNA sequencing. Data were validated by comparison with spinal cord oligonucleotide microarray profile. Here, glial fractions showed upregulation of Gpnmb (to 2082%), Cst7, Clec7a, Axl, Csf1, Lgals3, Lgals3bp, Slc11a1, and Usp18 as an unspecific neuroinflammatory signature, versus downregulation of axonal Nefh (to &lt;19%), and synaptic Scn4b, Camk2b, Rab15, and Grin1 mRNAs correlating with circuit disconnection. In all fractions, reductions of Kif5a, Rph3a, and Cplx1 were noted, versus disease-specific inductions of ribosomal subunits, presumably mirroring the partial loss-of-function of ATXN2 as RNA translation modulator. Selective accumulations of embryonic factors Rnu1b2 and Eef1a1 versus downregulation of adult Eef1a2, specify mutation impact on splicing and translation elongation. As a potential underpinning of toxic gain-of-functions, the proteostasis transcript Rnf213 appeared increased in astroglial and microglial fractions. These transcriptome data suggest altered ribosomal and spliceosome machinery, with massive microgliosis versus mild astrogliosis, at the core of SCA2 and ALS13.

Abstract: Common bunt of wheat (Tilletia spp.) remains a significant threat to wheat production in low-input and organic farming systems, where chemical seed treatments are restricted or avoided. Host resistance represents a key component of sustainable disease control, but it’s effective deployment requires detailed knowledge of race-specific virulence and the genetic basis of resistance. In this study, we analysed the reaction of a large and diverse wheat germplasm collection to current European populations of common bunt and mapped the underlying resistance genes using SNP-based approaches. A total of 2,731 wheat accessions were phenotyped from 2012 to 2025 using up to 42 purified bunt races with well-defined virulence profiles. Based on phenotypic responses to race-specific resistance patterns , accessions were grouped, and compared with established differential lines. A total of 1504 selected accessions were genotyped using Illumina 26k SNP arrays, and resistance loci were identified by genome-wide association studies followed by fine mapping using recombination analysis. All classical Bt resistance genes from Bt1 to Bt10 and Bt13 were mapped to defined physical intervals, and the genomic positions of 16 additional race-specific resistance genes were identified in the panel of germplasm. Our results confirm that several historically defined Bt genes including Bt11 and Bt12 represent multi-gene resistance complexes rather than single loci. Also, genes established as separate genes may possibly be identical, including Bt4 being identical to Bt6, Bt10 being identical BtZ and Bt9 possibly being identical to one of the genes in the Bt11 complex. These finding highlights the need for revised nomenclature of genes and differetial set of varieties. The identified resistance haplotypes provide an improved tool for marker-assisted selection, and support the development of wheat cultivars with durable resistance to common bunt.

Abstract: Background/Objectives: Delirium is an acute state of confusion associated with impaired consciousness and a decline in cognitive function. Delirium has significant clinical importance due to its substantial impact on morbidity and mortality. The primary objective of this systematic review is to assess Dexmedetomidine’s potential efficacy for delirium in critically ill and elderly patients, addressing a significant need in this high-risk population, and to evaluate its safety as a secondary objective. Methods: A systematic review (2018–2024) searched PubMed, Embase, and Cochrane for English-language studies on dexmedetomidine and delirium in older intensive care unit (ICU) patients. Dual reviewers independently screened, extracted data, and resolved disagreements. Eligible randamized control trials (RCTs) and observational studies were assessed using Risk of Bias 2 ( RoB-2) and Risk Of Bias In Non-randomized Studies - of Interventions (ROBINS-I) tools. Results: Dexmedetomidine emerges as a promising drug because of its unique pharmacological profile, which provides sedation without modulating gamma-aminobutyric acid (GABA) receptors, potentially lowering the risk of delirium. Its additional analgesic, anti-inflammatory, and organ-protective characteristics could provide broader clinical benefits in the ICU. However, the aged population’s heightened susceptibility to Dexmedetomidine’s hemodynamic effects may limit some of its potential benefits. Conclusions: Although existing research suggests short-term neuroprotective effects, these effects do not always translate into better long-term survival. As a result, further large-scale, well-designed randomized controlled trials are needed to determine the optimal use of Dexmedetomidine in this population and to understand its overall impact on morbidity and mortality in the ICU.

Abstract: Objective: This study aimed to estimate the trends, projections, and determinants of standalone and Coexisting Forms of Malnutrition (CFM) at global, regional, national, and individual level among children under five in low- and middle-income countries (LMICs). The study also assessed the projection trajectory towards the 2030 GNTs (GNT) for child growth. Methods: Data from 48 LMICs were analysed using the Multiple Indicator Cluster Surveys (MICS) and Demographic & Health Surveys (DHS). Children with complete anthropometry were included for national and individual-level descriptive analyses. Projected prevalence of each form of malnutrition, including CFM, was calculated using the Annual Rate of Change (ARR). Inferential analyses employed generalized linear regression models (GzLM) with two-way interaction terms to identify determinants of each malnutrition type. Findings: By 2030, 22 of 48 LMICs are projected to achieve all three GNT, up from 10 countries currently, while Yemen and Zimbabwe are expected to remain off-track. Stunting is the most prevalent form, affecting 42 countries, with nine nations projected to have over 50% of children affected by any form of malnutrition. Wasting, obesity, and CFM are rising in several countries. Maternal education and household wealth were the strongest determinants, with children of uneducated mothers and from poorest households at highest risk. Inequalities are narrowing slowly, by 1–2% per year, and marked regional disparities persist. Conclusion: Many LMICs are off track to meet child growth targets when CFM in considered alongside standalone indicators. The government and global health partners must strengthen nutrition surveillance systems and equity-focused policies and programs to routinely capture CFM and prevent as well as manage all forms of malnutrition at national and individual levels.

Abstract: Reinforcement learning is increasingly deployed in domains where reward feedback is sparse, delayed, and entangled with long-horizon constraints, making reliable credit assignment difficult. A central development in recent work is the insertion of large language model modules directly into the reinforcement learning loop, not as peripheral interfaces but as components that alter trajectory generation and supervision. In these systems, language modules provide planning priors, structured reward shaping, process verification, synthetic world traces, and tool-memory context that reconfigure optimization at trajectory level. This survey develops a mechanism-first synthesis of that shift. We formalize intervention operators for planning, reward and verifier channels, world construction, and tool-memory mediation; analyze how each operator changes update targets, bias pathways, and stability conditions; and organize the field into a unified taxonomy grounded in optimization effects rather than model branding. We then examine evaluation practice across embodied control, web interaction, games, continuous control, and multi-agent settings, highlighting reproducibility gaps and protocol confounds. Finally, we synthesize recurring failure modes and propose a concrete research agenda on calibration, module authority arbitration, uncertainty-aware simulation, and benchmark design. The resulting perspective positions LLM-in-the-loop reinforcement learning as a systems and optimization discipline centered on trustworthy credit assignment under heterogeneous supervision.

Abstract: The rapid proliferation of agentic AI, autonomous software systems capable of executing transactions, accessing sensitive data, and acting on behalf of human users, has created an unprecedented security challenge. The existing authentication systems which developers created to authenticate human users and fixed system accounts, face their most significant authentication challenge because they need to establish the identity and access rights and operational purpose of AI agents. Deepfake technology has developed to the point where it can generate synthetic identities that perfectly mimic actual human beings. The first complete framework for AI agent authentication in environments with widespread deepfake usage appears for the first time in this research paper. We propose a verification model that uses multiple security layers to establish machine identity through cryptography while holding users accountable through human identification and measuring user behavior against expected patterns with risk assessment based on transaction details. Drawing on emerging industry concepts including "Know Your Agent" frameworks (Rasmussen, 2026; Sumsub, 2026), agentic AI orchestration platforms (Veritas AI, 2025), and multi-modal deepfake detection research (Bank Rakyat Indonesia & Telkom University, 2025; Kubam, 2024), we present a unified architecture for establishing trust in autonomous digital entities. The framework we developed establishes a complete system which enables people to establish trust in autonomous digital entities. Our framework addresses the fundamental question of our era: when an AI agent appears at the digital gate requesting access, how do we know it is who it claims to be, acting for a legitimate purpose, and not a deepfake in disguise?

Abstract: This study investigates a multidimensional "choir–maqām–meaning" model for Qur’anic memorization (hifz) integrated with formal undergraduate education, ana-lyzed through the lens of Qira’at science and cognitive pedagogy. Departing from tra-ditional individualistic methods, this research evaluates the effectiveness of a collec-tive, melodic approach in sustaining student commitment. Utilizing a mixed-methods design at Kırşehir Ahi Evran University, data from a cohort of 20 students were ana-lyzed through open-ended questionnaires, thematic analysis, and descriptive statistics. Findings indicate that despite the high cognitive and physical demands of dual curric-ula, the integration of choir and maqām enhances long-term retention and minimizes phonetic errors while maintaining peak motivation through peer support. Crucially, this research serves as a pilot phase for an expansive interdisciplinary project. By es-tablishing a theoretical and practical foundation, it aims to pave the way for subse-quent neuroimaging stages utilizing fMRI, DTI, and EEG methodologies to investigate the impacts of this model on neuroplasticity and cognitive reserve. Thus, the study of-fers a novel perspective on how specialized religious training can contribute to brain-based learning and cognitive development within the higher education ecosys-tem.

Abstract: In this work, we present a generic classification of a class of n-dimensional piecewise smooth vector fields known as refractive systems. Our purpose is to characterize the local structural stability of refractive fold-folds of elliptic type. To this end, we reduce the study of such piecewise smooth vector fields to the analysis of the first return maps and their structural stability. Normal forms for such systems are also provided.

Abstract: Large Language Models face significant challenges in biomedical multi-hop reasoning, including noise interference, context sensitivity, and ensuring factual consistency. To address these limitations, we propose the Robust Biomedical Reasoning Agent (RBRA), a novel agent framework designed to significantly improve the robustness and accuracy of LLMs in this critical domain. RBRA integrates core mechanisms such as hierarchical query decomposition, dynamic context filtering and aggregation, and iterative fact verification and refinement, underpinned by Robustness-aware Metric Optimization. Zero-shot evaluations on BioMultiHopQA-Dynamic, a challenging dataset designed to rigorously assess robustness, confirm its efficacy. RBRA-GPT4o achieves state-of-the-art average accuracy (70.1\%) and robust accuracy (63.5\%). Crucially, RBRA significantly reduces the Robustness Gap to 6.6\% (RBRA-GPT4o) and 6.5\% (RBRA-Llama3-70B), marking a substantial improvement compared to baseline methods' gaps exceeding 12\%. Furthermore, RBRA empowers open-source models, enabling RBRA-Llama3-70B to surpass leading proprietary LLMs in robust accuracy. Ablation studies and detailed analyses confirm the critical contribution of each RBRA component and its superior resilience to various perturbations. RBRA thus represents a significant step towards more reliable and trustworthy AI systems in biomedical applications.