Central Police University

Age-based licensing renewal policies are implemented in aging societies to protect vulnerable road users (VRUs), yet the rationale for selecting age thresholds remains uncertain. This study examined whether age 70 represents a threshold for VRU case fatality among crash-involved older drivers and whether license status modifies this association. This population-based cross-sectional study analyzed Taiwanese police-reported crash data from 2011 to 2024 involving nonprofessional car drivers aged ≥ 50 years and VRUs. Driver age was modeled categorically and continuously using restricted cubic splines (RCSs). License status was examined as an exploratory effect modifier. A prespecified primary linear-spline analysis directly tested for a change in slope at age 70, while supporting RCS analyses evaluated overall nonlinearity. Multivariable logistic regression with cluster-robust standard errors estimated adjusted odds ratios (AORs) and 95% confidence intervals (CIs). Among 602,782 crash-involved VRUs, 2,789 sustained fatal injuries (0.46%). The prespecified analysis did not detect a change in slope at age 70 (change-in-slope AOR per year, 1.001; 95% CI, 0.977–1.025; P = 0.95), and the selected 3-knot RCS specification provided no evidence of overall nonlinearity (P = 0.24). Compared with drivers aged 50–64 years, odds of VRU case fatality were higher among drivers aged 65–69 years (AOR 1.15; 95% CI, 1.03–1.28) and ≥ 70 years (AOR 1.14; 95% CI, 1.01–1.29). However, among drivers aged 65–99 years, age ≥ 70 years was not independently associated with VRU case fatality (AOR 1.08; 95% CI, 0.93–1.25; P = 0.32). Unlicensed driving was associated with higher odds of VRU case fatality (AOR, 1.92; 95% CI, 1.62–2.28). Overall, the prespecified analysis did not detect a change in slope at age 70, while the association with unlicensed driving highlights a potentially important correlate of VRU case fatality among crash-involved older drivers.

We present the SemEval-2026 shared task on Dimensional Aspect-Based Sentiment Analysis (DimABSA), which improves traditional ABSA by modeling sentiment along valence-arousal (VA) dimensions rather than using categorical polarity labels. To extend ABSA beyond consumer reviews to public-issue discourse (e.g., political, energy, and climate issues), we introduce an additional task, Dimensional Stance Analysis (DimStance), which treats stance targets as aspects and reformulates stance detection as regression in the VA space. The task consists of two tracks: Track A (DimABSA) and Track B (DimStance). Track A includes three subtasks: (1) dimensional aspect sentiment regression, (2) dimensional aspect sentiment triplet extraction, and (3) dimensional aspect sentiment quadruplet extraction, while Track B includes only the regression subtask for stance targets. We also introduce a continuous F1 (cF1) metric to jointly evaluate structured extraction and VA regression. The task attracted more than 400 participants, resulting in 112 final submissions and 42 system description papers. We report baseline results, discuss top-performing systems, and analyze key design choices to provide insights into dimensional sentiment analysis at the aspect and stance-target levels. All resources are available on our GitHub repository.

Abstract Evidence-based policing (EBP) has gained global momentum, yet its reception within centralized, non-Western police systems remains largely unexplored. This study examines EBP receptivity among 1,189 officers from three Taiwanese police departments—representing urban, rural, and mixed jurisdictions—replicating the survey framework of Lum et al. (2012) (‘Receptivity to Research in Policing’, Justice Research and Policy, 14: 61–95) and the analytical approach of Telep (2017) (‘Police Officer Receptivity to Research and Evidence-Based Policing: Examining Variability Within and Across Agencies’, Crime & Delinquency, 63: 976–99) using a culturally adapted survey instrument. Findings reveal that Taiwanese officers demonstrate markedly lower research receptivity than Western counterparts: only 30.3 per cent found research useful and valued researcher collaboration, while fewer than 8 per cent believed science should outweigh experience in daily decision-making. Logistic regression analyses identify holding a master's degree and reading professional publications as the strongest predictors of receptivity, with publication reading alone distinguishing the most receptive officers (3.0 per cent of the sample) from all others. Drawing on institutional theory, we argue that Taiwan's centralized bureaucracy generates coercive isomorphism that suppresses individual- and job-level variation in officer attitudes, rendering external knowledge exposure disproportionately critical. These findings challenge assumptions of uniform EBP implementation and underscore the need for system-specific strategies that account for organizational structure.

The rising share of renewable energy has amplified electricity price volatility, underscoring the need for accurate forecasting and robust risk management. This study proposes an integrated machine learning framework that combines advanced forecasting models (Random Forest, XGBoost, LSTM) with feature engineering and probabilistic risk assessment. Value at risk (VaR) and conditional VaR (CVaR) are derived from predictive distributions to guide hedging strategies using futures and options. Empirical tests on 3 years of hourly data show that the approach reduces RMSE by up to 18

This paper presents a reversible data hiding method in encrypted images, combining pixel-wise encryption with block-wise keyed hashing and random map generation. The original image is first encrypted using a pixel-wise XOR operation with a secret key K1. The encrypted image is then divided into non-overlapping blocks, where each block carries secret information by embedding a symbol through seed-driven random maps generated from the concatenation of a second key K2, block indices, and the random seed generated via SHA-256 hashing. The resulting marked encrypted image ensures secure data embedding while preserving reversibility. At the receiver side, the same random maps are regenerated using the corresponding keys, enabling secret extraction and original image reconstruction. To enhance reliability, a block complexity evaluation is applied to candidate results, and the secret corresponding to the lowest complexity score is selected. Experimental results demonstrate that the proposed method has superior performance than the method proposed by Kim et al. [14].