Nanyang Normal University

Lung cancer is among the malignancies with the highest incidence and mortality rates worldwide, and it poses a serious threat to human health. Increasing the accuracy of pulmonary nodule detection in CT images is essential for the early diagnosis and treatment of lung cancer. However, the grayscale characteristics of lung CT images, together with the variability in the sizes and morphologies of nodules, make the existing detection models prone to false positives and false negatives, posing challenges for achieving accurate detection. To address these problems, an improved WTAM-YOLO model based on YOLOv11 is proposed in this study. The model features four main improvements: a wavelet convolution approach to expand the receptive field, a lightweight convolutional block attention module (CBAM) to enhance the key feature representations, a hierarchical residual attention mixer (HRAMi) module to improve the multiscale detection performance of the model, and an improved exponential moving average (iEMA) module to strengthen the detail capture ability of the model and reduce the number of false positives. Experiments are conducted with a pulmonary nodule dataset acquired from the Roboflow platform and the LUNA16 dataset. Compared with those of YOLOv11, the proposed model improves the mAP@50 values by 3.4% and 2.5%, the mAP@75 values by 9.5% and 7.0%, the precision values by 4.4% and 0.7%, and the recall values by 2.3% and 4.6% based on the Roboflow and LUNA16 datasets, respectively.

Coprophagous flies can convert livestock manure into protein-rich larval biomass for animal feed, but manure-based rearing raises biosafety concerns. This study characterized the internal bacterial community dynamics across development in Aldrichina grahami and Boettcherisca peregrina reared on swine manure, aiming to identify developmental stages with a lower microbial hazard profile. Using 16S rRNA gene amplicon sequencing of pooled internal samples, we analyzed communities from third-instar larvae, dispersing-stage larvae, pupae at multiple time points, and newly emerged adults. Developmental stage strongly structured bacterial composition and altered richness in both species. Communities were dominated by Bacillota and Pseudomonadota, reflecting substrate origin, with pronounced turnover during metamorphosis and stage-specific dominance patterns, indicating developmental filtering rather than uniform microbial clearance. Crucially, dispersing larvae did not show the marked dominance signatures seen in later pupal or adult stages, supporting this stage as a pragmatic harvest window with a comparatively lower microbial-hazard indicator profile. Since downstream processing such as drying or heating will further reduce viable hazards, stage selection serves as an effective upstream control to lower the initial hazard burden entering production.

The yellow peach moth, Conogethes punctiferalis, is a destructive polyphagous pest and poses a severe threat to the fruit industry and field crops worldwide with its continuously increasing population and expanding host range in recent years. Despite the severe damage caused by C. punctiferalis larvae, their antennae and mouthparts, equipped with abundant sensilla responsible for feeding behavior, have not been investigated in detail. In our study, the antennae, mouthparts, and associated sensilla of first-instar and mature larvae of C. punctiferalis were examined with light and scanning electron microscopy. Our results revealed no obvious morphological differences between the two instars in the basic composition of the antennae and mouthparts, or in the types, distribution, and numbers of sensilla. The antenna is three-segmented, with no sensilla on the scape, three sensilla basiconica and two sensilla chaetica on the pedicel, and three sensilla basiconica and one sensillum styloconicum on the flagellum. The mouthparts of C. punctiferalis are typically mandibulate and consist of a labrum-epipharynx, paired mandibles, a pair of maxillae, a labium, and a hypopharynx. Six types of sensilla were primarily concentrated on the labrum-epipharynx, maxilla, and labial palp, including sensilla chaetica, sensilla basiconica, sensilla styloconica, sensilla digitiformia, sensilla epipharyngea, and sensilla placodea. We conducted a systematic analysis of the characteristics of sensilla and discussed their variation in the context of Lepidoptera phylogeny. The potential functions of the sensilla have also been inferred. The study could advance our understanding of the behavioral ecology of C. punctiferalis and provide potentially useful information on the development of pest control technologies.

Large Language Models (LLMs) increasingly leverage Federated Learning (FL) to utilize private, task-specific datasets for fine-tuning while preserving data privacy. However, while federated LLM frameworks effectively enable collaborative training without raw data sharing, they critically lack built-in mechanisms for regulatory compliance like GDPR’s right to be forgotten. Integrating private data heightens concerns over data quality and long-term governance, yet existing distributed training frameworks offer no principled way to selectively remove specific client contributions post-training. Due to distributed data silos, stringent privacy constraints, and the intricacies of interdependent model aggregation, federated LLM unlearning is significantly more complex than centralized LLM unlearning. To address this gap, we introduce Oblivionis, a lightweight learning and unlearning framework that enables clients to selectively remove specific private data during federated LLM training, enhancing trustworthiness and regulatory compliance. By unifying FL and unlearning as a dual optimization objective, we incorporate 6 FL and 5 unlearning algorithms for comprehensive evaluation and comparative analysis, establishing a robust pipeline for federated LLM unlearning. Extensive experiments demonstrate that Oblivionis outperforms local training, achieving a robust balance between forgetting efficacy and model utility, with cross-algorithm comparisons providing clear directions for future LLM development.

• A moisture threshold governs when unfrozen STC exceeds frozen STC. • Crossover S r *​ predicted from measured properties matches field diagnostics. • Calibrated K e (S r ) schemes outperform the original Johansen scheme. • Bulk density, porosity, and quartz fraction shift S r * across sites. Soil thermal conductivity (STC) governs near-surface heat exchange and constrains simulations of active-layer evolution and permafrost change. Using a 10-year record from four Qinghai–Tibet Plateau sites (0–10 cm), laboratory Kersten number ( K e )–saturation ( S r ) calibrations, and a structure-aware Johansen implementation, we identify a moisture-threshold reversal: under low antecedent moisture the frozen state conducts less heat than the unfrozen state, while at higher moisture the conventional ordering returns. The crossover saturation S r * is traceable in calibrated K e – S r relations and observable from pre-freeze moisture, linking field diagnosis to model parameters. A compact, deployable correction follows: taper the frozen branch for S r < S r * , compute endmembers from measured bulk density, porosity, and quartz fraction (BD–n–q), and select the unfrozen K e ( S r ) form by soil class and dryness tendency. The scheme reduces unfrozen-season errors across the core sites and generalizes at an independent hold-out station (TGL) without site-specific tuning. The approach is transparent—inputs are observable and decisions are tied to S r * —and is most impactful in dry, coarse, and sparsely monitored regions.