Browse Articles

Overlap locking and nonperturbative effects in spin glasses

We study the phenomenon of the locking of the order parameter (or synchronization) in spin glasses at low temperatures. When two systems with independent disorders are coupled, their overlaps become similar. A crucial question is how this effect depends on the strength of the coupling between the two systems. Nonperturbative phenomena are present when 1 ≪ Δ H ≪ N , being Δ H the coupling Hamiltonian and N the system size. In this intermediate-coupling region, the effect is related to finite-size free-energy corrections in mean-field spin-glass models and to correlations in the Dyson hierarchical spin glass, a model that mimics the physics of finite-dimensional systems. We study this phenomenon in the mean-field approach, both analytically and numerically, and we finally compute the critical exponents for finite-volume corrections in mean-field theory and for the decay of correlations in the Dyson hierarchical model.

Trust-driven healthy engagement with conversational AI for mental health support in young adults: a mixed methods study

MYDGF promotes pathological and physiological retinal angiogenesis via the Gαi1/3-Gab1-Akt-mTOR signaling

TFPI2 promotes NK cell–mediated glioblastoma killing through adhesion and checkpoint control

Immune-mediated killing triggers dynamic transcriptional adaptations in tumor cells that can reciprocally regulate the cytolytic process. Unraveling such feedback mechanisms is crucial for advancing cancer immunotherapy. Here, we identified tissue factor pathway inhibitor 2 (TFPI2) as a central node in natural killer (NK)–glioblastoma cross talk. Using transcriptomic and functional approaches, we demonstrated that NK cell attack induces TFPI2 expression in glioblastoma cells via IL1β- and TNFα-driven activation of NFκB signaling. TFPI2 not only restrains tumor proliferation by suppressing the POU2F2–CCND1 axis but also enhances NK cytotoxicity through two complementary mechanisms: It supports optimal ICAM1 expression to promote NK–tumor adhesion, and it selectively represses the immune checkpoint molecule SIGLEC15, restoring NK cell effector function. In vivo, loss of TFPI2 accelerates glioblastoma progression and abrogates the efficacy of adoptive NK cell therapy in a context-dependent manner; the functionality is likely restricted to tumors retaining the capacity for TFPI2 induction upon inflammatory stimuli. Our findings identified the TFPI2–ICAM1 and TFPI2–SIGLEC15 axes as conditional regulators of immune–tumor adhesion and checkpoint control, supporting TFPI2 as a candidate therapeutic target for a subset of glioblastomas amenable to inflammatory reprogramming.

Insights into germination, physiological, and molecular changes in aniseeds induced by magnetic fields

Abstract Improving seed germination is essential for enhancing crop establishment under increasingly variable environmental conditions associated with climate change. Magnetic field (MF) treatment represents a clean, non-chemical, and sustainable seed-priming approach; however, frequency-dependent biological responses remain insufficiently understood. This study investigates the responses of aniseed ( Pimpinella anisum L.) to static (DC) and low-frequency alternating magnetic fields (5, 10, and 15 Hz) across different exposure durations. Germination parameters (percentage, speed, vigor index), physiological traits, and activities of key hydrolytic and antioxidant enzymes (α-amylase, protease, catalase) were assessed. Furthermore, the molecular expression of the stress-responsive superoxide dismutase (SOD) and the cytoskeletal actin genes was analysed. MF significantly enhanced germination percentage (up to a 25% increase), mean germination time, and vigor indices compared to the untreated controls. Physiologically, treated seedlings exhibited higher antioxidant defense levels. At moderate frequencies, catalase activity increased, while α-amylase and protease were markedly elevated at higher frequencies, enabling reserve mobilization and stress tolerance. At the molecular level, sod transcripts were down-regulated across all MF treatments compared to the control, indicating a functioning oxidative stress response. These findings demonstrate that MF frequency modulates the integration of physiological (enzyme-driven metabolism) and molecular (antioxidant gene regulation) pathways to optimize during aniseed germination. This research provides mechanistic insights and presents low-frequency MF as a viable seed priming for sustainable crop improvement under dynamic environments.

MiT fusions, TSC1–TSC2 divergence, and stem-like programs reveal distinct origins and vulnerabilities in PEComa

Collapsible scissored surfaces

We introduce an additive approach for the design of a class of transformable structures based on two-bar linkages (“scissor mechanisms”) joined at vertices to form a two-dimensional mesh which we call a pantograph lattice. Our approach shows how these lattices unfold from a one-dimensional collapsed state to two-dimensional surfaces of single and double curvature. We provide an algorithm for growing pantograph structures that allows us to explore the full space of possible mechanisms, and we use it to computationally design and physically assemble a series of examples of varying complexity. We finally demonstrate a streamlined method for automated fabrication of pantograph lattices using multimaterial 3D printing.

3D-FPFH-Int: a hybrid geometric–radiometric descriptor for structural surface anomaly detection in tropical heritage

Abstract Anomaly detection in tropical heritage structures is often constrained by geometry-only point-cloud descriptors that inadequately capture moisture-related radiometric variation, while prior evaluations frequently confound descriptor contributions with detector-specific behaviour and provide limited statistical attribution. This study isolates the contribution of a hybrid geometric–radiometric descriptor through controlled multi-detector validation. We introduce 3D-FPFH-Int, extending Fast Point Feature Histogram with local three-dimensional intensity histograms. An explicit ablation study (geometric-only, radiometric-only, hybrid) is conducted across three detectors—PatchCore, Isolation Forest, and kNN—using 500 synthetic instances (including adversarial weak-contrast subsets) and three non-overlapping field scan segments (two bridge spans and one tunnel segment) comprising 28.9 million points acquired via terrestrial LiDAR. Training and evaluation employ spatially disjoint partitions to prevent data leakage. The statistical protocol includes two-way ANOVA (Descriptor × Detector), Bonferroni-adjusted post-hoc comparisons, bias-corrected and accelerated 95% bootstrap confidence intervals, Cohen’s d with confidence bounds, and post-hoc power analysis (1 − β = 0.82 for moderate interaction effects, η 2  ≥ 0.05). The hybrid achieves a weighted mean F 1  = 0.559 [0.538–0.580], representing a 114% relative improvement over FPFH-only (Cohen’s d = 1.38 [1.27–1.49]). The Descriptor × Detector interaction was not statistically significant (p = 0.15, η 2  = 0.02), indicating that the relative ranking of descriptors remains broadly consistent across the evaluated detectors within the tested conditions. Under 40% intensity contrast reduction, ΔF 1 remains + 0.294 relative to FPFH. Early crack detection (0.3–0.5 mm) yields F 1  = 0.158 with localization error < 16 mm. Moisture-related anomaly detection achieves recall = 0.85 [0.80–0.90] with 68% fewer condensation-induced false positives than intensity-only baselines. Performance degradation remains < 12% under ± 20% point-density perturbation and controlled intensity noise. Validation is restricted to masonry and concrete structures in tropical humid environments using terrestrial LiDAR; generalization to other materials, climates, or sensing modalities requires independent verification. Code and synthetic data are publicly available to support reproducibility.

Reasoning in machine vision by learning fast and slow thinking

Abstract Reasoning is a hallmark of human intelligence, enabling adaptive decision-making in complex unfamiliar scenarios. In contrast, machine intelligence remains bound to training data, unable to dynamically refine solutions at inference. While recent advances have explored machine reasoning - trading inference-time compute for improved performance - they focus on verbal domains such as mathematical problem-solving where explicit rules govern step-by-step solution generation. Many tasks lack sufficient labelled data and require alternative performance improvement mechanisms, such as inference-time compute. Here we present a paradigm for machine reasoning in vision, enabling performance improvements with increasing thinking time (inference-time compute), even with limited labelled data. Our approach is inspired by dual-process theories of human cognition, integrating a fast-thinking System I module for generating and verifying solutions in familiar tasks, with a slow-thinking System II module that iteratively refines predictions using self-play reinforcement learning, even when task-specific data is limited. This paradigm involves proposing, competing over, and refining solutions until convergence. We demonstrate that extended inference-time compute yields superior performance compared to large-scale supervised learning, foundation models, and human experts in vision tasks. These include computer-vision benchmarks and cancer localisation across five organs, highlighting the potential of inference-time compute for data-scarce problems.

AI-driven vibration-based event classification in railway switches and crossings

Abstract Automated condition monitoring of railway switches and crossings (S&C) requires classification models whose reported accuracy reflects genuine generalization rather than evaluation artefacts. This paper presents a methodologically rigorous, leak-free machine-learning framework for vibration-based event classification, evaluated on accelerometer data from a full-scale outdoor S&C test facility. The pipeline enforces strict ordering (split, select, augment, standardize, train, evaluate) and partitions the data at the level of physical events, so that all measurements of a given event are assigned together to either the training or the test subset. A symmetric tabular autoencoder generates synthetic minority-class samples through latent-space interpolation. Twenty-one classifiers spanning eight families are benchmarked on held-out data and by group-aware five-fold cross-validation. The strongest models reach 81.5% held-out accuracy (ROC-AUC $$\approx 0.94$$ ) and $$80.4\%\pm 2.1\%$$ under cross-validation; ensemble methods are the most stable. Feature standardization is essential: without it, neural networks collapse below chance level. Computational profiling (inference latency 0.005–0.63 ms per one-second segment; model size 0.002–2.4 MB) maps three deployment scenarios to specific algorithm recommendations. Because the minority crossing class has only six held-out samples, its per-class metrics carry wide confidence intervals and should be interpreted with caution.

A belt-buckle checkpoint regulates the onset of botulinum neurotoxin intoxication

Abstract Fast-acting botulinum neurotoxins (BoNTs) are highly desirable for both medical and aesthetic indications, but the underlying mechanism for the differing onset of BoNTs’ action remains unknown. Here, we demonstrate that the “belt” of BoNTs, a largely unstructured loop wrapping around their catalytic light chain (LC), is key to onset of intoxication. The more flexible BoNT/E belt promotes quicker LC translocation into the neuronal cytosol, leading to faster onset of action compared to BoNT/A. Furthermore, we discover a “belt-buckle” checkpoint that regulates this process. By loosening the BoNT/A belt-buckle via protein engineering, we enhance its sensitivity to acidic pH, leading to an accelerated onset of action. Conversely, locking the belt-buckle with an antibody neutralizes BoNT/A. Our findings open avenues for developing fast-acting BoNTs and effective countermeasures.

Numerical study of viscous and rarefaction effects on choked flow in two-dimensional convergent conical nozzle

Creation of an air-stable surface electrene and its application to ammonia synthesis

Binomial prolate spheroidal functions, Pascal matrices, and arithmetic of elliptic curves

The ( N + 1 ) × ( N + 1 ) symmetric Pascal matrix T N is a generalized discrete time and band-limiting operator for the binomial transform and its eigenvectors are generalized discrete prolate spheroidal wave functions which we call binomial prolates. Their generating functions are also generalized prolate spheroidal functions in the sense that they are simultaneously eigenfunctions of a third-order differential operator and an integral operator over the line { z ∈ C :   Re ( z ) = 1 / 2 } . For even, positive integers N , we obtain an explicit formula for the generating function of an eigenvector of the symmetric Pascal matrix with eigenvalue 1. When N = p − 1 for an odd prime p , we show that the generating function is equivalent modulo p to ( # E z ( F p ) − 1 ) 2 , where # E z ( F p ) is the number of points on the Legendre elliptic curve y 2 = x ( x − 1 ) ( x − z ) over the finite field F p . Furthermore when N = p n − 1 , our generating function is the square of a period of E z modulo p n in the open p -adic unit disk.

Psychological resilience, physical self-efficacy, and physical activity among college students: developing and testing a model based on social cognitive theory

Learning missing physics from legacy simulators with alternating neural integrators

Targeting the DNA methylation–H3K27me3 switch reverses castration resistance and immunosuppression via ADAMTS1-driven collagenolysis

Castration-resistant prostate cancer (CRPC) lethality arises from epigenetic-driven resistance to androgen deprivation therapy (ADT). Here, we uncover a compensatory epigenetic switch between DNA methylation and H3K27me3-mediated repression as a critical barrier to epigenetic therapy in CRPC. Integrative multiomics analyses reveal that DNMT inhibitors (DNMTis) trigger EZH2-dependent H3K27me3 accumulation at the ADAMTS1 locus—a master collagenase essential for extracellular matrix (ECM) remodeling—perpetuating fibrotic niche formation and therapy resistance. Dual targeting of DNMTs and EZH2 disrupts this epigenetic plasticity, synergistically reactivating ADAMTS1 to degrade collagen-rich stroma, suppress FAK/MAPK mechanotransduction signaling, and reverse epithelial–mesenchymal transition (EMT). Crucially, in immunocompetent models, this strategy achieves >90% tumor suppression and reverses immunosuppression by enhancing cytotoxic CD8 + T cell infiltration 11.4-fold while depleting immunosuppressive macrophages and Tregs. Mechanistically, dual therapy inactivates the FAK/MAPK/EMT axis via ADAMTS1-mediated ECM degradation, overcoming stromal-mediated resistance. Our work establishes epigenetic-ECM coevolution as a hallmark of CRPC and provides a rationally designed combination therapy to dismantle the therapy-resistant niche.

Energy-efficient wireless network control via spatio-temporal deep learning and multi-agent reinforcement learning

Two-dimensional surface melting with an intermediate quasi-hexatic layer

Photoexcitation induces translocation of a common fluorescent pH and proton-transfer probe confined in reverse micelles

The photoacid 8-hydroxypyrene-1,3,6-trisulfonate (HPTS) is one of the most widely used fluorescent probes for studying proton transfer and local pH in systems from advanced materials to plants, environmental sensors to medicine. HPTS exists as two different species: the acid and its conjugate base, which lead to unique protonation-state-dependent translocation of the molecule when it is nanoconfined within anionic AOT reverse micelles. Using steady-state and time-resolved optical spectroscopy, molecular simulations, and IR solvation shell spectroscopy, we report that the protonated HPTS species associates strongly with the micelle interface via hydrogen bonding. In contrast, its deprotonated species resides in the micelle’s aqueous interior. Our results show that photoexcitation of the acid species and its subsequent deprotonation leads the conjugate base to rapidly move away from the interface into the water pool. This light-induced translocation, an effect observed for a range of micelle sizes, challenges the prevailing view where molecular probes are assumed to be static reporters of their environments, remaining in a fixed location for the duration of an experiment. This is especially relevant for interpreting results in the numerous studies enlisting optical spectroscopy of HPTS to report on complex systems. Our findings reveal the potential for molecular probes as dynamic explorers capable of mapping environmental heterogeneity on the timescale of the very processes they are designed to measure.