Explore the vast range of titles available.

Urban green infrastructure is increasingly recognized as a core component of urban sustainability, providing regulating ecosystem services (ES) that support climate resilience, environmental quality, and long-term urban livability. However, empirical evidence on the long-term stability of ecosystem services (ES) in rapidly urbanizing cities remains limited. Despite widespread assumptions that urbanization inevitably leads to irreversible ecological decline, few studies have quantitatively examined whether ES can persist, or even recover, over multi-decadal time horizons relevant to sustainable urban development. This study investigates the long-term trajectories of eight urban ES in Seoul, South Korea, across nearly five decades (1978–2025) and eight congressional districts, providing one of the longest temporal assessments of urban ES in East Asia. Using i-Tree Canopy and high-resolution aerial imagery across four benchmark years (1978, 1989, 2010, 2025), this study quantified standardized indicators for carbon sequestration (CSeq), avoided runoff (AVRO), and removal of six atmospheric pollutants (O3, NO2, SO2, CO, PM10, PM2.5). Paired-sample t-tests and Cohen’s dz (effect size) were used to assess within-district temporal shifts and the magnitude of ecological change. Results reveal a pronounced period of early ecological stress during rapid industrialization (1978–1989), with negative standardized effect sizes across all services (dz between −0.65 and −0.72). However, these early losses were not sustained. Structural services such as CSeq and AVRO exhibited long-term functional stability, with effect sizes converging toward zero and the 1978–2025 change in CSeq showing no statistical difference (p = 0.784). Pollutant removal services followed an early-decline–followed-by-recovery trajectory, exemplified by CO removal shifting from a large early decline (dz = −0.72) to a modest positive effect in later decades dz = 0.31). These findings indicate that Seoul’s sustained urban greening and environmental policies were effective in preventing further deterioration and maintaining core ecological functions, even if they produced stabilization rather than significant long-term gains in ES delivery.

One of the chronic problems of deep-learning models is shortcut learning. In a case where the majority of training data are dominated by a certain feature, neural networks prefer to learn such a feature even if the feature is not generalizable outside the training set. Based on the framework of Neural Tangent Kernel (NTK), we analyzed the case of linear neural networks to derive some important properties of shortcut learning. We defined a feature of a neural network as an eigenfunction of NTK. Then, we found that shortcut features correspond to features with larger eigenvalues when the shortcuts stem from the imbalanced number of samples in the clustered distribution. We also showed that the features with larger eigenvalues still have a large influence on the neural network output even after training, due to data variances in the clusters. Such a preference for certain features remains even when a margin of a neural network output is controlled, which shows that the max-margin bias is not the only major reason for shortcut learning. These properties of linear neural networks are empirically extended for more complex neural networks as a two-layer fully-connected ReLU network and a ResNet-18.

Backward error analysis allows finding a modified loss function, which the parameter updates really follow under the influence of an optimization method. The additional loss terms included in this modified function is called implicit regularizer. In this paper, we attempt to find the implicit regularizer for various federated learning algorithms on non-IID data distribution, and explain why each method shows different convergence behavior. We first show that the implicit regularizer of FedAvg disperses the gradient of each client from the average gradient, thus increasing the gradient variance. We also empirically show that the implicit regularizer hampers its convergence. Similarly, we compute the implicit regularizers of FedSAM and SCAFFOLD, and explain why they converge better. While existing convergence analyses focus on pointing out the advantages of FedSAM and SCAFFOLD, our approach can explain their limitations in complex non-convex settings. In specific, we demonstrate that FedSAM can partially remove the bias in the first-order term of the implicit regularizer in FedAvg, whereas SCAFFOLD can fully eliminate the bias in the first-order term, but not in the second-order term. Consequently, the implicit regularizer can provide a useful insight on the convergence behavior of federated learning from a different theoretical perspective.

Among physical stimulation modalities, magnetism has clear advantages, such as deep penetration and untethered interventions in biological subjects. However, some of the working principles and effectiveness of existing magnetic neurostimulation approaches have been challenged, leaving questions to be answered. Here we introduce m-Torquer, a magnetic toolkit that mimics magnetoreception in nature. It comprises a nanoscale magnetic torque actuator and a circular magnet array, which deliver piconewton-scale forces to cells over a working range of ~70 cm. With m-Torquer, stimulation of neurons expressing bona fide mechanosensitive ion channel Piezo1 enables consistent and reproducible neuromodulation in freely moving mice. With its long working distance and cellular targeting capability, m-Torquer provides versatility in its use, which can range from single cells to in vivo systems, with the potential application in large animals such as primates. A magnetic torque actuator has been developed and is capable of modulation of neurons expressing the mechanosensitive ion channel, Piezo1, resulting in long-distance control of locomotion of mice.

Innate lymphoid cells (ILCs) are innate lymphocytes that do not express antigen-specific receptors and largely reside and self-renew in mucosal tissues. ILCs can be categorized into three groups (ILC1–3) based on the transcription factors that direct their functions and the cytokines they produce. Their signature transcription factors and cytokines closely mirror those of their Th1, Th2, and Th17 cell counterparts. Accumulating studies show that ILCs are involved in not only the pathogenesis of mucosal tissue diseases, especially respiratory diseases, and colitis, but also the resolution of such diseases. Here, we discuss recent advances regarding our understanding of the biology of ILCs in mucosal tissue health and disease. In addition, we describe the current research on the immune checkpoints by which other cells regulate ILC activities: for example, checkpoint molecules are potential new targets for therapies that aim to control ILCs in mucosal diseases. In addition, we review approved and clinically- trialed drugs and drugs in clinical trials that can target ILCs and therefore have therapeutic potential in ILC-mediated diseases. Finally, since ILCs also play important roles in mucosal tissue homeostasis, we explore the hitherto sparse research on cell therapy with regulatory ILCs. This review highlights various therapeutic approaches that could be used to treat ILC-mediated mucosal diseases and areas of research that could benefit from further investigation. Immunology: the role of innate lymphoid cells Cells of the immune system called innate lymphoid cells (ILCs) patrol the soft ‘mucosal’ tissues lining the respiratory, digestive and reproductive organs, where they can be involved in both the recovery from and development of disease. Researchers in South Korea led by Hye Young Kim at Seoul National University, review recent advances in understanding the biology of ILCs in mucosal tissue health and disease. ILCs are part of the innate immune system, which elicits immediate and general responses to infection and disease, rather than responses to specific foreign substances. The authors discuss physiological mechanisms that regulate ILCs, which could be targeted for future drug development. They also review the potential of existing drugs that could target ILCs to treat disease, especially inflammatory conditions, and the possible use of ILCs as cell-based therapies for various diseases.

The quantification of supramolecular host–guest interactions is important for finely modulating supramolecular systems. Previously, most host–guest interactions quantified using force spectroscopic techniques have been reported in force units. However, accurately evaluating the adhesion energies of host–guest pairs remains challenging. Herein, using a surface forces apparatus, we directly quantify the interaction energies between cyclodextrin (CD)-modified surfaces and ditopic adamantane (DAd) molecules in water as a function of the DAd concentration and the CD cavity size. The adhesion energy of the β-CD–DAd complex drastically increased with increasing DAd concentration and reached saturation. Moreover, the molecular adhesion energy of a single host–guest inclusion complex was evaluated to be ~9.51 k B T . This approach has potential for quantifying fundamental information toward furthering the understanding of supramolecular chemistry and its applications, such as molecular actuators, underwater adhesives, and biosensors, which require precise tuning of specific host–guest interactions. Quantification of supramolecular host–guest interactions is important for modulating supramolecular systems but accurately evaluating the adhesion energies of host–guest pairs remains challenging. Here, the authors directly quantify the interaction energies between cyclodextrin modified surfaces and ditopic adamantane using a surface force apparatus.

Designing an electrocatalyst that simultaneously satisfies high catalytic activity and surface stability is essential for realizing high-performance lithium-sulfur (Li||S) batteries. Here, we propose an advanced electrocatalyst by constructing a thin protective catalytic layer (PCL) on the surface of metal nanoparticle catalysts. This few atomic layer thicknesses of the PCL composed of pyridinic N embedded graphitic carbon allows electrons to transfer from a metal nanoparticle to pyridinic N, resulting in an optimized p -orbital level of pyridinic N of PCL favorable for highly active conversion reaction of lithium sulfide. Further, PCL suppresses the direct contact of sulfur species with metal electrocatalysts. This surface protection effect inhibits the phase change of metal electrocatalysts to metal sulfide impurities, which maintains a highly active Li||S electrocatalysis for long-term cycling. Consequently, A h-level Li||S pouch cell with >500 W h kg −1 (specific energy based on current collector, anode, separator, electrolyte, and cathode), Coulombic efficiency (>95%), and stable life of 20 cycles was successfully realized. Design of electrocatalyst in lithium sulfur batteries is important to improve electrochemical performance. Here, authors use protective carbon layer to modulate the metal p -band center and prevent undesirable phase change of electrocatalyst for realizing 1 A h level lithium sulfur pouch cell.

Amyloid-β (Aβ) oligomers are implicated in the onset of Alzheimer’s disease (AD). Herein, quinoline-derived half-curcumin-dioxaborine (Q-OB) fluorescent probe was designed for detecting Aβ oligomers by finely tailoring the hydrophobicity of the biannulate donor motifs in donor-π-acceptor structure. Q-OB shows a great sensing potency in dynamically monitoring oligomerization of Aβ during amyloid fibrillogenesis in vitro. In addition, we applied this strategy to fluorometrically analyze Aβ self-assembly kinetics in the cerebrospinal fluids (CSF) of AD patients. The fluorescence intensity of Q-OB in AD patients’ CSF revealed a marked change of log ( I / I 0 ) value of 0.34 ± 0.13 (cognitive normal), 0.15 ± 0.12 (mild cognitive impairment), and 0.14 ± 0.10 (AD dementia), guiding to distinguish a state of AD continuum for early diagnosis of AD. These studies demonstrate the potential of our approach can expand the currently available preclinical diagnostic platform for the early stages of AD, aiding in the disruption of pathological progression and the development of appropriate treatment strategies. In this work, the authors characterize a small molecule fluorescent probe pioneering early diagnosis of Alzheimer’s disease through identification of amyloid-β oligomers in patients’ cerebrospinal fluid, demonstrating potential for clinical application.

Owing to the emergenceof energy storage and electric vehicles, the desire for safe high‐energy‐density energy storage devices has increased research interest in anode‐free lithium metal batteries (AFLMBs). Unlike general lithium metal batteries (LMBs), in which excess Li exists to compensate for the irreversible loss of Li, only the current collector is employed as an anode and paired with a lithiated cathode in the fabrication of AFLMBs. Owing to their unique cell configuration, AFLMBs have attractive characteristics, including the highest energy density, safety, and cost‐effectiveness. However, developing AFLMBs with extended cyclability remains an issue for practical applications because the high reactivity of Li with limited inventory causes severely low Coulombic efficiency (CE), poor cyclability, and dendrite growth. To address these issues, tremendous effort has been devoted to stabilizing Li metal anodes for AFLMBs. In this review, the importance and challenges of AFLMBs are highlighted. Then, diverse strategies, such as current collectors modification, advanced electrolytes, cathode engineering, and operation protocols are thoroughly reviewed. Finally, a future perspective on the strategy is provided for insight into the basis of future research. It is hoped that this review provides a comprehensive understanding by reviewing previous research and arousing more interest in this field. To comprehend the anode‐free lithium metal batteries, it is gone into detail in this review about current collector modification, electrolyte optimization, cathode engineering, and regulation of cycling parameters.