Reprogramming of cellular signaling associated with diseases involves a concerted effort on the parts of transcriptional programs, chromatin dynamics, pathway activities, and cellular interactions. This work established a method of using single-cell multi-omics to define diverse cell states relevant to diseases and the dynamic signaling reprograming. After quality control, a total of 68,421 single cells/nuclei profiles were used, including 35,876 disease-related cells and 32,545 control-related cells/nuclei. Weighted k-nearest neighbor integration revealed 17 clusters of cells, and the disease-enriched abnormal cells grew from 0.052 ± 0.014 in controls to 0.241 ± 0.031 in diseases, resulting in an enrichment of +1.42 ± 0.18*. In addition, intermediate cell states were enriched significantly, reaching an enrichment of +0.77 ± 0.12*. Pseudotime analysis showed that the pseudotime value in disease cells was 0.67 ± 0.12*, higher than the control group, which was 0.24 ± 0.09*. The signaling that showed the most significant change was inflammatory signaling (a trajectory coefficient of +0.86 ± 0.10*), followed by hypoxia response (+0.69 ± 0.08*) and stress response (+0.74 ± 0.09*). The multi-omics analysis also found 1,284 up-regulated genes, 18,742 disease-enriched chromatin peaks, and 4,216 peak-to-gene connections.

With the deep penetration of artificial intelligence in education, its impact on educational equity needs to be clarified. This study uses a quasi-experimental design to control interfering variables through randomized blocks and compare the multi-dimensional impact of rule-driven (RBI) and data-driven algorithm (ABI) AI intervention strategies on educational equity. The study found that the ABI strategy has an advantage in optimizing opportunity equity and outcome equity, significantly improving access to quality resources (ERFI=0.81) for students with low SES and reducing the variance of performance gain, while the RBI strategy performs better in process equity, procedural justice and transparency. The conclusion indicates that a single logic cannot balance efficiency and ethics. Building a mixed intervention mechanism of "rule as the bottom line and algorithm as the efficiency" is the best solution for achieving educational process equity.

The work offers a methodology for low-carbon intelligent path planning in machining operations of carbon fiber reinforced polymers used in aerospace parts. In this case, the object of research is a lightweight bracket made from a carbon composite material with an irregular profile, two different slots, and areas of variable thickness. A set of 30 samples of such brackets was developed using Siemens NX 2306 software. Machining data have been collected on a VMC 850 machine equipped with Yokogawa WT5000 energy meter with a period of data sampling of 0.1 seconds. Genetic algorithm path optimization was carried out in MATLAB R2024a software, whereas data processing and computation of carbon footprint values was done with Python 3.11. The comparison of optimized paths was done against reference trajectories provided by Siemens NX under equal conditions of tools, spindles, feeds, depth of cuts, and inspections. Optimization showed a path shorter by 286.4 ± 42.7 mm, travel distance by 214.8 ± 36.5 mm, machining duration by 18.6 ± 3.9 seconds, energy consumed by 0.184 ± 0.037 kWh, and carbon emissions by 0.105 ± 0.021 kg CO2, with surface roughness being unchanged at 1.42 ± 0.16 μm.

With the development of large language models (LLMs) and generative artificial intelligence, non-player characters (NPCs) are shifting from traditional rule-driven systems to generation-driven systems. Traditional NPCs mainly rely on finite state machines and behavior trees to implement predefined behavioral logic. Although these methods provide strong stability and controllability, they cannot fully meet the demand for high freedom, strong immersion, and natural interaction in open-world games. Generative intelligent NPCs introduce memory, reasoning, planning, and natural language understanding capabilities, allowing game characters to evolve into agents with persistent interaction abilities. This paper examines the evolutionary history, interaction loops, system architecture, functional paradigms, industrial implementation, and alignment challenges of generative intelligent NPCs. Drawing on the theory of the digital labyrinth and wax wings, the Generative Agents framework, LLM-based agent architecture, and industrial cases such as NVIDIA ACE and NetEase Fuxi, this paper analyzes the application paths of generative NPCs in emotional companionship, strategic opposition, and real-time interaction. The study argues that generative NPCs are transforming game interaction from script-driven content delivery to experience-driven world building. However, issues such as behavioral loss of control, value drift, computational cost, and ethical governance still need further attention. In the future, balancing intelligence, real-time response, and controllability will become the core direction for the sustainable development of generative NPCs.

Cracks in mechanical structures cause numerous hazards, such as stress concentration, reduced structural strength and fatigue performance; continuous crack propagation easily leads to component fracture and equipment failure; it results in downtime maintenance and increased costs, and even safety accidents in severe cases. Therefore, accurate crack detection is crucial. Traditional detection relies on manual work, which has pain points such as low efficiency, strong subjectivity, obvious missed detection of micro-cracks, and difficulty in identifying internal hidden cracks, failing to meet the needs of real-time large-scale detection. This study takes YOLOv5 as the core and combines deep learning technology to carry out research on mechanical crack detection. A mechanical crack dataset is constructed, and sample annotation and data augmentation are completed to improve sample diversity; parameters are modified based on the YOLOv5 model, and after multiple rounds of training iteration and verification, the model finally achieves a mAP@0.5 of 31.37%, effectively reducing the missed detection rate and false detection rate; after model deployment, it can realize fast, automatic and high-precision identification and location of mechanical surface cracks, effectively alleviating the core pain points of traditional detection, and providing an efficient and feasible technical solution for mechanical structure crack detection.

Aiming at the problems of insufficient prediction accuracy, weak interpretability and inconsistent data caliber of traditional models, this paper constructs a Bayesian Optimization-LightGBM model, and combines SHAP values to carry out grain yield prediction and influencing factor analysis. The research results show that the coefficient of determination is 1.0000, the root mean square error is 10.7267, and the mean absolute error is 6.4542, indicating excellent model fitting accuracy. The national grain yield in 2026 will still be highly concentrated in major producing areas, with Heilongjiang, Henan and Shandong ranking the top three steadily. The prediction results are consistent with the actual production pattern. SHAP value analysis shows that sown area indicators are the core driving factors. This study provides effective methods and reference basis for accurate grain yield prediction and policy formulation.

Rapid urbanization has promoted the continuous expansion of urban built-up space. Green spaces, water bodies, woodlands, and open spaces have been increasingly divided by roads, residential areas, commercial land, and high-intensity functional zones. This process has caused ecological patch fragmentation, blocked ecological flows, and reduced the accuracy of corridor identification. To support the optimization of urban landscape ecological networks, this study takes the central urban area of Guangzhou as the case area. It integrates Sentinel-2 remote sensing imagery, 30 m land-use data, POI data, and road data from 2020 to 2024. A multi-source spatial data-driven method is developed to evaluate urban landscape ecological connectivity. NDVI and MNDWI are used to extract vegetation and water information. MSPA is combined to identify ecological sources. Land-use resistance, road proximity resistance, built-up intensity, POI kernel density, and the inverse NDVI indicator are integrated into a comprehensive ecological resistance surface. The MCR model, circuit theory, and graph-theoretic indicators are then used to extract potential ecological corridors and identify pinch points, barrier points, and key restoration nodes. The results show that ecological spaces in central Guangzhou present a pattern of relative continuity in peripheral areas and clear fragmentation in the urban core. The Pearl River green belts, Haizhu Wetland, the southern edge of Baiyun Mountain, and large parks form the main ecological connection framework. Road intersections, high-density commercial areas, and hardened riverbanks still create significant barriers. The study provides a computational basis for green-space restoration, ecological corridor planning, and landscape spatial optimization.

The reasoning illusion of large language model is the core bottleneck restricting its safe landing. The existing research mostly focuses on post-event detection and lacks pre-prediction mechanism. In order to improve the accuracy of illusion prediction and generalization performance, a BiLSTM-Attention classification algorithm based on RBMO optimization is proposed in this paper. The model uses a two-way long-short-term memory network to capture the pre-and post-sequence dependencies, focuses on key features with the help of attention mechanism, and introduces the red-billed blue magpie optimization algorithm to adaptively optimize the network weights and hyperparameters, so as to alleviate the local optimal problem of gradient descent. A standardized data set containing 1729 samples was constructed in the experiment, and a variety of machine learning algorithms were selected for comparison and verification. The results show that the proposed algorithm is optimal in accuracy, recall, accuracy, F1 value and AUC value, and all indexes are 0.975 or above, which is overall superior to traditional machine learning and single deep learning models, and can provide effective technical support for large model hallucination pre-prediction and safe deployment.