Graphic visual art is undergoing a profound transformation driven by generative artificial intelligence (AI). This transformation extends beyond technological innovation, reshaping the paradigms of artistic image creation. This paper examines human-centered interactive systems for artistic image creation, aiming to bridge the gap between technology-oriented and art-oriented research. We identify three core challenges in interactive generative art creation: controllability, integrity, and creativity, and analyze in depth the underlying trade-offs among them. Building on this analysis, we systematically review the evolution of generative models, approaches to integrating artistic knowledge, and the emergence of collaborative agentic AI systems. Looking ahead, we emphasize the need to develop human-AI collaborative frameworks that offer greater controllability, preserve human creative agency, and ensure regulatory compliance, while exploring a wide range of potential applications. This study proposes a unified analytical framework that connects AI technologies, human-AI interaction processes, and the integration and innovation of artistic image creation. It further outlines a task-oriented roadmap to guide future research and the development of collaborative systems for artistic image creation.

Technology managers are instrumental in facilitating technology transfer. While large language models (LLMs) have demonstrated potential in this domain, evaluating them on the core cognitive capabilities expected of technology managers remains challenging due to the lack of domain-specific benchmarks. To address this challenge, we construct TMBench, a domain-specific benchmark for evaluating the cognitive capabilities expected of technology managers. It comprises a collection of 3474 questions spanning diverse scenarios, covering three tasks and five question types. We integrate typical metrics with LLM-as-a-Judge using both fixed and entropy-based weights to calculate the corresponding dimensional and composite scores. Leveraging this mechanism, we evaluate nine representative models under varying prompt configurations and temperature settings to investigate their capabilities, characterize their error patterns, and benchmark them against human experts. Our results show heterogeneous capabilities, with no single model ranking first across all three capability levels. For higher-order capabilities, LLMs show more favorable point estimates than the human baseline, but the differences are not statistically significant. Chain-of-thought-based prompting is associated with prompt-sensitive capability profiles, with more favorable open-weight performance in the two higher-order tasks. Error analysis identifies knowledge base, retrieval, and application deficits as the most prevalent error category. Building on these findings, this study provides a reference for model selection, prompt engineering, and targeted optimization when applying LLMs in technology transfer.

Area and power need to be considered at all stages of integrated circuit design. Area and power cooperative optimization for mixed polarity Reed‒Muller (MPRM) circuits aims to find the Pareto non-dominated sets with minimum area and power. Existing multiobjective optimization approaches for MPRM circuits have a long runtime and cannot handle large-scale circuits. In this paper, we propose a multi-strategy quantum wolf pack algorithm (MQWPA), which includes quantum encoding, dual-population strategy, dynamic reverse learning strategy, mutation strategy, and termination criterion. Moreover, based on the proposed MQWPA, we propose an efficient area and power optimization approach (EAPOA), which searches for XNOR/OR-based MPRM circuits with Pareto non-dominated solution sets. Experimental results show that compared with state-of-the-art evolutionary algorithms, MQWPA can search for optimal solutions in multiobjective test problems; compared with commonly used Boolean minimization methods, EAPOA has advantages in area and power consumption. After field-programmable gate array technology and standard cell mapping, by selecting area as the optimal cost implementation, we obtain average improvements in the actual area of 16.33% and 2.27%, respectively.

Virtual staining converts label-free autofluorescence (AF) images into hematoxylin and eosin (H&E)-like brightfield contrast for nondestructive pathology. We adapt the unpaired neural Schrödinger bridge (UNSB) framework to AF-to-H&E virtual staining and develop an enhanced UNSB (E-UNSB) model for pathology-oriented image translation. E-UNSB retains the unpaired, unidirectional, and progressive stochastic transport formulation of UNSB while introducing pathology-specific structural regularization to preserve cellular morphology and tissue architecture. In particular, saliency constraints and global feature consistency provide multi-scale structural guidance to reduce distortion during staining generation. Experiments demonstrate improved distributional similarity and morphometric consistency over the evaluated unsupervised baselines. Quantitative downstream analysis further indicates that virtual H&E images retain morphometric features relevant to pathological assessment, supporting their potential for high-throughput multimodal tissue analysis.

Industrial anomaly detection is important for intelligent manufacturing since defects can degrade product reliability. Existing methods based on knowledge distillation typically use a single teacher, limiting feature diversity and robustness. Multi-teacher methods often require explicit feature alignment across different network architectures, which may introduce semantic conflicts and weaken anomaly responses. We propose implicit multi-teacher distillation (IMTD) for industrial anomaly detection. IMTD employs three complementary teacher networks, namely ResNet18, ResNet34, and Wide-ResNet50, to capture texture, structural, and semantic representations from different feature spaces. Each teacher is paired with an independent decoder trained only on normal samples, enabling implicit knowledge transfer without cross-architecture feature alignment. A false-negative-aware maximum response aggregation (FN-MRA) module is further introduced to preserve the strongest anomaly evidence during fusion. Experiments on MVTec anomaly detection (MVTec AD), visual anomaly benchmark (VisA), and metal part defect dataset (MPDD) show that IMTD consistently outperforms recent state-of-the-art methods in both image-level and pixel-level anomaly detection. The results demonstrate the effectiveness of combining complementary teachers, implicit reconstruction, and maximum-response aggregation under anomaly-scarce conditions.

This study analyzes the price of anarchy (PoA) for a scheduling game on two hierarchical uniform machines under the parallel processing policy, a coordination mechanism that combines the well-studied ShortestFirst and makespan policies. This scheduling game corresponds to the round-robin scheduling mechanism widely adopted in time-sharing operating systems and network packet processors. This study focuses on makespan minimization. We derive the PoA as a function of the ratio of the speeds of the two machines. We demonstrate that all derived bounds are tight for any ratio, thereby characterizing the inefficiency of Nash equilibria in scheduling games on two uniform machines under parallel processing.

Semi-active suspension systems equipped with continuous damping control (CDC) dampers place stringent demands on accurate damper-level force tracking. However, the inherent hysteresis, compression-rebound asymmetry, and strong current dependence of the damper force make force control persistently challenging. This paper addresses this challenge with a four-component dynamic model that describes elastic restoring force, velocity-dependent viscous damping force, Bouc-Wen hysteresis force, and an additional force induced by directional asymmetry in a unified framework. Each current-dependent parameter is represented by an exponential saturation function, which naturally ensures bounded and monotonic coefficient mappings at the model level while avoiding empirical clipping. Because the four force components dominate in distinct operating regions, a three-stage hierarchical parameter identification method is further proposed to replace the coupled, single-step identification of all 20 parameters by direct global optimization. Based on the identified model, a feedforward inverse controller is constructed that combines online Bouc-Wen state reconstruction, passivity-based clipping, and Newton iteration to achieve real-time current solution and force tracking. Replacing equal-weight global fitting with the proposed saturation-mapped hierarchical scheme substantially reduces the identification error, from 0.155 to 0.079. In simulation, the inverse controller built on this model attains a force-tracking RMSE of 4.9 N, and hardware-in-the-loop experiments further verify the feasibility and effectiveness of the proposed framework under power-amplifier transport delay and other bench-level non-idealities.

With its excellent anti-interference capability, low probability of interception, and code-division multiple access capability, direct sequence spread spectrum plays a crucial role in unmanned aerial vehicle swarm communications, high-precision ranging, and massive connectivity in beyond fifth-generation and sixth-generation networks. However, low spectral efficiency and poor energy concentration of conventional rectangular chip waveforms used in spread spectrum modulation make it difficult to meet the high-rate and high-sensitivity reception requirements of future systems. To address this issue, this study proposes a spread spectrum chip waveform design method based on prolate spheroidal wave functions (PSWFs). By constructing a set of four orthogonal PSWF-based chip waveforms, the proposed method enhances signal dimensionality and spectral efficiency under finite time-bandwidth product constraints. Furthermore, an equivalent low-pass filter structure is integrated to reduce implementation complexity. Theoretical analysis and simulation results demonstrate that the proposed waveform achieves approximately double the spectral efficiency of conventional waveforms while maintaining comparable error performance. It also exhibits superior anti-interference and anti-interception capabilities, providing a viable waveform design approach for next-generation spread spectrum communication systems.