Jeonju University

Current multimodal artificial intelligence suffers from fragmentation, with models typically optimized for single tasks, impeding efficient and uniform handling of diverse tasks like Text-to-Image (T2I), Image-to-Text (I2T), and Visual Question Answering (VQA) within a single framework. To address this, we propose the Synergistic Multimodal Diffusion Transformer (SyMDit), a novel unified discrete diffusion model. SyMDit integrates an Adaptive Cross-Modal Transformer (ACMT) with a Synergistic Attention Module (SAM) for dynamic interaction, alongside Hierarchical Semantic Visual Tokenization (HSVT) for multi-scale visual understanding and Context-Aware Text Embedding with special tokens for nuanced textual representation. Trained under a unified discrete diffusion paradigm, SyMDit employs a multi-stage strategy, including advanced data augmentation and selective masking. Our extensive evaluations demonstrate that SyMDit consistently achieves superior performance across T2I, I2T, and VQA tasks, outperforming existing baselines. Furthermore, SyMDit significantly enhances inference efficiency, offering substantial speedups compared to autoregressive and prior discrete diffusion methods. This work presents a significant step towards truly unified and efficient multimodal AI, offering a robust framework for general-purpose multimodal intelligence.

Skin aging is caused by extrinsic and intrinsic factors and can lead to various aesthetic concerns and psychosocial stress. Exosome-based treatments have recently gained attention in the field of skin rejuvenation. This study aimed to retrospectively analyze the clinical efficacy of an exosome (CUREDOC EXOSOME REPAIR Advanced Solution Skin Booster, CUREDOC, Sejong, Korea) combined with microneedling treatment. We analyzed the medical records of 40 patients who received exosome skin booster (50 mcg/mL) and microneedling combination therapy at a single medical institution from January to March 2024. Physicians performed the treatment on each patient every 2weeks for a total of 4 sessions. The physicians followed each patient for 8weeks after the final treatment. The physicians evaluated each patient by taking clinical photographs, having each patient complete a satisfaction survey, assessing each patient using the Global Aesthetic Improvement Scale (GAIS), and documenting adverse events. The patient group had a mean age of 41.2±6.3years. The results of the study showed a high patient acceptance rate, with an overall satisfaction of 87.5%, an 82% intention to undergo repeat treatment, and an 85% willingness to recommend the treatment. According to the GAIS assessments, 85% of patients scored "improved" or greater. Adverse events were transient and mild, and there were no serious adverse reactions reported. Based on the results of this study, the combination of exosome skin booster and microneedling treatment can be considered a safe and effective option for improving facial aging.

The self-efficacy of individuals with intellectual disabilities is considered an important factor in the psychological adjustment process. The General Self-Efficacy Scale (GSES) is commonly used to measure self-efficacy. However, previous studies have not examined the psychometric properties of the GSES among individuals with intellectual disabilities. Therefore, this study investigated the psychometric properties of the GSES using the Rasch model based on the item response theory. This study used secondary data from the Employment Panel Survey of Persons with Disabilities provided by the Republic of Korea Employment Agency for Persons with Disabilities. The panel survey collected data from individuals with intellectual disabilities in the Republic of Korea using the GSES. This study analyzed data from 232 individuals to determine GSES item fitness, item difficulty, rating scale fit, and reliability. The results revealed that eight of ten GSES items exhibited appropriate fit. Item difficulty required modification, indicating the need for items with lower difficulty. The four-point Likert scale used for responses was appropriate. Person and item separation indices demonstrated good scale reliability. These findings suggest that the GSES is effective for measuring self-efficacy in people with intellectual disabilities; however, some adjustments, such as changes in difficulty level, are required.

This study explores the impact of female directors on the investment efficiency of Korean firms, especially under conditions of cost stickiness. Using data from firms listed on the Korean Stock Exchange between 2014 and 2021, regression is employed to determine the relationship between female board representation and investment efficiency. The results show that female directors do not significantly affect investment efficiency under typical economic conditions, but their influence becomes more pronounced during periods of cost stickiness. In these circumstances, the presence of female directors is associated with enhanced investment efficiency, suggesting that they promote more disciplined financial strategies under economic stress. These findings advance the understanding of gender diversity in corporate governance, highlighting the conditions under which female directors are most effective. The study also offers valuable insights for policymakers and investors, underscoring the strategic importance of gender diversity on corporate boards in enhancing firm performance during challenging economic times.

This literature review examines the state of Text-to-SQL technology, which translates natural language queries into SQL. It analyzes rule-based, neural, and hybrid approaches, assessing their strengths and weaknesses, and surveys commonly used datasets, benchmarks, and evaluation metrics. The study identifies research gaps concerning generalization, scalability, and interpretability, and suggests integrating user feedback and domain knowledge. To better understand the implementation and potential improvements of machine learning in this domain, we conducted a systematic literature review (SLR) of publications from 2015 to 2023. From 439 gathered papers, 23 were identified as highly relevant. The review analyzes these works across four areas: (i) datasets employed, (ii) evolution of learning methods, (iii) development of evaluation procedures, and (iv) a meta-analysis of model performance. The findings confirm significant room for improvement in learning strategies. Persistent research gaps include cross-domain generalization, schema linking for complex databases, a lack of robust multilingual models, and the trade-off between model accuracy and interpretability. We propose future directions such as integrating contrastive schema linking, zero-shot/few-shot learning, explainability-driven design, and developing diverse, large-scale benchmarks that reflect real-world database complexity.

Personalized nutrition aims to optimize health by addressing interindividual differences in metabolism, microbiota composition, and dietary responses. Modulating the gut microbiota through probiotics, prebiotics, and synbiotics is promising, yet conventional systems such as capsules or fermented foods offer limited control over dosage, release kinetics, and microbial viability. These formats often cause 2–4 log reductions in viable counts during processing and gastrointestinal transit, underscoring the need for advanced delivery technologies. Three-dimensional (3D) food printing enables digital design of edible matrices with programmable geometry and composition to enhance microbial protection and controlled release. Coaxial and gel-in-gel architectures have retained over 90–96% of probiotic cells after printing and 80–85% after simulated digestion. Synbiotic formulations combining probiotics with fructooligosaccharides or whey protein achieve 98–99% survival and stability for 35 days. This review summarizes advances in formulation, encapsulation, and printing strategies, highlighting how 3D food printing uniquely overcomes challenges of viability, release control, and personalized dosage in microbiota-based nutrition.