Search Results (3,189)

Task-oriented dialogue systems face a tension between comprehensive constraint elicitation (task adequacy) and conversational efficiency (minimizing turns). Current preference learning frameworks treat preferences as static, unable to capture the dynamic evolution of interaction states that evolve across dialogue progression. We present Dual-DPO, a framework that embeds multi-objective preferences into data construction via turn-aware scoring. Our approach decouples objective balancing from policy updates through offline preference scalarization, addressing the optimization instability challenges in online multi-objective reinforcement learning. Experiments on MultiWOZ 2.4 demonstrate 28–35% dialogue turn reduction while maintaining Joint Goal Accuracy > 89% ($p<0.001$). Pareto frontier analysis shows 94% coverage with hypervolume $HV=0.847$. Independent expert evaluation by 10 PhD-level researchers ($n=300$ assessments, inter-rater agreement $\alpha =0.78$) confirms 32% user satisfaction improvement ($p<0.001$). Theoretical analysis demonstrates that offline scalarization, which correlates with improved optimization stability, achieves $3.2\times$ lower gradient variance than online multi-reward optimization by eliminating sampling stochasticity. Our approach enables balanced treatment of competing objectives through Pareto-optimal trade-offs. These results highlight a symmetric and balanced treatment of competing objectives within a Pareto-optimal optimization framework. Full article

Unconsolidated sandstone reservoirs, particularly in offshore and marine environments, are highly susceptible to sand production, which leads to flow-capacity degradation, plugging evolution, sand-retention instability, and erosion–corrosion damage in gravel-pack completion systems. To address the lack of system-level and quantitative integrity evaluation methods, a unified assessment framework is developed by coupling flow behavior, sand-retention mechanisms, and erosion–corrosion damage processes. The gravel-pack completion system is idealized as a concentric multilayer porous-medium structure under steady-state radial Darcy flow, and an equivalent radial permeability model is established to characterize flow capacity and anti-plugging performance, which enables consistent comparison of different completion schemes under identical plugging conditions. Based on sand-retention mechanisms, a sand-retention capacity index is proposed by integrating formation particle size distribution, screen aperture, gravel size, and sand-leakage risk. An erosion–corrosion coupled damage model is further developed to predict screen damage rates in CO₂-containing environments, and an integrity index is formulated to link damage evolution with long-term service performance. By integrating flow capacity, anti-plugging performance, sand-retention capacity, and structural integrity using a weighted geometric mean, a comprehensive evaluation index is established for overall system integrity assessment. Using the proposed framework, a representative formation sand with d₁₀ = 30  μm, d₅₀ = 180  μm, and d₉₀ = 500 μm  is evaluated. The optimal sand control design corresponds to a gravel median size of 971.53 μm (equivalent to a standard 16/20 mesh gravel) and an optimal screen aperture of 523.11 μm, with a screen porosity of 0.56. Under these conditions, the selected screen aperture and gravel size are well matched with the formation sand size, falling within recommended engineering ranges and achieving a favorable balance among sand retention, flow capacity, anti-plugging performance, and structural integrity. The proposed framework provides a quantitative and engineering-applicable basis for the optimization and integrity classification of offshore gravel-pack sand control completions under multi-constraint operating conditions. Full article

This paper presents a multi-layer architecture designed to enhance the reliable autonomous flight of single and multiple quadcopters in simulation. The architecture leverages concepts inspired by the resilient spacecraft executive to hierarchically organize trajectory planning and flight control and integrates an extended Simplex framework that employs multiple candidate algorithms to provide safety assurance at each layer, with a supervisory program that adapts Simplex behavior based on system states and environmental conditions to enable high-level mission management. The approach is evaluated in bridge-inspection simulations under environmental uncertainties, including varying wind conditions and obstacles. Across multiple operating configurations and Monte Carlo simulation runs, the architecture achieves high coverage rates; notably, under high-wind conditions, it reduces average trajectory deviation by 66.2%. The results demonstrate proactive safety through graceful degradation in both trajectory planning and flight control under stress and off-nominal conditions. Full article

Temporal action detection (TAD) aims to localize and recognize action instances in untrimmed videos, and serves as a key component in practical intelligent electronic systems such as smart video surveillance and real-time human–machine interaction. In these scenarios, accurate temporal localization is essential for reliable event understanding and downstream decision-making in edge computing and real-time streaming scenarios. To handle long video durations and diverse action dynamics, existing methods typically rely on hierarchical temporal feature integration to capture multi-scale contextual information. However, such integration often leads to intra-segment inconsistency and boundary ambiguity, as indiscriminate temporal smoothing across scales degrades segment coherence and blurs critical boundary cues. In this work, we propose FreqAct, a multi-frequency feature fusion framework that explicitly models complementary low-frequency and high-frequency temporal components within hierarchical representations. Specifically, low-frequency modulation suppresses undesired temporal fluctuations to stabilize segment-level representations, while high-frequency enhancement preserves boundary-sensitive cues essential for precise localization. Furthermore, we introduce a boundary-aware regression loss to emphasize learning at action boundaries and an intra-segment consistency regularization to encourage coherent predictions within each action instance. Extensive experiments on THUMOS14 and ActivityNet1.3 demonstrate that FreqAct outperforms state-of-the-art TAD methods, further highlighting its effectiveness and practical potential for real-world electronics applications. Full article

Lymphatic filariasis (LF) is a mosquito-borne neglected tropical disease targeted by the World Health Organization (WHO) for global elimination as a public health problem. Sixteen Pacific Island countries and territories were historically endemic, and eight have now met the WHO criteria for elimination as a public health problem. Elimination as a public health problem does not imply zero transmission. Rather, it denotes that LF prevalence has been reduced below a defined threshold at which community transmission can be sustained. Following validation of elimination, the WHO recommends post-validation surveillance (PVS) to detect potential re-emergence of LF as a public health problem. However, implementing PVS is challenging in Small Island Developing States with dispersed populations, limited workforce capacity, resource constraints, and competing health priorities. The ‘Voices and Visions: Building Partnerships for Integrated Serosurveillance of LF and Other Infectious Diseases in the Pacific Islands’ meeting was held in Brisbane, Australia, from 8–10 July 2025. Fifty-one delegates, including Pacific LF programme managers, WHO representatives, global health partners, and academic researchers, reviewed regional PVS progress, discussed the newly released WHO guidelines for the implementation, monitoring, and evaluation of PVS, planned for PVS implementation, and explored novel multiplex bead assay (MBA) serological analysis methods to strengthen regional coordination for its development as a public health tool. Five broad themes emerged. First, the new WHO Monitoring and Epidemiological Assessment of Mass Drug Administration in the Global Programme to Eliminate Lymphatic Filariasis: A Manual for National Elimination Programmes, 2nd edn needs to be operationalised to meet decision-making needs across diverse Pacific settings. Second, integrating LF-PVS with existing surveys and health service activities could improve efficiency and long-term sustainability. Third, regional coordination and alignment of funding cycles will require high-level collaboration. Fourth, community engagement is essential to strengthen demand for PVS. Finally, while at an early stage and with further evidence needed, MBA laboratory methods hold promise for cost-effective, feasible integrated multi-pathogen serosurveillance. Full article

It is well established that early diagnosis and subsequent intervention can result in significant benefits in infants with neurodevelopmental disorders such as cerebral palsy (CP). This scoping review aimed to assess the current state of the literature regarding the use of innovative and emerging technologies for early CP screening, diagnosis and phenotyping in pre-ambulatory children. Searches were performed across PubMed, Embase and Cochrane databases; articles were screened by four independent reviewers at the title/abstract and full-text levels. Forty-eight studies met the inclusion criteria. The most frequently used modalities included wearable sensors (e.g., accelerometers, inertial measurement units) and video-based motion analysis. These movement-tracking systems were used to screen for a variety of pediatric-onset neurodevelopmental disorders and have been useful in quantifying spontaneous infant movements, detecting the absence or abnormality of fidgety movement, or identifying atypical motor patterns. Although CP was our primary focus, several studies applied a similar pipeline to autism spectrum disorder (ASD) and spinal muscular atrophy (SMA), underscoring broader relevance for early neurodevelopmental screening, diagnosing and phenotyping. Overall, technology-assisted motor assessment demonstrated promising feasibility and diagnostic potential; however, most studies are limited by small sample sizes, short follow-up durations, and heterogeneous validation methods. Given the benefits of early intervention and the emerging capabilities of wearable and video-based analytics, larger multi-site and longitudinal datasets are needed to support early diagnosis, risk stratification, and functional phenotyping in CP. Full article

Context: Code recommendation plays an important role in improving programming efficiency and software quality. Existing approaches mainly focus on method- or API-level recommendations, which limits their effectiveness to local code contexts. From a multi-stage recommendation perspective, class-level code recommendation aims to efficiently narrow a large candidate code space while preserving essential structural information. Objective: This paper proposes RioCC, a class-level code recommendation framework that leverages deep forest-based code clone detection to progressively reduce the candidate space and improve recommendation efficiency in large-scale code spaces. Method: RioCC models the recommendation process as a coarse-to-fine candidate reduction procedure. In the coarse-grained stage, a quick search-based filtering module performs rapid candidate screening and initial similarity estimation, effectively pruning irrelevant candidates and narrowing the search space. In the fine-grained stage, a deep forest-based analysis with cascade learning and multi-grained scanning captures context- and structure-aware representations of class-level code fragments, enabling accurate similarity assessment and recommendation. This two-stage design explicitly separates coarse candidate filtering from detailed semantic matching to balance efficiency and accuracy. Results: Experiments on a large-scale dataset containing 192,000 clone pairs from BigCloneBench and a collected code pool show that RioCC consistently outperforms state-of-the-art methods, including CCLearner, Oreo, and RSharer, across four types of code clones, while significantly accelerating the recommendation process with comparable detection accuracy. Conclusions: By explicitly formulating class-level code recommendation as a staged retrieval and refinement problem, RioCC provides an efficient and scalable solution for large-scale code recommendation and demonstrates the practical value of integrating lightweight filtering with deep forest-based learning. Full article

General-purpose large language models excel at open-domain question answering, but in railway operation and maintenance (O&M) scenarios they still suffer from hallucinated knowledge and poor domain adaptation. In practice, railway O&M knowledge mainly arises from two heterogeneous sources: spatio-temporal data such as train trajectories, which are organized along the spatial layout of railway lines, and domain documents such as operating rules, which exhibit varying degrees of structural regularity. Traditional retrieval-augmented generation (RAG) systems usually flatten these multi-source data into a single unstructured text space and perform global retrieval in one embedding space, which easily introduces noisy context and makes it difficult to precisely target knowledge for specific lines, sections, or equipment states. To overcome these limitations, we propose CARAG, a context-aware RAG framework tailored to railway O&M data. CARAG treats domain documents and spatial data as a unified knowledge substrate and builds a spatial knowledge graph with concept and instance levels. On top of this knowledge graph, a GraphReAct-based multi-turn interaction mechanism guides the LLM to reason and act over the concept knowledge graph, dynamically navigating to spatially and semantically relevant candidate regions, within which vector retrieval and instance-level graph retrieval are performed. Experiments show that CARAG significantly outperforms baseline RAG methods on RAGAS metrics, confirming the effectiveness of structure-guided multi-step reasoning for question answering over multi-source heterogeneous railway O&M data. Full article

Accurate reservoir water level forecasting is essential for water resource management, flood risk mitigation, and hydropower operation. However, it remains challenging due to pronounced geographical heterogeneity and complex multi-scale temporal dynamics. Existing deep-learning approaches typically overlook explicit geographical and climatic conditioning. They struggle to capture temporal dependencies across multiple time scales. They also exhibit limited transferability across reservoirs with similar hydrological characteristics. To address these limitations, this paper proposes GeoFormer, a geography-aware adaptive Transformer framework designed for reservoir water level forecasting across diverse geographical contexts. GeoFormer integrates three key innovations. First, a Geography-Aware Embedding Module conditions temporal representations on geographical location, climate regimes, and reservoir attributes. Second, an Adaptive Multi-Scale Temporal Fusion mechanism dynamically aggregates information across daily, weekly, and monthly temporal resolutions. Third, a Cross-Reservoir Knowledge Transfer strategy enables effective knowledge sharing among hydrologically similar reservoirs. Extensive experiments on six reservoirs distributed across multiple continents and climate zones demonstrate that GeoFormer consistently outperforms state-of-the-art baselines, including iTransformer, DLinear, and Informer. The model achieves average reductions of 23.7% in RMSE, 19.4% in MAE, and 15.8% in MAPE, while maintaining strong robustness and generalization across geographically heterogeneous hydrological systems. Full article

Multi-stack Polymer electrolyte Membrane Fuel Cell (PEMFC) systems are increasingly adopted in heavy-duty mobility to overcome the power limitations and thermal instability of single-stack configurations. However, the overall energy efficiency, hydrogen utilization, and thermal behavior of multi-stack fuel cell trucks are highly dependent on the applied Power Management System (PMS). In this study, high-fidelity, system-level dynamic model of multi-stack fuel cell truck was developed using Matlab/Simscape^TM, and three PMS approaches (rule-based control, state-machine control, and fuzzy logic control) were comparatively evaluated. The analysis includes coolant temperature regulation, hydrogen consumption, battery State of Charge (SoC) dynamics, and the parasitic power demand of Balance of Plant (BoP) components. Results show that the fuzzy logic PMS provides the most balanced operating profile by smoothing transient fuel cell loading and actively leveraging the battery during high-demand periods. In the thermal domain, the fuzzy logic PMS reduced temperature overshoot by up to 61.20%, demonstrating the most stable thermal control among the three strategies. Hydrogen consumption decreased by 3.08% and 0.89% compared with the rule-based and state-machine PMS, respectively, while parasitic power consumption decreased by 7.12% and 3.32%, confirming improvements in overall energy efficiency. TOPSIS-based multi-criteria decision analysis further showed that the fuzzy logic PMS achieved the highest closeness coefficient (0.9112), indicating superior system-level performance. These findings highlight the importance of PMS design for achieving energy-optimal and thermally stable operation of multi-stack PEMFC trucks and provide practical guidance for future control strategies, heavy-duty mobility applications, and next-generation hydrogen powertrain optimization. Full article

Dysarthria in patients post-stroke is often accompanied by central facial paralysis, which impairs facial motor control and emotional expression. Current assessments rely on acoustic modalities, overlooking facial pathological cues and their correlation with emotional expression, which hinders comprehensive disease assessment. To address this issue, we propose a multimodal severity classification framework that integrates facial and acoustic features. Firstly, a multi-level annotation algorithm based on a pre-trained model and motion amplitude was designed to overcome the problem of data scarcity. Secondly, facial topology was modeled using Delaunay triangulation, with spatial relationships captured via graph convolutional networks (GCNs), while abnormal muscle coordination is quantified using facial action units (AUs). Finally, we proposed a multimodal feature set fusion technology framework to achieve the compensation of facial visual features for acoustic modalities and the analysis of disease classification. Our experimental results using the THE-POSSD dataset demonstrate an accuracy of 92.0% and an F1 score of 91.6%, significantly outperforming single-modality baselines. This study reveals the changes in facial movements and sensitive areas of patients under different emotional states, verifies the compensatory ability of visual patterns for auditory patterns, and demonstrates the potential of this multimodal framework for objective assessment and future clinical applications in speech disorders. Full article

Understanding the differences in the infiltration processes of soda saline–alkali soils with varying degrees of salinization and their underlying mechanisms is of great significance for the rational use of regional soil and water resources. This study was conducted in the Songnen Plain, one of the world’s three major saline–alkali soil distribution areas, where the salt composition is dominated by sodium bicarbonate and sodium carbonate. Five types of soda saline–alkali soils with different degradation levels were selected from the study area. Using a one-dimensional vertical constant-head single-ring infiltration method, characteristic parameters of the infiltration process were measured through in situ experiments. Based on principal component analysis (PCA), a comprehensive multi-parameter infiltration capacity index (SICI) was constructed. Pathway analysis was further employed to explore the potential relationship between soil physical and saline–alkali characteristics and the infiltration process. The results showed that compared to the initial infiltration rate, the steady-state infiltration rates of the five soils decreased significantly by 41.81%, 64.87%, 97.20%, 99.24%, and 99.59%, respectively. Notably, the steady-state infiltration rate of the most severely degraded saline–alkali soil (Suaeda glauca) was only 0.13 mm·h⁻¹. Correspondingly, Suaeda glauca soil exhibited the lowest SICI. Correlation and pathway analyses indicated that SICI was significantly associated with physical and saline–alkali parameters of the soda saline–alkali soils. Besides the direct associations of the fractal dimension of particle size distribution (D), non-capillary porosity (NCP), and salt content (SC) on SICI, D was also linked to lower SICI indirectly through its relationship with NCP, sodium adsorption ratio (SAR), and SC. The findings suggest that soil physical structure, particularly the fractal dimension of particle size distribution and pore characteristics, appears to be a primary factor influencing the infiltration capacity of highly soda saline–alkali soils, and that improving soil texture structure and enhancing soil porosity could be prioritized in the restoration and management of severely degraded soda saline–alkali lands. Full article

To address the inherent mismatch between the fluctuating power output of renewable energy and the continuous production requirements of ammonia in off-grid wind–solar–hydrogen–ammonia systems, this paper proposes a “day-ahead–intraday–real-time” multi-time-scale coordinated optimization scheduling strategy. In the day-ahead layer, Wasserstein Distributionally Robust Optimization (WDRO) is employed to determine a conservative and stable baseline plan for ammonia load under high uncertainty of wind and solar output. The intraday layer utilizes Model Predictive Control (MPC) with a 2-h prediction horizon and 15-min rolling steps to correct short-term forecast deviations. The real-time layer achieves minute-level power balancing through priority dispatch and deadband control. Furthermore, hydrogen storage tanks serve as a material buffer between hydrogen production and ammonia synthesis, with their state variables transmitting across layers to achieve flexible multi-time-scale coupling. Simulation results demonstrate that, although this strategy slightly reduces the theoretical maximum ammonia yield, it completely avoids load-shedding risks. Compared with the deterministic scheduling (Scheme 1), which suffers a net loss due to severe penalty costs, the proposed strategy achieves a positive daily profit of CNY 277,700, representing an absolute increase of CNY 429,300. Furthermore, it provides an additional daily profit of CNY 65,800 compared to the stochastic optimization approach (Scheme 2), demonstrating superior economic robustness in off-grid environments. Full article

Under extreme fault conditions or during maintenance restarts, DC collection wind farms may experience a total blackout due to protective isolation. Addressing the black-start challenges arising from the unidirectional power flow structure and weak damping characteristics inherent to DC step-up collection wind farms, this paper proposes a sequential black-start control scheme predicated on grid-source coordination. A representative topology and an equivalent black-start model of the DC collection system are established to analyze the start-up mechanism and to design an active voltage build-up strategy with virtual impedance for the grid-side Modular Multilevel Converter (MMC). Meanwhile, generator-side permanent-magnet direct-drive wind turbines exploit their self-excitation capability and optimized pitch control to realize islanded self-bootstrapping and stable rotational speed. In addition, we develop a two-stage soft cut-in strategy that combines open-loop voltage scanning for pre-synchronization with closed-loop constant-current ramping of DC/DC converters, together with control logic for sequentially connecting multiple units to the DC grid. Simulation results show that the proposed approach smoothly restores the system from a zero-energy state to the rated operating point without external power sources, confirming the feasibility of full-farm start-up using the grid-side converter station and unit self-bootstrapping. Full article

Extracting structured threat intelligence from unstructured cybersecurity texts requires accurate identification of entities together with their underlying semantic relations. However, threat reports often exhibit intricate sentence structures, long-range contextual dependencies, and tightly coupled entity–relation patterns, which pose substantial challenges for existing extraction approaches. To address these challenges, this study investigates joint entity–relation extraction from the perspective of semantic dependency modeling and develops HDIM-JER, a unified framework that captures structured interactions among heterogeneous linguistic features. HDIM-JER integrates character-level cues, contextual representations, and higher-order semantic dependency evidence to enhance structural awareness during joint inference, where different second-order dependency configurations provide an interpretable perspective on structurally symmetric and hierarchically asymmetric interaction patterns among entity–relation instances. By incorporating multi-level dependency interactions, HDIM-JER effectively alleviates error propagation associated with pipeline-based architectures and improves the modeling of complex relational dependencies. Extensive experiments on a threat intelligence corpus and a public benchmark dataset demonstrate consistent performance improvements over representative state-of-the-art methods in both entity recognition and relation extraction, confirming the effectiveness of higher-order semantic dependency interaction modeling for threat intelligence analysis. Full article