Search Results (527)

As large language models (LLMs) move onto resource-constrained devices, maintaining factual reliability without adding another expensive decoding pass becomes a practical inference problem. Instead of introducing another complex hallucination detector, this paper presents an empirical study of which low-cost white-box features remain useful under a controlled single-pass benchmark. Across repeated candidate-answer reruns on Qwen2.5-1.5B-Instruct and Llama-3.2-1B-Instruct, truthful and incorrect internal states are most separable in the middle-to-late layers, with the peak consistently falling at 50–70% of total network depth across both model families. The depth-relative pattern is more stable than any single detector ranking: simple residual-space baselines, including Mahalanobis scoring, remain competitive with more elaborate residual-plus-spectral fusion features under the same protocol, although detector ranking still changes by task. A separate preliminary two-seed Qwen2.5-7B-Instruct BF16 probe under that same white-box benchmark reproduces the same middle-to-late peak, and auxiliary Int8 checks on Qwen2.5-1.5B and Qwen2.5-7B remain consistent with that same localization under moderate quantization. Taken together, the results point away from detector complexity and toward a more reproducible question of where hallucination cues emerge, which internal statistics remain reliable, and how cautiously such conclusions should be transferred to deployment settings. Full article

Cattle behavior constitutes important phenotypic information reflecting animals’ health status, activity level, and welfare condition, and is therefore of considerable significance for automated monitoring and precision management in smart livestock farming. However, under complex barn conditions, cattle behavior recognition is easily affected by factors such as illumination variation, partial occlusion, background interference, and individual differences, thereby reducing recognition stability and generalization capability. To address these challenges, this study proposes a pose-driven method for cattle behavior recognition in complex barn environments. First, a 16-keypoint annotation scheme suitable for describing bovine posture, termed cow16, was constructed. Based on this scheme, OpenPose was employed to extract heatmaps (HMs) and part affinity fields (PAFs), which were then used to build an intermediate HM/PAF posture representation. Subsequently, this representation was taken as the input to a lightweight convolutional neural network for classifying three behavioral categories: stand, walk, and lying. On this basis, class-imbalance correction during training and a multi-random-seed logits ensemble strategy during inference were further introduced. In addition, knowledge distillation was adopted to transfer knowledge from a high-performance teacher model to a lightweight student model. Experimental results demonstrate that training-stage class-imbalance correction and inference-stage multi-random-seed logits ensembling exhibit strong complementarity; when combined, the AB configuration improves the test-set Macro-F1 by 3.83 percentage points. Moreover, the distilled student model still achieves competitive recognition performance while maintaining 1× inference cost, indicating a favorable trade-off between accuracy and efficiency. This study provides a useful reference for deployment-oriented cattle behavior recognition in smart farming scenarios and offers a lightweight technical basis for subsequent practical applications. Full article

Recent automated search methods have improved lower bounds for several Ramsey numbers, but the strongest gains often depend on structured seeding and cell-specific heuristic discovery. This leaves open a more fundamental question: Can a useful search structure be transferred across related Ramsey cells rather than rediscovered independently for each target instance? This work proposes a teacher–student framework for transferable structural search in Ramsey graph construction, inspired by the structure-distillation logic of Physics Structure-Informed Neural Networks (Ψ-NNs). The framework builds compressed structural representations from teacher witnesses and search traces, extracts reusable motifs and relations, and reconstructs transfer candidates. These are refined by balanced search and, for weak R(3, s) cells, by exact small-cell supervision. The framework is evaluated as a proof of concept across five Ramsey cells under transfer, matched-compute, search, ablation, and interpretability settings, including a proportional shift-scaling baseline and a greedy triangle-closing baseline that probe the structure-validity frontier from complementary directions. Supplementary experiments cover seed robustness, budget sensitivity, transfer-neighborhood variation, structural-resolution changes, stronger exact supervision, cross-r teacher pooling, single-teacher configurations, and scaling behavior across graph sizes. The results show that the portfolio version of the framework is the strongest balanced transfer method in the current study, while a structure-dominant oracle achieves stronger witness-shape agreement but worse Ramsey-valid construction. These findings reveal a clear structure-validity frontier and suggest that transferable Ramsey search should be evaluated by how well structural priors survive the validity constraints of new cells. Full article

Background/Objectives: Castanopsis tibetana Hance (C. tibetana) is a valuable timber species in southern China. Its chloroplast and nuclear genomes have been characterized, but its mitochondrial genome (mitogenome) remains unknown. This study assembles and characterizes the first complete mitogenome of C. tibetana, elucidating its structural and evolutionary features. Methods: A hybrid approach combining Oxford Nanopore long reads and Illumina short reads was used. The mitogenome was assembled via iterative seed-based mapping and annotated via GeSeq and tRNAscan-SE. Repeats were identified via MISA, TRF, and REPuter. The RNA editing sites were predicted with the PREP suite. Phylogenetic analysis was performed on 14 conserved protein-coding genes from 13 species via maximum likelihood and Bayesian inference. Results: The mitogenome is a 554,078 bp circular molecule (GC 45.27%) encoding 51 genes (32 PCGs, 16 tRNAs, 3 rRNAs). It contains 202 simple sequence repeats (37.1% tetrameric). We predicted 53 C-to-U RNA editing sites, most frequently in nad7 and nad5. Codon usage showed bias, with 28 codons having RSCU > 1. Twenty fragments (6001 bp, 1.08% of the mitogenome) were transferred from the chloroplast. Phylogenomic analysis placed C. tibetana within Fagaceae, close to other Castanopsis species. Conclusions: This study provides the first comprehensive characterization of the C. tibetana mitogenome, revealing its structural architecture, repetitive landscape, RNA editing profile, and phylogenetic placement. These findings offer valuable genomic resources for understanding mitogenome evolution in Fagaceae and support future research on the conservation genetics and molecular breeding of this important tree species. Full article

Seed priming is a simple, economical, and sustainable technique capable of enhancing crop resilience to abiotic stresses. A plastic greenhouse experiment was conducted on the durum wheat cultivar, Karim, sown in a 375 L volume container under semi-controlled conditions. Plots were arranged in a completely randomized design regarding treatments (control, salinity) and priming agents (indole-3-acetic acid, IAA; gibberellic acid, GA₃; and salicylic acid, SA). Some physiological, biochemical, and morphometric traits were analyzed at vegetative and reproductive stages. The obtained results demonstrated that salinity stress reduced plant growth and the SPAD index, hampered photosynthetic efficiency through disrupted PSII integrity and energy management in the electron transfer chain, and significantly affected ear filling (EF) and grain caliber (marked by mean weight of 100 grains, MW100G). However, seed hormonal priming allowed the alleviation of salinity stress effects on durum wheat growth and yield. Although IAA and GA₃ have shown significant potential in improving durum wheat tolerance to salinity, SA was found to be the most effective priming agent. It promotes the biosynthesis of chlorophyll pigments, restores the functional integrity of PSII, enhances photosynthetic efficiency, increases plant growth, and stimulates ear filling and wheat grain development. The principal component analysis demonstrated the interdependence of the vegetative and reproductive traits and presents SA as the most effective treatment that brings plants close to control conditions, despite the salinity. Full article

In the present study, we conducted three treatments with different cadmium concentrations and three different soil types selected for grape cultivation to assess the accumulation and migration characteristics of Cadmium (Cd) in the soil–grape system in different years. The change in Cd fractions in soil and the transfer and accumulation of Cd in different soil–grape systems were analyzed to evaluate the health risks of pulp cadmium accumulation to grape consumers. The results showed that after the planting of the grape, the active Cd fraction increased by 1~3 times and the stable fraction decreased by 1~3 times compared to before planting grapes. It gradually began to stabilize as the cultivation period progressed. The bioaccumulation factor (BCF) of Cd in different parts of grape was ranked as: root (0.094~2.590) > stem (0.117~2.112) > leaf (0.008~0.621) > seed (0.010~0.195) > skin (0.000~0.148) > pulp (0.000~0.156). High Cd concentration inhibited the transfer of Cd from the soil and root to the aboveground part of the grape. The Cd of grape pulp has no health risks. Cd pollution significantly altered the soil microbial community, suppressing Actinobacteria while enriching Acidobacteria. The results of this study will help to clarify migration patterns between different soil–grape systems and providing effective data and theoretical support for the management of Cd pollution in vineyard soils. Full article

We introduce Self-Competitive Distillation (SCD), a parameter-neutral training strategy aimed at influencing optimization dynamics without increasing model size or relying on external teachers. Two identical instances of the same architecture, initialized with different random seeds, are trained jointly and dynamically exchange asymmetric teacher–student roles based on instantaneous performance, enabling knowledge transfer between diverging optimization trajectories. Under fixed parameter and training budgets, SCD is observed to improve the realized effective capacity of lightweight architectures, yielding a higher test accuracy at matched epochs. Across multiple lightweight vision models and datasets, SCD demonstrates gains in both in-domain performance and cross-domain generalization, as measured by xScore. These results suggest that, within the evaluated experimental conditions, SCD can help mobile models make more effective use of training dynamics, while the underlying architecture remains the primary determinant of effective capacity in resource-constrained settings. Full article

Pinus oocarpa Schiede ex Schltdl. is the most important resin-producing conifer in Mexico, yet its morphological variation and seed transfer guidelines remain poorly defined for the Sierra Madre del Sur (SMS). This study evaluated variation in cone, seed, fascicle sheath, and needle traits, analyzed their associations with geographic and climatic factors, and delineated altitudinal seed zones and assisted migration distances. Most variation occurred among individual trees, with smaller but significant components among populations and provenances. All traits differed significantly among populations, provenances, and trees (p ≤ 0.0325), except for cone length, which showed no significant differences among populations (p = 0.0714). Multivariate analyses at both tree and provenance levels identified two differentiated population groups within the SMS. Several traits, including needle thickness, seed size, cone length, and seed weight, showed significant associations with environmental gradients. To realign provenances with projected climates for the 2030s, 2060s, and 2090s, upward altitudinal shifts of 165, 255, and 400 m are required. These findings support the implementation of a modified climate-adjusted provenancing strategy to reduce maladaptation risks and enhance climate resilience in reforestation and restoration programs. Full article

The fabrication of high-resolution and mechanically robust copper patterns remain a critical challenge in flexible electronics. Here, we present a universal metallization strategy that combines sequential two-step laser transfer, including laser-induced backward transfer and laser-induced forward transfer, with subsequent electroless copper plating. In this approach, laser-induced backward transfer first generates a transferable copper particle donor layer; subsequently, laser-induced forward transfer selectively embeds these catalytic copper particles into the surface of target substrates, constructing spatially confined activation networks while minimizing direct thermal exposure. These embedded seeds are then amplified into continuous copper conductors via electroless copper plating, achieving a high-resolution pattern (average minimum linewidth of approximately 20 μm) with robust interfacial integrity. Benefiting from laser-induced mechanical interlocking, the resulting copper patterns exhibit a low electrical resistivity of ~2.0 × 10⁻⁸ Ω·m (comparable to bulk copper) and maintain stable electromechanical performance even after 8000 bending cycles across a radius range of 3 to 6 mm. Furthermore, the fabricated versatile electrodes are successfully integrated into a triboelectric nanogenerator for tactile sensing and Morse code transmission. With its inherent substrate universality (e.g., polyimide, wood, fabric, and paper) and process scalability, this strategy provides a versatile route for manufacturing reliable copper electrodes in next-generation flexible electronic systems. Full article

Moist chilling is widely used to overcome seed dormancy in zoysiagrass (Zoysia japonica Steud.), but the coordinated physiological and molecular basis remains unclear. Here, freshly matured seeds were subjected to moist chilling at 4 °C in darkness for 0 (Control), 1 (CS1), 2 (CS2), 3 (CS3), or 4 weeks (CS4) and then transferred to germination conditions (30/20 °C, day/night). Prolonged moist chilling progressively improved dormancy release: final germination percentage increased from 40.5% (Control) to 73.5% (CS4), accompanied by a higher germination index and earlier, faster cumulative germination dynamics. Moist chilling also enhanced early seedling vigor, with stronger treatment differentiation in root elongation than in shoot growth. Physiologically, abscisic acid (ABA) content declined while gibberellic acid (GA) content increased, resulting in an elevated GA/ABA ratio with prolonged chilling. Metabolic activation was evidenced by increased α-amylase activity, greater soluble sugar and soluble protein accumulation, and stimulated oxygen uptake. In addition, CAT, SOD, and POD activities were enhanced under prolonged moist chilling, whereas H₂O₂ levels remained relatively stable, suggesting that redox adjustment during dormancy release was characterized by strengthened antioxidant buffering rather than pronounced oxidative accumulation. qRT-PCR supported a mechanistic transition from dormancy maintenance to germination execution, showing moist chilling-associated regulation of ABA/GA metabolism and signaling genes (e.g., NCED, CYP707A, ABI3/ABI5, and GA20ox) and downstream metabolic modules (e.g., GAMYB, AMY, ISA, INV, and HXK1), together with concurrent modulation of respiration- and ROS-related markers (e.g., AOX1a, RBOH, and CAT). Correlation analysis linked germination performance most strongly with α-amylase activity, oxygen uptake, and the GA/ABA ratio. Collectively, our data support a working model in which moist chilling rebalances the ABA–GA gate and activates downstream metabolic and redox adjustment modules to promote dormancy release and improve germination performance in zoysiagrass, providing practical markers for optimizing seed establishment through moist chilling treatment. Full article

Battery state-of-health (SoH) estimation is central to transportation electrification because it conditions safety limits, warranty accounting, power capability management, and long-horizon fleet optimization. Although deep temporal architectures can achieve high laboratory accuracy, field deployment is frequently limited by laboratory (Lab)-to-field (L2F) domain shift that alters input statistics, feature definitions, and noise regimes. Under such a shift, predictors may remain strongly monotonic, preserving degradation ordering and become operationally unreliable due to systematic output distortion (e.g., compression/warping of the SoH scale). A deployment-complete L2F transfer learning pipeline is presented, built around a gated Temporal Convolutional Network (TCN)–Transformer fusion backbone, domain-specific adapters and heads, alignment-regularized fine-tuning, and row-level inference via sliding-window overlap averaging. To address the dominant deployment failure mode, a Safe Calibration stage robustly filters calibration pairs and selects among candidate calibrators under a strict do-no-harm criterion. On an unseen deployment stream (2154 labeled rows), overlap-averaged raw inference achieves MAE = 0.0439, RMSE = 0.0501, and R² = 0.7451, consistent with mid-to-high SoH range compression, while Safe Calibration (Isotonic-Balanced selected) corrects nonlinear scaling without violating monotonic structure, improving to MAE = 0.0188, RMSE = 0.0252, and R² = 0.9357 to obtain a complete understanding of the challenges due to domain shifts, evaluation is extended to include other architecture baselines such as TCN-only, Transformer-only, Gated Recurrent Unit (GRU), and Long Short-Term Memory (LSTM), and a Ridge regression baseline. Also added is explicit alignment and calibration ablations that include CORAL off/on, that is, none vs. Safe-Global vs. Context-Aware under identical leakage-safe splits and the same overlap-averaged deployment inference operator. This work goes beyond peak-score reporting and looks at the robustness of a pipeline under domain shift, which is quantified across four random seeds and multiple deployment streams, with uncertainty summarized via mean ± std and bootstrap confidence intervals for Mean of Absolute value of Errors (MAE)/Root of the Mean of the Square of Errors (RMSE) computed from per-example absolute errors. Full article

Modern diets often provide insufficient health-promoting nutrients, prompting the development of enriched staple foods. This study investigated the impact of incorporating germinated spelt (Triticum spelta), naked oat (Avena nuda), and buckwheat (Fagopyrum esculentum) seeds at 30% and 60% levels on the nutritional, technological, and sensory properties of wheat bread. Liquid chromatography–mass spectrometry (LC–MS/MS) analysis verified the successful transfer of grain-specific bioactive compounds into the dough and bread matrix—benzoxazinoids (BOA, MBOA) from spelt, avenanthramides (AVN A, B, C) from oats, and flavonoids (e.g., rutin, vitexin, orientin) from buckwheat—emphasizing both free and bound metabolite fractions. Multigrain breads exhibited a complementary phytochemical profile. The antioxidant properties of the enriched breads were markedly enhanced, with germinated buckwheat providing the most pronounced increase. Analysis confirmed a significant increase in dietary fibre content proportional to the level of germinated grain addition, with almost double the content in 60% multigrain bread. Texture analysis indicated that the control crumb exhibited the greatest relative firming over 48 h during storage. Sensory evaluation showed that all of the enriched breads received high acceptability scores (>18/20). The incorporation of germinated seeds effectively enhances the nutritional value of bread, offering a promising strategy for developing health-promoting bakery products. Full article

There is a high risk of transfer of foodborne pathogens to the edible part of microgreens when seeds, irrigation water or soilless substrates are contaminated. Post-harvest sanitizer treatments are generally not preferred due to the fragility of microgreens. In this study, the effectiveness of post-harvest UV-C treatment was evaluated against Salmonella enterica, Shiga toxin-producing Escherichia coli O157:H7, and Listeria monocytogenes in sunflower and radish microgreens. Agricultural perlite soaked with plant nutrient solution was artificially contaminated with foodborne pathogens at a concentration of 10⁵–10⁶ CFU/g to serve as the soilless substrate. UV-C was applied to harvested microgreens uni- and bidirectionally with doubled exposure at varying distances (10, 20, and 30 cm) and exposure times (5, 10, 20, 30, 60, and 120 s). UV-C doses ranged from 0.03 to 2.07 kJ/m², depending on treatment distance and exposure time. The survival of pathogens in treated microgreens was also determined at 4 °C for 14 days. The highest pathogen inhibition was achieved with bidirectional UV-C treatment at a 10 cm distance for 120 s (p < 0.05), yielding reductions of up to 3.1, 3.0, and 2.0 log CFU/g for S. enterica, E. coli O157:H7, and L. monocytogenes, respectively. Pathogen inhibition decreased significantly with increasing distance (p < 0.05). During subsequent refrigerated storage after UV-C treatment, pathogen populations increased by 0.3–1.7 log CFU/g. These results demonstrate that UV-C treatment can significantly reduce pathogen populations on microgreens as a post-harvest treatment strategy but cannot fully address food safety concerns about these immature seedlings. Full article

The rapid synthesis of high-performance silicalite-1 membranes was systematically investigated by examining the effects of seed size, solution volume, and reactor configuration on membrane growth, microstructure, and gas separation performance. Silicalite-1 seeds (~100 nm and ~1 µm) were dip-coated onto capillary α-alumina supports, followed by secondary growth under controlled conditions. Small seeds (~100 nm) produced high nucleation density, uniform intergrowth, and defect-free membranes, yielding consistently high ideal separation factor for H₂/SF₆ (181–295) and low SF₆ permeance (~10⁻⁹ mol m⁻² s⁻¹ Pa⁻¹) after only 45 min of synthesis. In contrast, larger seeds (~1 µm) enabled faster growth but resulted in less uniform layers with inferior selectivity. Furthermore, a novel reactor design with enhanced heat transfer enabled the rapid silicalite-1 membrane synthesis on conventional large-diameter tubular supports, producing well-intergrown and uniform membranes with high H₂ permeance (4.7 × 10⁻⁶ mol m⁻² s⁻¹ Pa⁻¹) and high ideal separation factors of up to 349 for H₂/SF₆ and 223 for N₂/SF₆. Overall, this study demonstrates that optimization of seed properties, synthesis parameters, and reactor design enables rapid and scalable fabrication of silicalite-1 membranes with robust molecular sieving performance, highlighting their strong potential for SF₆ purification applications. Full article

Magnetic fields, either imposed externally or produced spontaneously, are often present in laser-driven high-energy-density systems. In addition to changing plasma conditions, magnetic fields also directly modify laser–plasma interactions (LPI) by changing the participating waves and their nonlinear interactions. In this paper, we use two-dimensional particle-in-cell (PIC) simulations to investigate how magnetic fields directly affect crossbeam energy transfer (CBET) from a pump to a seed laser beam when the transfer is mediated by the ion-acoustic wave (IAW) quasimode. Our simulations are performed in the parameter space where CBET is the dominant process and in a linear regime, where pump depletion, distribution function evolution, and secondary instabilities are insignificant. We use a Fourier filter to separate out the seed signal and project the seed fields onto two electromagnetic eigenmodes, which become nondegenerate in magnetized plasmas. By comparing the seed energy before CBET occurs and after CBET reaches quasi-steady state, we extract the CBET energy gains for both eigenmodes in lasers that are initially linearly polarized. Our simulations reveal that, starting from a few MG fields, the two eigenmodes have different gains, and magnetization alters the dependence of the gains on laser detuning. The overall gain decreases with magnetization when the laser polarizations are initially parallel, while a nonzero gain becomes allowed when the laser polarizations are initially orthogonal. These findings qualitatively agree with theoretical expectations. Full article