Search Results (46)

Large language models (LLMs) are playing an increasingly important role in data science applications. In this study, the performance of LLMs in generating code and designing solutions for data science tasks is systematically evaluated based on different real-world tasks from the Kaggle platform. Models from different LLM families were tested under both default settings and configurations with hyperparameter tuning (HPT) applied. In addition, the effects of few-shot prompting (FSP) and Tree of Thought (ToT) strategies on code generation were compared. Alongside technical metrics such as accuracy, F1 score, Root Mean Squared Error (RMSE), execution time, and peak memory consumption, LLM outputs were also evaluated against Kaggle user-submitted solutions, leaderboard scores, and two established AutoML frameworks (auto-sklearn and AutoGluon). The findings suggest that, with effective prompting strategies and HPT, models can deliver competitive results on certain tasks. The ability of some LLMS to suggest appropriate algorithms reveals that LLMs can be seen not only as code generators, but also as systems capable of designing machine learning (ML) solutions. This study presents a comprehensive analysis of how strategic decisions such as prompting methods, tuning approaches, and algorithm selection, affect the design of LLM-based data science systems, offering insights for future hybrid human–LLM systems. Full article

Accurate estimation of soil properties is crucial for optimizing agricultural practices and promoting sustainable resource management. Hyperspectral imaging provides a non-invasive means of quantifying key soil parameters, but effectively utilizing the high-dimensional hyperspectral data presents significant challenges. In this paper, we introduce HyperSoilNet, a hybrid deep learning framework for estimating soil properties from hyperspectral imagery. HyperSoilNet leverages a pretrained hyperspectral-native CNN backbone and integrates it with a carefully optimized machine learning (ML) ensemble to combine the strengths of deep representation learning with traditional ML techniques. We evaluate our framework on the Hyperview challenge dataset, focusing on four critical soil properties: potassium oxide, phosphorus pentoxide, magnesium, and soil pH. Comprehensive experiments demonstrate that HyperSoilNet surpasses state-of-the-art models, achieving a score of 0.762 on the challenge leaderboard. Through detailed ablation studies and spectral analysis, we provide insights on the components of the framework, and their contribution to performance, showcasing its potential for advancing precision agriculture and sustainable soil management practices. Full article

In response to challenges around student engagement and teacher technological proficiency, this paper looks at the impact of gamification on students’ mathematical resilience whilst monitoring their mathematical anxiety plus investigating teachers’ experiences, willingness, and professional development ambitions to utilise gamified instructional tools in the mathematics classroom. Drawing on strategies to motivate students, the aim of this paper is to unbundle gamification in enacting curriculum change and the role of teacher professional development in using the pedagogical approach in mathematics in Ireland. Ireland is currently experiencing second-level curriculum reforms that are placing particular emphasis on digital competence and technological fluency from both teachers and students. With teachers highlighting the gap in educators’ pedagogical skills for the smooth roll out of recent curriculum reform due to the lack of knowledge and competency in technological teaching strategies, this study is both relevant and timely. Games have been used in multiple industries aiming to motivate participants and increase engagement on a particular matter. However, the term “gamification” has been coined by Pelling as the use of games in a non-gaming context. Current students are very technologically savvy due to the exposure of software applications from a young age and the integration of technological appliances in all walks of life. Traditional teaching and learning strategies are potentially seen as monotonous and somewhat boring to today’s students. Utilising game-based design such as leaderboards, points, and badges encourages motivation and enhances engagement of students. With this in mind, and the rate of change in mathematics curricula globally in recent years, there is a significant emphasis on the necessity of professional development initiatives to adapt at the same rate. Full article

Citizen science now relies heavily on digital platforms to engage the public in environmental data collection. Yet, many projects face declining participation over time. This study examines the effect of three elements of gamification—points, daily streaks, and real-time leaderboards—on student engagement, achievement, and immersion during a five-day campus-wide intervention utilising the GAME and a spatial crowdsourcing app. Employing a convergent mixed-methods design, we combined behavioural log analysis, validated psychometric scales (GAMEFULQUEST), and post-experiment interviews to triangulate both quantitative and qualitative dimensions of engagement. Results reveal that gamified elements enhanced students’ sense of accomplishment and early-stage motivation, which is reflected in significantly higher average scores for goal-directed engagement and recurring qualitative themes related to competence and recognition. However, deeper immersion and sustained “flow” were less robust with repetitive task design. While the intervention achieved only moderate long-term participation rates, it demonstrates that thoughtfully implemented game mechanics can meaningfully enhance engagement without undermining data quality. These findings provide actionable guidance for designing more adaptive, motivating, and inclusive citizen science solutions, underscoring the importance of mixed-methods evaluation in understanding complex engagement processes. While the sample size limits the statistical generalizability, this study serves as an exploratory field trial offering valuable design insights and methodological guidance for future large-scale, controlled citizen science interventions. Full article

In traditional programming education, teachers typically design fixed questions and standard answers, manually grading the solutions submitted by students. This process not only requires significant time and effort from educators but may also fail to provide timely and personalized feedback due to limited teaching resources. To alleviate these burdens and enhance teaching efficiency, this study leverages generative artificial intelligence (AI) technology to develop a system capable of automatically generating questions and grading answers. Students engage in programming exercises through a gamified approach, with the system providing instant feedback on their answers. Additionally, student performance is displayed via leaderboards, incorporating peer competition to boost learning motivation. According to a user survey, the gamified system demonstrates significant advantages: 56.67% of students found the system easy to use; 40% considered the system well-integrated; 60% indicated that they quickly mastered the system’s functionality, and over half (53.33%) believed that the leaderboard effectively enhanced their competitive awareness and motivation. These results suggest that the system not only reduces teachers’ workload but also increases student engagement and learning outcomes through gamified design. Full article

As technology advances, interest in video games is extending to broader audiences. This makes gamification as a mechanism for improving educational outcomes increasingly attractive. This article reports on a study in which a 3D third-person video game was used to develop proficiency in spatial reasoning abilities relating to symmetry. The video game, called Triskelion, interleaves elements of traditional gameplay with educational elements. Gameplay includes non-violent shooter elements, parkour, searching, and exploring. Educational elements include points, leaderboards, a theme, clear goals, feedback, and a group challenge in the form of a clan-based match. This composition of elements makes Triskelion unique in the genre of academic educational games. Our study compares one Triskelion match to a longer educational sequence consisting of a “practice test”, followed by engagement with a 3D digital experience called the Kessel Run. Analysis of the results using the Mann–Whitney U test revealed (p = 0.9297) that both pedagogical pathways yielded similar proficiency results. This suggests that Triskelion might create a learning environment more aligned with the characteristics of deliberate practice. Full article

As virtual reality (VR) sports training apps start to become more mainstream, it is important that human performance is measured from VR gameplay interaction data in a more meaningful way. CleanSheet is a VR training app that is played by over 100,000 users around the world. Many of those players are aspiring goalkeepers who want to use the app as a new way to train and improve their general goalkeeping performance. Whilst the leaderboards display how many shots players saved, these data do not take into account the difficulty of the shot faced. This study presents a regression model developed from a combination of existing expected goals (xG) models, goalkeeper performance metrics, and psychological research to produce a new shot difficulty metric called CSxG. Utilizing user save rate data as the target variable, a model was developed that incorporated three input variables relating to ball flight and in-goal positioning. Our analysis showed that the required rate of closure (RROC), adapted from Tau theory, was the most significant predictor of the proportion of goals conceded. A validation process evaluated the new xG model for CleanSheet by comparing its difficulty predictions against user performance data across players of varying skill levels. CSxG effectively predicted shot difficulty at the extremes but showed less accuracy for mid-range scores (0.4 to 0.8). Additional variables influencing shot difficulty, such as build-up play and goalpost size, were identified for future model enhancements. This research contributes to the advancement of predictive modeling in sports performance analysis, highlighting the potential for improved goalkeeper training and strategy development using VR technology. Full article

Background/Objectives: Antimicrobial resistance (AMR) is a global public health concern exacerbated by inappropriate antimicrobial prescribing practices, particularly in low-resource settings such as Uganda. The research aimed to develop a culturally sensitive behavioral intervention, leveraging a “nudge” strategy, to improve healthcare provider adherence to the 2016 Uganda Clinical Guidelines (UCG 2016) in five Ugandan hospitals. This intervention formed part of broader antimicrobial stewardship initiatives led by the United States Agency for International Development Medicines, Technologies, and Pharmaceutical Services Program. Methods: This study employed a mixed-methods approach, combining formative research and behavioral intervention. Guided by the Deloitte Behavioral Insights Framework, the research team conducted key informant interviews to identify prescribing barriers and motivators and developed three suitable behavioral interventions: perceived monitoring, ward leaderboards, and educational workshops. The study evaluated the interventions’ impact through point prevalence surveys (PPS), using the World Health Organization PPS methodology at three stages: pre-intervention, immediate post-intervention, and one-month post-intervention. Results: Key behavioral themes across individual, social, environmental, and organizational elements informed the intervention design and implementation. The behavioral intervention package increased antimicrobial prescription compliance with the UCG 2016 from 27% at baseline to 50% immediately post-intervention, though these effects diminished at one-month post-intervention. Conclusions: Our study addresses an existing gap in behavioral nudges-based operational research on antimicrobial prescribing in low- and middle-income countries. These results showed an immediate improvement in adherence to the UCG 2016 among healthcare providers in Ugandan hospitals, though the effect was attenuated at one-month follow-up. Despite the attenuation, behavior change presents a feasible, cost-effective, and sustainable approach to improving antimicrobial prescribing practices and addressing AMR. Full article

Advergames represent a novel product placement strategy that surpasses traditional advertising methods by fostering interaction between brands and their target audiences. This study investigates the unique engagement opportunities provided by video games, focusing mainly on the ‘flow experience’, an intensified state of immersion frequently encountered by players of computer games. Such immersive experiences have the potential to significantly influence a player’s perception, offering a new avenue for advertisements to impact and engage audiences effectively. The primary objective of this research was to examine the influence of advergames on players who are deeply immersed in the gaming experience, with a specific focus on the subsequent effects on brand recognition over time. The study involved 44 software developers, who were evenly divided into two groups for the experiment. Both groups were exposed to an identical gaming environment with the task of locating a designated product within the game. However, one group interacted with an enhanced version of the game, which included additional stimuli—such as dynamic music, an engaging narrative, time constraints, a competitive leaderboard, and immersive voice acting—to intensify the gaming experience. The experiment strategically placed various products within the game, and their detectability was assessed using eye-tracking technology. Following gameplay, participants completed questionnaires that measured their experience with flow state and brand recall. The data were analyzed using the Mann–Whitney U test and correlation analysis to facilitate comparisons. The findings indicated that the product associated with the primary task achieved the highest recall rate between both groups. Furthermore, eye-tracking technology identified the areas in the game that attracted the most attention, revealing a preference for mid- and high-level placements over lower-level ones. Full article

Skin diseases and lesions can be ambiguous to recognize due to the similarity of lesions and enhanced imaging features. In this study, we compared three cutting-edge deep learning frameworks for dermoscopic segmentation: U-Net, SegAN, and MultiResUNet. We used a dermoscopic dataset including detailed lesion annotations with segmentation masks to help train and evaluate models on the precise localization of melanomas. SegAN is a special type of Generative Adversarial Network (GAN) that introduces a new architecture by adding generator and discriminator steps. U-Net has become a common strategy in segmentation to encode and decode image features for limited data. MultiResUNet is a U-Net-based architecture that overcomes the insufficient data problem in medical imaging by extracting contextual details. We trained the three frameworks on colored images after preprocessing. We added incremental Gaussian noise to measure the robustness of segmentation performance. We evaluated the frameworks using the following parameters: accuracy, sensitivity, specificity, Dice and Jaccard coefficients. Our accuracy results show that SegAN (92%) and MultiResUNet (92%) both outperform U-Net (86%), which is a well-known segmentation framework for skin lesion analysis. MultiResUNet sensitivity (96%) outperforms the methods in the challenge leaderboard. These results suggest that SegAN and MultiResUNet are more resistant techniques against noise in dermoscopic segmentation. Full article

(1) Background: In developed countries, public health faces a number of problems, including sedentary lifestyles and poor diets, which collectively contribute to the occurrence of preventable diseases. Noncommunicable diseases represent the leading cause of global mortality. Despite the promotion of healthy living, compliance remains a significant challenge. The integration of gamification into health apps has been demonstrated to facilitate behavioral change. Blockchain technology enhances the effectiveness of gamification by providing data trustability and support for auditable incentives. This feature is possible and easy due to the inherent characteristics of blockchain automating processes through Smart Contracts, rewarding participants and creating leaderboards in a transparent and reliable manner. The use of smart contracts and events enhances the traceability and reliability of decentralized applications, including healthcare. Interoperability in blockchain tools facilitates the deployment of complex environments. The aim of this research is the deployment of a tool for the implementation and testing of a gamification platform based on blockchain technology. (2) Methods: Pre-experimental research was carried out to assess the usability of the decentralized application developed. (3) Results: A decentralized application was developed with the objective of gamifying healthy habits. The application was evaluated using the System Usability Scale, obtaining a score of 80.49, and the Cronbach’s Alpha score, which was found to be 0.75. (4) Conclusions: A prototype of a decentralized application connected with a blockchain network to reward challenge fulfilment was deployed. Despite being in early development, it demonstrated high usability. Employing blockchain technology guarantees transparency and traceability while remaining in compliance with legal requirements like the General Data Protection Regulation. Full article

Fine-grained representation is fundamental to species classification based on deep learning, and in this context, cross-modal contrastive learning is an effective method. The diversity of species coupled with the inherent contextual ambiguity of natural language poses a primary challenge in the cross-modal representation alignment of conservation area image data. Integrating cross-modal retrieval tasks with generation tasks contributes to cross-modal representation alignment based on contextual understanding. However, during the contrastive learning process, apart from learning the differences in the data itself, a pair of encoders inevitably learns the differences caused by encoder fluctuations. The latter leads to convergence shortcuts, resulting in poor representation quality and an inaccurate reflection of the similarity relationships between samples in the original dataset within the shared space of features. To achieve fine-grained cross-modal representation alignment, we first propose a residual attention network to enhance consistency during momentum updates in cross-modal encoders. Building upon this, we propose momentum encoding from a multi-task perspective as a bridge for cross-modal information, effectively improving cross-modal mutual information, representation quality, and optimizing the distribution of feature points within the cross-modal shared semantic space. By acquiring momentum encoding queues for cross-modal semantic understanding through multi-tasking, we align ambiguous natural language representations around the invariant image features of factual information, alleviating contextual ambiguity and enhancing model robustness. Experimental validation shows that our proposed multi-task perspective of cross-modal momentum encoders outperforms similar models on standardized image classification tasks and image–text cross-modal retrieval tasks on public datasets by up to 8% on the leaderboard, demonstrating the effectiveness of the proposed method. Qualitative experiments on our self-built conservation area image–text paired dataset show that our proposed method accurately performs cross-modal retrieval and generation tasks among 8142 species, proving its effectiveness on fine-grained cross-modal image–text conservation area image datasets. Full article

This study investigates the strategic decision-making abilities of large language models (LLMs) via the game of Tic-Tac-Toe, renowned for its straightforward rules and definitive outcomes. We developed a mobile application coupled with web services, facilitating gameplay among leading LLMs, including Jurassic-2 Ultra by AI21, Claude 2.1 by Anthropic, Gemini-Pro by Google, GPT-3.5-Turbo and GPT-4 by OpenAI, Llama2-70B by Meta, and Mistral Large by Mistral, to assess their rule comprehension and strategic thinking. Using a consistent prompt structure in 10 sessions for each LLM pair, we systematically collected data on wins, draws, and invalid moves across 980 games, employing two distinct prompt types to vary the presentation of the game’s status. Our findings reveal significant performance variations among the LLMs. Notably, GPT-4, GPT-3.5-Turbo, and Llama2 secured the most wins with the list prompt, while GPT-4, Gemini-Pro, and Mistral Large excelled using the illustration prompt. GPT-4 emerged as the top performer, achieving victory with the minimum number of moves and the fewest errors for both prompt types. This research introduces a novel methodology for assessing LLM capabilities using a game that can illuminate their strategic thinking abilities. Beyond enhancing our comprehension of LLM performance, this study lays the groundwork for future exploration into their utility in complex decision-making scenarios, offering directions for further inquiry and the exploration of LLM limits within game-based frameworks. Full article

Autonomous driving navigation relies on diverse approaches, each with advantages and limitations depending on various factors. For HD maps, modular systems excel, while end-to-end methods dominate mapless scenarios. However, few leverage the strengths of both. This paper innovates by proposing a hybrid architecture that seamlessly integrates modular perception and control modules with data-driven path planning. This innovative design leverages the strengths of both approaches, enabling a clear understanding and debugging of individual components while simultaneously harnessing the learning power of end-to-end approaches. Our proposed architecture achieved first and second place in the 2023 CARLA Autonomous Driving Challenge’s SENSORS and MAP tracks, respectively. These results demonstrate the architecture’s effectiveness in both map-based and mapless navigation. We achieved a driving score of $41.56$ and the highest route completion of $86.03$ in the MAP track of the CARLA Challenge leaderboard 1, and driving scores of $35.36$ and $1.23$ in the CARLA Challenge SENSOR track with route completions of $85.01$ and $9.55$, for, respectively, leaderboard 1 and 2. The results of leaderboard 2 raised the hybrid architecture to the first position, winning the edition of the 2023 CARLA Autonomous Driving Competition. Full article

Virtual reality (VR) is increasingly employed in various domains, notably enhancing learning and experiences in cultural heritage (CH). This study examines the effects of gamified and non-gamified VR experiences within virtual museum environments, highlighting the concept of a digital twin and its focus on cultural heritage. It explores how these VR modalities affect visitor motivation, engagement, and learning outcomes. For this purpose, two versions were developed: a gamified virtual reality version incorporating interactive gaming elements like achievements, profiles, leaderboards, and quizzes and a non-gamified virtual reality version devoid of these elements. This study, using an experimental design with 76 participants (38 in each group for the gamified and non-gamified experiences), leverages the Wieng Yong House Museum’s digital twin and its fabric collection to assess the educational and experiential quality of virtual museum visits. The findings indicate that while gamification significantly boosts the reward dimension of visitor engagement, its influence is most pronounced in the effort dimension of motivation; however, its impact on learning outcomes is less marked. These insights are instrumental for integrating VR and gamification into museum environments. Full article