Metaphor translation quality estimation requires models to track shifts in conceptual structure across languages, rather than simply comparing surface similarity. This study proposes a cognitive-linguistics-driven prompting framework that injects conceptual metaphor information into a dual-encoder architecture and calibrates its decisions through a meta-learned transferable validator. Cognitive prompt templates encode source–target domain mappings, imageability, and contextual constraints, while a contrastive objective encourages consistent alignment between metaphorical inputs and their translations. A meta-learning layer further adapts validation weights to new metaphor families and language pairs with limited supervision. Experiments on two bilingual datasets (English–Chinese and English–Spanish, 7,420 and 6,985 annotated sentence pairs respectively) show that the framework improves correlation with human ratings, conceptual mapping consistency, and metaphor retention in both in-domain and zero-shot transfer settings. Quantitative analyses and ablation studies indicate that cognitive prompting contributes most of the gains in conceptual alignment, whereas transferable validation stabilizes performance under domain and language shifts. The findings suggest that cognitively grounded prompting can bridge linguistic theory and neural evaluation, providing interpretable and robust decisions for metaphor translation quality estimation across languages.

The intelligent manufacturing industry is accelerating its upgrading. Industrial robots have become the core support of the flexible production system, and their operational status directly affects the continuous operation capacity and safety of the production line. In large-scale production scenarios, industrial robots are constantly operating under high loads and multiple conditions. Potential faults such as joint wear, abnormal load transmission, and motor performance degradation tend to accumulate continuously. If not predicted in time, it can easily lead to unplanned shutdowns, resulting in reduced production efficiency and increased operation and maintenance costs. Aiming at the problems of insufficient feature focusing and inadequate dynamic correlation capture in the fault prediction of industrial robots by existing machine learning algorithms, this paper proposes the RBMO-BiLSTM-Attention classification algorithm. Firstly, violin graph analysis and correlation analysis are carried out, and then comparative experiments are conducted through multiple machine learning algorithms. The results show that among the traditional machine learning algorithms, decision trees and random forests perform relatively outstandingly, with accuracy and recall rates both reaching 94.1%, while the performance of other algorithms is slightly inferior. The proposed algorithm has significant advantages in all evaluation indicators. Its accuracy rate, recall rate, precision rate, F1 value all reach 95.9%, and the AUC value is as high as 97.8%. It has stronger feature extraction capabilities and classification performance, and can more accurately complete the risk level classification task of industrial Internet of Things devices. This research provides an effective technical solution for the precise prediction of industrial robot faults, which is of great significance for reducing unplanned downtime, improving production efficiency and the economic efficiency of operation and maintenance.

Rising demand and persistent capacity constraints in mental health care have exposed the limitations of rule-based triage and single-shot trial designs, especially under volatile symptoms, delayed outcome feedback, and fragmented digital data. To address these challenges, this study develops a reinforcement learning–driven counterfactual policy evaluation framework that models care allocation as a partially observable Markov decision process and combines doubly robust off-policy estimation with conservative policy improvement. In-hospital follow-up trajectories and community wearable data are integrated to construct latent state representations, and candidate policies for visit frequency, session duration, and intervention modality are evaluated offline for value and safety before deployment. On real-world cohorts, the doubly robust estimator achieves an 8.3% improvement in policy value over the baseline while retaining 76.2% of the original effective sample size and reducing bias and mean squared error under optimized clipping. Operationally, the learned policy shortens median waiting time to 8.7 days, increases per-staff throughput by 22.6%, and lowers readmission rates, while opportunity and treatment intensity gaps shrink and stability metrics improve. These findings indicate that RL combined with counterfactual evaluation can support more efficient, fair, and auditable dynamic allocation of mental health services under fixed resources and offer a transferable pipeline from offline data to safe policy deployment.

Generative AI has rapidly become foundational infrastructure for knowledge-intensive urban industries, yet empirical evidence on how it reshapes skill structures and matching mechanisms at the city level remains limited. Most existing work relies on occupational exposure indices and local productivity evaluations, and it rarely delivers causal identification of skill reallocation. This paper builds an unbalanced city–occupation–week panel for 2018–2025 that combines online vacancy postings, firm adoption signals and city-level digital infrastructure. We use a large language model to perform instruction-based skill extraction and ontology alignment, and we construct fine-grained measures of skill shares, diversity and embedding-based migration. On top of these measures, we build a GenAI diffusion index at the city–time level and estimate its effects using a staggered difference-in-differences design with event-time coefficients, complemented by a shift-share instrumental variable strategy. These findings suggest that generative AI operates in urban labor markets primarily through skill reallocation rather than simple job destruction, with task bundles inside jobs being reshaped toward more collaborative and coordination-intensive activities. The study provides quantitative support for reskilling policies and firm-level human–AI task design that target specific skill dimensions instead of whole occupations.

In this study, through Artificial Intelligence Generated Content (AIGC) for the local ethnic minority Pumi people's dress patterns of China's local culture as a carrier, we mainly extract the dress patterns characterized by the Lonicera pattern, combining AIGC technology and the Lonicera pattern to further generate the IP image with national characteristics. Using the stable diffusion model, based on the dataset of a complete Lonicera pattern, the LoRA (Low Rank Adaptation) model is trained for the Pumi dress pattern, and an IP series of image designs based on the Chinese Pumi ethnic group is innovatively designed. This ultimately achieves the purpose of integrating traditional culture and technology, publicizing the local characteristics of the dresses, and passing down the excellent traditional dress patterns of this ethnic group. This research provides reference value for the integration of the Pumi ethnic minority and promotes the digital transformation and innovative development of national culture.

Unstructured software requirements text faces challenges of low processing efficiency and poor consistency and needs to be transformed into a structured form. This paper presents an automation framework based on a large language model that integrates text preprocessing, fine prompt engineering, and requirements association network construction techniques. It precisely parses the semantics of requirements in order to automatically convert them into structured form and analyzes the impact of changes, providing strong support for requirements engineering. Experimental results show that the method performs exceptionally well in terms of requirements structuring accuracy, with an F1 score of 90.8%, and in terms of change tracking efficiency, with an average time consumption of only 3.5 seconds, both significantly better than traditional methods. This study enhanced the automation and intelligence of requirements engineering.

Amid the in-depth integration of artificial intelligence technology into the field of education, natural language processing and knowledge graph technologies have provided new possibilities for the digital transformation of classical text teaching. Faced with the widespread dilemma of in-depth interpretation in classical text teaching, this study proposes a hybrid intelligent auxiliary system featuring the collaboration of "human teacher-AI agent-learner". By integrating deep semantic understanding models with dynamic knowledge graphs, the system constructs a three-layer architecture supporting dynamic classroom inquiry, and realizes key technologies such as domain-adaptive pre-training for literature, semi-automated knowledge graph construction, and teaching-oriented interactive recommendation. A six-week quasi-experiment shows that the system can effectively improve students' depth of interpretation and ability of meaning connection, while enhancing teachers' teaching guidance efficiency. This research provides an operable and verifiable hybrid intelligent practice path for the cultivation of higher-order humanistic literacy in the era of artificial intelligence.

Ascochyta blight of chickpea is a major constraint in semi-arid rainfed systems, and epidemics are strongly driven by rainfall-mediated splash dispersal, yet most weather-based models still rely on an empirical 2 mm daily rainfall threshold and lack experimental quantification of the rainfall–lesion relationship and associated thresholds. To address this gap, a calibrated open-channel rainfall simulator was used in a greenhouse to impose six event rainfall levels from 0 to 8 mm per 10 minutes on potted chickpea plants supplied with residue inoculated by Ascochyta rabiei, generating 96 pot-level lesion counts. A negative binomial GLMM combined with segmented regression was then fitted to derive the rainfall–lesion response function and to estimate a breakpoint at an event rainfall of about 1.1 mm. Lesion numbers increased slowly with rainfall at low amounts but rose much more steeply once the threshold was exceeded. After normalisation, this function was incorporated into the ascotraceR model, and simulations for 4 regions and 28 site–years showed an increase in mean AUDPC R2R^2R2 from 0.65 to 0.81 and a reduction in root mean square error from 210 to 145 disease·day, with clear improvement in capturing the cumulative effects of repeated light rain events. The study provides quantitative threshold evidence for rain-splash dispersal of A. rabiei and illustrates a practical pathway in which physical experiment parameterisation is used to update computational epidemiological models and refine rainfall triggers and fungicide timing rules.

This study presents a structure-aware large language model for accurate prediction of RNA secondary structures directly from primary sequence. The model extends a transformer backbone with a dual-stream architecture that jointly encodes nucleotides and dot-bracket tokens, a grammar-constrained decoder that enforces context-free validity, and a graph-informed attention module that injects candidate base-pair priors during inference. Training is performed on 1.36×10⁶ sequence-structure pairs spanning 203±11 RNA families with lengths from 40 to 1,024 nucleotides. Evaluation on non-overlapping family-level splits shows that the proposed model achieves base-pair F1 scores of 0.884±0.012 on medium-length and 0.837±0.017 on long RNAs, outperforming thermodynamic and neural baselines by 0.041-0.079 absolute F1 while keeping bracket validity above 0.998±0.001. Ablation experiments indicate that grammar constraints account for 63.2±7.4% of the gain in structural validity, whereas graph fusion explains 58.9±6.1% of the improvement on long-range motifs. Despite richer structural bias, decoding throughput reaches 1.27×10³±84 sequences per second on commodity GPUs.

Adolescents exposed to family dysfunction, violence, and school bullying often show trauma-related symptoms and disengagement from class, and current digital interventions rely on static templates that do not handle trauma safety and individual differences well. This study designs an adaptive recommendation framework for generative art therapy that integrates content generation and tagging, a constrained contextual bandit policy, and behavioral and scale-based evaluation, and manages image, music, and narrative micro-tasks with structured labels of goal type, difficulty, and arousal. A school-based trial compares the adaptive policy with a static rule-based sequence. Results indicate higher task completion, deeper sessions, and higher next-day login in the adaptive group, together with larger improvements in emotion-regulation difficulty, self-efficacy, and trauma symptom scores and no increase in the proportion of “trigger/strong discomfort” sessions. The findings suggest that trauma-informed online learning policies can enhance engagement and psychological benefit in adolescent digital art therapy without an observable rise in safety incidents.