This study presents a comprehensive empirical evaluation of six state-of-the-art large language models (LLMs) for code generation, including both general-purpose and code-specialized models. Using a dataset of 944 real-world LeetCode problems across five programming languages, we assess model performance using rigorous metrics: compile-time errors, runtime errors, functional failures, and algorithmic suboptimalities. The results reveal significant performance variations, with DeepSeek-R1 and GPT-4.1 consistently outperform others in terms of correctness, efficiency, and robustness. Through detailed case studies, we identify common failure scenarios such as syntax errors, logical flaws, and suboptimal algorithms, highlighting the critical role of prompt engineering and human oversight in improving results. Based on these findings, we provide actionable recommendations for developers and practitioners, emphasizing that successful LLM deployment depends on careful model selection, effective prompt design, and context-aware usage to ensure reliable code generation in real-world software development tasks.
Current large language model (LLM) serving systems, primarily designed for text completion, are neither efficient nor adaptable for increasingly complex LLM applications due to their inflexible design. We propose a new LLM serving system architecture that serves programs instead of prompts to address this problem. These programs, called LLM Inference Programs (LIPs), allow users to customize token prediction and KV cache management at runtime and to offload parts of their application logic, such as tool execution, to the server. We describe an example of this architecture through a system named Symphony, which functions as an operating system for LIPs. Symphony exposes LLM model computations via system calls and virtualizes KV cache with a dedicated file system, while ensuring GPU efficiency with a two-level process scheduling scheme. Symphony has the potential to open the door to a more efficient and extensible ecosystem for LLM applications.
Constructing a compressed latent space through a variational autoencoder (VAE) is the key for efficient 3D diffusion models. This paper introduces COD-VAE that encodes 3D shapes into a COmpact set of 1D latent vectors without sacrificing quality. COD-VAE introduces a two-stage autoencoder scheme to improve compression and decoding efficiency. First, our encoder block progressively compresses point clouds into compact latent vectors via intermediate point patches. Second, our triplane-based decoder reconstructs dense triplanes from latent vectors instead of directly decoding neural fields, significantly reducing computational overhead of neural fields decoding. Finally, we propose uncertainty-guided token pruning, which allocates resources adaptively by skipping computations in simpler regions and improves the decoder efficiency. Experimental results demonstrate that COD-VAE achieves 16x compression compared to the baseline while maintaining quality. This enables 20.8x speedup in generation, highlighting that a large number of latent vectors is not a prerequisite for high-quality reconstruction and generation. The code is available at https://github.com/join16/COD-VAE.
Samuel Maddock, Shripad Gade, Graham Cormode
et al.
State-of-the-art differentially private synthetic tabular data has been defined by adaptive 'select-measure-generate' frameworks, exemplified by methods like AIM. These approaches iteratively measure low-order noisy marginals and fit graphical models to produce synthetic data, enabling systematic optimisation of data quality under privacy constraints. Graphical models, however, are inefficient for high-dimensional data because they require substantial memory and must be retrained from scratch whenever the graph structure changes, leading to significant computational overhead. Recent methods, like GEM, overcome these limitations by using generator neural networks for improved scalability. However, empirical comparisons have mostly focused on small datasets, limiting real-world applicability. In this work, we introduce GEM+, which integrates AIM's adaptive measurement framework with GEM's scalable generator network. Our experiments show that GEM+ outperforms AIM in both utility and scalability, delivering state-of-the-art results while efficiently handling datasets with over a hundred columns, where AIM fails due to memory and computational overheads.
Generative Adversarial Networks (GAN) have greatly influenced the development of computer vision and artificial intelligence in the past decade and also connected art and machine intelligence together. This book begins with a detailed introduction to the fundamental principles and historical development of GANs, contrasting them with traditional generative models and elucidating the core adversarial mechanisms through illustrative Python examples. The text systematically addresses the mathematical and theoretical underpinnings including probability theory, statistics, and game theory providing a solid framework for understanding the objectives, loss functions, and optimisation challenges inherent to GAN training. Subsequent chapters review classic variants such as Conditional GANs, DCGANs, InfoGAN, and LAPGAN before progressing to advanced training methodologies like Wasserstein GANs, GANs with gradient penalty, least squares GANs, and spectral normalisation techniques. The book further examines architectural enhancements and task-specific adaptations in generators and discriminators, showcasing practical implementations in high resolution image generation, artistic style transfer, video synthesis, text to image generation and other multimedia applications. The concluding sections offer insights into emerging research trends, including self-attention mechanisms, transformer-based generative models, and a comparative analysis with diffusion models, thus charting promising directions for future developments in both academic and applied settings.
This paper presents an artistic and technical investigation into the attention mechanisms of video diffusion transformers. Inspired by early video artists who manipulated analog video signals to create new visual aesthetics, this study proposes a method for extracting and visualizing cross-attention maps in generative video models. Built on the open-source Wan model, our tool provides an interpretable window into the temporal and spatial behavior of attention in text-to-video generation. Through exploratory probes and an artistic case study, we examine the potential of attention maps as both analytical tools and raw artistic material. This work contributes to the growing field of Explainable AI for the Arts (XAIxArts), inviting artists to reclaim the inner workings of AI as a creative medium.
In this study, we have analysed the determinants of consumer insurance premium expenditures in India using regional variation in insurance market characteristics, as well as demographic and economic influences. This finding, that of the significant geographic differences in spending behaviors, suggests important implications for the effects of local economic conditions and possibly cultural attitudes on insurance choices, which is supported by the findings of rough statistical significance from ANOVA and regression results. So descriptive statistics give us an average spending of ₹40,000 with a decent dispersion, and the distribution is also well behaved around this mean, revealing that different consumers spend differently. Likewise, age and income significantly influence spending on life insurance premiums, as older individuals and those with higher income spend more on such products both now and six months from now. The results of this research provide valuable indicators for insurance providers, underlining the need to adapt offerings and marketing approaches according to the regional or demographic groups they are intended for. It increases the consumer behavior knowledge within the insurance industry and lays a solid foundation for strategic decisions and further research efforts.
The purpose of this study is to examine the effects on digital competencies (digital literacy, digital self-efficacy, digital mindset), creativity, and self-expression after conducting digital storytelling activities utilizing metaverse avatar customization in a mandatory liberal arts course for international students. To conduct this study, a case from a mandatory liberal arts course based on artistic creation activities at C University was analyzed. To achieve the research objectives, a paired sample t-test was conducted using SPSS 25.0 to verify the pre-post differences, and a correlation analysis was performed to examine the relationships between variables. The study results are as follows: First, the analysis of pre-post differences revealed that all variables showed significantly higher values in the post-test. Second, the correlation analysis between variables showed a positive (+) correlation among digital literacy, digital efficacy, digital mindset, creativity, and self-expression, while self-expression and digital mindset did not show a significant correlation. Third, the analysis of learner perceptions revealed that most learners experienced an improvement in their digital skills through this course. They perceived it as a positive experience that allowed them to discover and develop their creative potential. Additionally, they responded that the process of customizing characters from their stories as metaverse avatars and completing videos using digital tools provided them with a sense of accomplishment. Based on the results of this study, it is hoped that this will serve as a valuable resource for instructors and learners in digital storytelling activities using the metaverse, particularly targeting foreign learners.
Shogher Poghosyan, Meri Manucharyan, Gevorg Martirosyan
et al.
This article aims to identify the possibilities of combining existing approaches in ecology with the challenges of the labor market in the current era of digitalization and to identify what qualitative changes the labor market may undergo under the impact of digital knowledge acquisition. This article contains original ideas as it proposes a methodology framework that combines the long-known population prediction model in ecology (the Leslie matrix) with empiric economic methods for labor studies. The proposed framework, which includes a two-stage empirical approach and employs the Leslie matrix, can help determine how digital knowledge can affect the quality of structural changes occurring in labor. The proposed approach is useful for policymakers for various purposes, such as predicting structural changes in a labor market, assessing the effectiveness of state support programs, and for the private sector, big corporations, and state administration bodies to predict the possibilities of changes in job structure resulting from digital knowledge acquisition.
The present study concerns the analysis of the reality and functionality of an identity that is based on diasporic space. The analysis tends to elaborate on the location of hybrid identity in a diasporic space, with a reliance on a comparative reading of Homi Bhabha’s concept of hybridity and Jean Baudrillard’s concept of simulacrum in the particular case of Mehrnoush Mazarei’s short story, "Sangam". From a formal perspective, "Sangam" is a compilation of chronology and autobiography; the content denotes the narration of a one-year span of the life of the narrator, a migrant Iranian woman, her memories of acquaintance and friendship with an Indian woman in a minor diasporic society of United States. Employing what results from the dialogue between Bhabha’s hybridity and Jean Baudrillard’s simulacrum, and with an applied reference to the form of the narrative and also the narrator’s characterization, the analysis undertaken here illumines how the identity of the narrator, even if harvested and developed as hybrid, is a simulation; a simulation that she herself builds via gazing at the other, cultural fascination and mimicry; activities which in certain ways enables her to absorb the Indian woman’s identity. Furthermore, in the course of these actions and as their consequences, there emerge different layers of repetition and imitation which reduplicate and reinforce the differing nature and the layers of simulation, in a way that all the possible signification of hybridity can disintegrate.
Erik M. Fredericks, Denton Bobeldyk, Jared M. Moore
Generative art is a rules-driven approach to creating artistic outputs in various mediums. For example, a fluid simulation can govern the flow of colored pixels across a digital display or a rectangle placement algorithm can yield a Mondrian-style painting. Previously, we investigated how genetic improvement, a sub-field of genetic programming, can automatically create and optimize generative art drawing programs. One challenge of applying genetic improvement to generative art is defining fitness functions and their interaction in a many-objective evolutionary algorithm such as Lexicase selection. Here, we assess the impact of each fitness function in terms of the their individual effects on generated images, characteristics of generated programs, and impact of bloat on this specific domain. Furthermore, we have added an additional fitness function that uses a classifier for mimicking a human's assessment as to whether an output is considered as "art." This classifier is trained on a dataset of input images resembling the glitch art aesthetic that we aim to create. Our experimental results show that with few fitness functions, individual generative techniques sweep across populations. Moreover, we found that compositions tended to be driven by one technique with our current fitness functions. Lastly, we show that our classifier is best suited for filtering out noisy images, ideally leading towards more outputs relevant to user preference.
This visual narrative is a first-person reflection of the previous pictorial at the 1st International Workshop on Explainable AI for the Arts (XAIxArts) at ACM Creativity and Cognition 2023. The initial workshop pictorial explored a relationship between researcher and artificial intelligence, navigating creative challenges throughout the 2023 teaching block. It concluded by raising crucial questions regarding attribution transparency, the ethical dimensions of the creative process, and the delicate balance between inspiration and plagiarism. Subsequent discussions at the workshop yielded valuable insights, particularly concerning interpreting the creative journey. This follow-up visual narrative reflects the enduring impact of Makayla Lewis's interaction with AI. A self-portrait that delves into the interplay of creativity and introspection.
Although automatically animating audio-driven talking heads has recently received growing interest, previous efforts have mainly concentrated on achieving lip synchronization with the audio, neglecting two crucial elements for generating expressive videos: emotion style and art style. In this paper, we present an innovative audio-driven talking face generation method called Style2Talker. It involves two stylized stages, namely Style-E and Style-A, which integrate text-controlled emotion style and picture-controlled art style into the final output. In order to prepare the scarce emotional text descriptions corresponding to the videos, we propose a labor-free paradigm that employs large-scale pretrained models to automatically annotate emotional text labels for existing audiovisual datasets. Incorporating the synthetic emotion texts, the Style-E stage utilizes a large-scale CLIP model to extract emotion representations, which are combined with the audio, serving as the condition for an efficient latent diffusion model designed to produce emotional motion coefficients of a 3DMM model. Moving on to the Style-A stage, we develop a coefficient-driven motion generator and an art-specific style path embedded in the well-known StyleGAN. This allows us to synthesize high-resolution artistically stylized talking head videos using the generated emotional motion coefficients and an art style source picture. Moreover, to better preserve image details and avoid artifacts, we provide StyleGAN with the multi-scale content features extracted from the identity image and refine its intermediate feature maps by the designed content encoder and refinement network, respectively. Extensive experimental results demonstrate our method outperforms existing state-of-the-art methods in terms of audio-lip synchronization and performance of both emotion style and art style.
This study aims to analyze the development and operation cases of applying the team teaching method to ‘Mind Immune Coaching’, a liberal arts course at a university, and to suggest improvement measures in the future. The research method was to conduct a survey of 85 students who took the ‘Mind Immune Coaching’ course in the first semester of the 2024 academic year to objectively understand the learners' experiences and perceptions. The research results showed that the team teaching method provides students with various perspectives and interdisciplinary perspectives and promotes creative and convergent problem-solving activities. Based on these results, three major improvement measures were derived. First, research on learner awareness and satisfaction was strengthened as an important factor for realizing learner-centered education. Second, it is necessary to strengthen the interaction between instructors and learners in a distance education environment and develop an efficient question-and-answer method. Third, it is necessary to study and develop an efficient operation plan to improve the educational quality of distance education through the team teaching method. Accordingly, this study provides an empirical analysis of the effectiveness and challenges of team teaching methods, explores possibilities for realizing learner-centered education in a university distance education environment, and suggests implications for developing customized teaching methods based on edutech. In addition, it is expected to contribute to the development of various liberal arts subjects by suggesting a new model of convergence liberal arts education as a direction for innovation in university liberal arts education.
Devi Fauziyah Ma’rifat, Suyami Suyami, Ninawati Syahrul
et al.
The construction of Harmonious Life is a value which has been practiced by the Malay community since long ago. This context can be viewed and found in the scroll text of Sang Nila Utama Museum’s collection in Riau Province. This study aims to explain and reflect on the construction of harmonization values that exist within the Riau Malay community. To explore harmonious life construction values in the manuscripts of the Riau Malay community, this study uses a qualitative descriptive approach. The result of this study shows that the construction of harmonious life in the Malaya Riau community manuscripts refers to three contexts; First, transcendental values foster compassion for the Creator. The second consists of values that are transactional in the context of blessed commerce. The third refers to relational values to maintain human relations in interacting and communicating. It can be concluded from these three contexts that the construction of harmonious life in the manuscripts of the Riau Malay community does not only emphasize transcendental-based values but also accommodates communal values.
Computational creativity has contributed heavily to abstract art in modern era, allowing artists to create high quality, abstract two dimension (2D) arts with a high level of controllability and expressibility. However, even with computational approaches that have promising result in making concrete 3D art, computationally addressing abstract 3D art with high-quality and controllability remains an open question. To fill this gap, we propose to explore computational creativity in making abstract 3D art by bridging evolution strategies (ES) and 3D rendering through customizable parameterization of scenes. We demonstrate that our approach is capable of placing semi-transparent triangles in 3D scenes that, when viewed from specified angles, render into films that look like artists' specification expressed in natural language. This provides a new way for the artist to easily express creativity ideas for abstract 3D art. The supplementary material, which contains code, animation for all figures, and more examples, is here: https://es3dart.github.io/
إن فعل الطلب هو احد أهم الأفعال الكلامية في اللغة وان استخدامه بشكل الدقيق من قبل المتعلمين للغة الانجليزية إنما يدل على اكتسابهم القدر اللازم من المعرفة. إلا أن عدد كبير منهم لا يعكسون ذلك فيما يخص استخدام فعل الطلب. إذا جاءت هذه الدراسة لتحديد ما هي الأنماط المختلفة للطلب التي يستخدمها طلبة الجامعة, بينما تبرز أوجه الصعوبات في طرح واستيعاب هذا الفعل الكلامي من قبلهم, وتقييم استخدامهم له, و فيما إذا كان هناك أي تأثير للغة الأم_ أي العربية _على ذلك. وتفترض الدراسة لأجل ذلك كله إن الطلبة العراقيين كمتعلمين أجانب للغة الانجليزية, لا يمتلكون القدر اللازم من المعرفة لاستيعاب و طرح فعل الطلب في اللغة الانجليزية هذا أولا. ثانيا إن مستوى أدائهم أفضل في استيعاب فعل الطلب من ما هو الحال في طرحه. و تفترض الدراسة أيضا إن هؤلاء الطلبة ومهما اختلف الموقف يكررون استخدام ذات الأنماط للتعبير عن الطلب, لإلمامهم بعدد محدد من أنماطه.
ولبرهنت مدى مصداقية هذه الفرضيات من عدمها تبنيت الدراسة الإجراءات التالية: (1) عرض الإطار النظري لنظرية فعل الكلام و مبادئ التأدب اللغوي ومحاور أخرى ذات صلة بفعل الطلب. (2) إدارة اختبار ذو شقين لقياس مستوى طلبه جامعه الكوفة/ كليه الآداب/ قسم اللغة الانجليزية في استيعاب وطرح فعل الطلب. (3) إجراء عدد من العمليات الإحصائية التي تبرهن صحة الفرضيات أعلاه أو تدحضها. وفي الختام تم التوصل إلى عدد من الاستنتاجات والتوصيات العلمية.
History of scholarship and learning. The humanities, Arts in general
Matthias Springstein, Stefanie Schneider, Christian Althaus
et al.
Gesture as language of non-verbal communication has been theoretically established since the 17th century. However, its relevance for the visual arts has been expressed only sporadically. This may be primarily due to the sheer overwhelming amount of data that traditionally had to be processed by hand. With the steady progress of digitization, though, a growing number of historical artifacts have been indexed and made available to the public, creating a need for automatic retrieval of art-historical motifs with similar body constellations or poses. Since the domain of art differs significantly from existing real-world data sets for human pose estimation due to its style variance, this presents new challenges. In this paper, we propose a novel approach to estimate human poses in art-historical images. In contrast to previous work that attempts to bridge the domain gap with pre-trained models or through style transfer, we suggest semi-supervised learning for both object and keypoint detection. Furthermore, we introduce a novel domain-specific art data set that includes both bounding box and keypoint annotations of human figures. Our approach achieves significantly better results than methods that use pre-trained models or style transfer.