Abstract
ABSTRACT The study aims to understand how, to what extent, and through what mechanisms startups contribute to the achievement of the sustainable development goals (SDGs), distinguishing between impact-driven initiatives and approaches in which sustainability is used instrumentally. To this end, a mixed-methods methodology was applied, combining quantitative and qualitative analysis on a sample of 1864 startups accelerated by Y Combinator. The results indicate that the most addressed SDGs are SDG 13 (Climate Action) and SDG 3 (Good Health and Well-being), with an emphasis on clean technologies, low-carbon models, data-based climate monitoring, personalized and preventive medicine, mental health, and clinical data management. In contrast, goals related to biodiversity and reducing inequalities are less represented, revealing gaps and areas with potential for expansion. The study's contributions include identifying patterns and gaps in startups' engagement with the SDGs, reinforcing the importance of public policies and incentives to balance efforts in less addressed areas, and demonstrating the central role of digital technologies and data analysis in promoting sustainability.

Abstract
Data leakage is a critical issue in deep learning inflating performance and compromise validity, especially in sensitive areas like medical imaging. This study systematically evaluates two common leakage types in oral squamous cell carcinoma classification from histopathology images: (1) preprocessing leakage (global normalization before dataset splitting) and (2) a severe sample-related (patient-related) contamination scenario created by mixing closely related original and augmented images across splits. We trained 11 CNN and Transformer-based models on a public oral cancer histopathology dataset, benchmarking results against published leakage-free baselines. The results obtained show that the configuration with random splitting of original and augmented images (Scenario 2) artificially increased accuracy by up to 18% (mean +14.3%) compared to leakage-free conditions, while the preprocessing-based leakage (Scenario 1) showed smaller deviations (+1.8%). These inflated metrics arise from a combination of cross-split contamination between closely related samples and increased dataset redundancy, rather than genuine gains in generalization ability. Transformers improved leak-free accuracy (+3.9%) but degraded performance in Scenario 2 (-1.4%), revealing sensitivity to sample-specific biases. The observed performance gains under data leakage conditions are methodological artifacts that undermine clinical reliability, with a severe sample-related contamination scenario (Scenario 2) with random splitting of original and augmented images being particularly detrimental due to its promotion of non-generalizable feature learning. The quantitative benchmarks established here-including a mean accuracy gap of 12.5% (Scenario 2 vs. Scenario 1) across 11 models and Transformer architectures’ sensitivity to contamination-reveal fundamental tradeoffs between metric inflation and model trustworthiness. These findings establish quantitative benchmarks for leakage impacts in medical imaging and inform future guidelines for trustworthy AI development in pathology. © The Author(s) 2026.

Abstract
The polarization curve characteristics of proton exchange membrane (PEM) hydrogen electrolyzers lead to large variations in the equivalent load impedance over the operating current range. This results in a varying closed-loop system time response when traditional fixed-gain PI controllers are employed. In this work, the design and experimental validation of a 3-phase interleaved buck converter controlled via a proposed adaptive lead-lag current control strategy for a PEM hydrogen electrolyzer load is presented. The incremental load conductance method is used to obtain a control-oriented model of the converter-electrolyzer system, enabling real-time calculation of controller parameters via pole-zero cancellation and user-specified transient performance. A laboratory prototype is implemented to experimentally verify the approach under step-load changes, ramp-load changes, and 50% input voltage sag conditions. The results show less than 1% current ripple, identical transient performance over the entire operating range, and improved disturbance ride-through performance compared to a traditional PI controller. The proposed approach offers a viable and robust control solution for high-current PEM electrolyzer applications.

Abstract
Short-form destination videos often rely on music to carry cultural meaning. This paper links Cognitive Metaphor Theory with the circumplex dyad of pleasure and arousal to explain how music-image pairings build destination brand resonance (DBR). Three experiments show that pleasure is the stable route to DBR, arousal helps only under favorable tone, and their effects are additive. A Meaning-Access Prime (MAP) raises both emotions under identical clips and, in Bayesian structural models, also exerts a direct path to DBR, strongest when pleasant tone is low. DBR then predicts destination brand identification and destination consumption intention. We also show a useful state view: Resonant versus Emergent DBR. The framework provides design rules for co-tuning tone, activation, and cultural cues in creator-made clips that improve resonance, identification, and intention.

Abstract
The widespread adoption of embedded systems has led to their deployment in critical real-world applications, making them attractive targets for malicious actors. This paper presents RunPBA, a hardware-based runtime attestation system designed to defend against control flow attacks while maintaining minimal performance overhead and adhering to strict power consumption constraints. RunPBA leverages Pointer Authentication and Branch Target Identification (PACBTI), a new processor extension tailored for the ARM Cortex M processor family, allowing robust protection without requiring hardware modifications, a limitation present in similar solutions. We implemented a proof-of-concept and evaluated it using two benchmark suites, Coremark PRO and BEEBS. Experimental results indicate that RunPBA imposes a geometric mean performance overhead of only 1.3% and 6.8% across the benchmarks, underscoring its efficiency and suitability for real-world deployment.

Abstract
This paper presents a case study to illustrate the application of the directed qualitative content analysis (DQCA) technique to focus group transcriptions for data-driven qualitative persona creation, with broader applicability in human-computer interaction and software development. Using a case study from a project focused on creating an e-grocery marketplace for facilitating short agrifood supply chain trade in the Portuguese context, we demonstrate and validate how DQCA can systematically generate personas that reflect real user needs. For the focus group session, we involved one of the project's stakeholders: family farmers. Furthermore, we propose how these personas can be integrated into the Rational Unified Process software development methodology, guiding decision-making, user-centered design, and prioritization throughout all its phases. Despite being rooted in the e-grocery domain, this paper's methodological approach and insights into generating and integrating user-centered personas in software development processes apply to a broader range of industries and projects, offering guidelines for practitioners and researchers in diverse contexts.