Multi-modal hashing aims to succinctly encode heterogeneous modalities into binary hash codes, facilitating efficient multimedia retrieval characterized by low storage demands and high retrieval speed. Despite the commendable achievements of existing methods, they still face three crucial challenges: 1) Inadequate bridging of the heterogeneous modality gap through coarse, global feature-level alignment and fusion. 2) The erosion of bit independence and consequent limitations on the semantic representation capacity of hash codes during feature-level hash code learning. 3) The insufficiency of binary label-based pairwise semantic preservation strategies in capturing intricate fine-grained semantic correlations within multi-modal data. To address these challenges, this paper introduces the Dynamic Bit-wise Semantic Transformer Hashing (DBSTH) framework. Remarkably, it treats each hash bit as a unique semantic concept, facilitating concept-level alignment of heterogeneous modalities. This safeguards bit independence and augments representation capabilities. Specifically, we devise a dynamic unit fusion strategy for the adaptive combination of local multi-modal information units, facilitating the acquisition of bit-wise semantic concepts. Subsequently, we incorporate a transformer encoder to refine these concepts by uncovering latent correlations among distinct concepts. Finally, we perform the multi-modal alignment and fusion on the fine-grained concept-level, independently encoding each concept to its corresponding hash bit. To provide enhanced guidance for concept learning, a label prototype learning mechanism is introduced, which learns prototype embeddings for all categories through the consideration of co-occurrence priors. This mechanism effectively captures fine-grained explicit semantic correlations and generates supervising hash codes. Additionally, to improve the robustness of the hashing model in handling noisy multi-modal data, a masked concept learning strategy is introduced, facilitating the acquisition of resilient semantic concepts. Extensive experiments conducted on three widely tested multi-modal retrieval datasets demonstrate the superiority of our method in conventional, noisy, and open-set retrieval scenarios.

The rapid digital transformation in the Industry 4.0 era has reshaped vocational education, emphasizing the need for graduates who possess strong English communication skills aligned with global business demands. This study investigates students’ perceptions of using Free4Talk, an online language exchange platform, to enhance English for Business Communication among vocational students. Employing a mixed method design, the research collected quantitative data through a five point Likert scale questionnaire and qualitative data through semi-structured interviews with 20 Business Administration students at Kupang State Polytechnic. Descriptive statistics were used to analyze the questionnaire, while thematic analysis was applied to interpret interview responses. The findings revealed that students hold positive perceptions of Free4Talk across four key dimensions: usability and accessibility (M=3.73), motivation and engagement (M=3.59), language and business communication skills (M=3.45), and intercultural communication competence (M=3.43) all categorized as “high.” The results indicate that Free4Talk effectively fosters speaking fluency, confidence, and intercultural awareness through authentic, real time interaction with global users. Students valued its accessibility, user friendly design, and motivational features that encouraged frequent speaking practice beyond the classroom. In conclusion, integrating Free4Talk into vocational English instruction bridges academic and workplace communication needs, supporting the development of globally competent communicators. It is recommended that future research employ experimental or longitudinal approaches to examine measurable gains in fluency, accuracy, and pragmatic competence, as well as to explore the integration of teacher guidance and feedback mechanisms in technology-mediated language learning

The Sustainable Development Goals (SDGs) have emerged as a global agenda to address challenges such as social inequality, environmental degradation, and the quality of education. Higher education plays a strategic role in achieving these goals by fostering students’ critical thinking, academic literacy, and social responsibility. This study employs a Systematic Literature Review (SLR) of scholarly publications from 2025 to examine the integration of SDG values into Indonesian Language courses in higher education. The review identifies key themes, including the role of higher education in SDG implementation, academic literacy development, character formation, social awareness, and strategies for integrating the SDGs into language learning. The findings indicate that Indonesian Language courses possess significant potential as a medium for SDG integration because of their emphasis on academic literacy, critical thinking, and scientific communication. However, the review also reveals that SDG integration remains partial and unsystematic, highlighting the need for a structured conceptual framework. This study recommends the planned incorporation of SDG values into learning objectives, instructional materials, teaching strategies, and assessments, as well as the development of SDG-based learning guides for lecturers. The results provide theoretical and practical implications for enhancing Indonesian Language instruction in alignment with sustainable development.

Purpose This study aims to investigate how a single, well-designed instructional video can influence students’ perceptions, learning outcomes and response heterogeneity in an Applied Mathematics course, focusing on the topic “domain of a function.” It aims to quantify short-term learning gains and identify learner profiles based on attitudes, prior difficulties and study habits. Design/methodology/approach A pre/post experimental design was implemented with 81 first-year Business students enrolled in a Applied Mathematics course. Students completed a pretest, watched an instructional video and then solved three posttasks of increasing complexity. A stepwise scoring rubric measured learning gains, and Latent Class Analysis (LCA) was used to identify learner subgroups. Wilcoxon signed-rank and nonparametric tests examined overall and subgroup differences. Findings Results show statistically significant improvements across all posttasks (p < 0.001), including among students who reported frequent difficulties with Mathematics. Nevertheless, higher-complexity items continued to reveal persistent gaps in solving inequalities and applying set operations. Students highly valued the video’s clarity, audiovisual quality and usefulness but consistently indicated that it cannot replace face-to-face instruction. Research limitations/implications Results derive from one institution, one topic and a short-term intervention without delayed posttest; LCA indicators were self-reported. External validity is limited. Future work should include multiple topics/institutions, link to clickstream measures and assess retention/transfer with follow-up testing (Bergdahl et al., 2024). Practical implications Actionable guidance for Technology-Enhanced Learning (TEL): concise, segmented videos with signaling and worked examples; high audiovisual quality; prompt-and-pause moments; and immediate practice with feedback. Pair videos with targeted activities addressing inequalities and set operations, and deploy in flipped/blended models to support both recent and long-gap learners. Social implications Scalable microvideos can expand flexible access for diverse higher education (HE) cohorts, including working students and those resuming Mathematics after long gaps, helping narrow opportunity gaps. Equitable deployment requires accessibility features (e.g. captions) and preservation of teacher–student interaction to sustain motivation and inclusion. Originality/value While video-based learning in HE and Mathematics Education has been widely investigated, most empirical studies focus on full flipped-course implementations, video series or multiresource learning environments. This study adopts a fine-grained, microanalytic perspective by isolating the learning impact of a single, carefully designed instructional video on a specific mathematical topic (“Domain of a Function”). Methodologically, it combines a transparent stepwise scoring rubric with pre/post assessment and LCA to capture learner heterogeneity beyond average performance effects. This allows identification of distinct learner profiles (“Casual Learners” and “Struggling Video Users”) and examination of how short-term procedural gains vary across profiles. By demonstrating that a single procedural video can generate statistically significant short-term learning gains across heterogeneous student profiles while simultaneously revealing persistent conceptual difficulties, this study contributes empirical evidence that refines current understandings of the instructional scope and pedagogical limits of microvideo use in Higher Education Mathematics. It further distills evidence-based design principles for integrating short instructional videos into blended and flipped learning environments.

Gaps exist between critical thinking importance and an organized agenda for the development of critical thinking learning outcomes. Multiple critical thinking skill sets were developed by the authors, forming a symbiotic network. The key concept for each critical thinking exercise follows one learning model emulating the master clinician's thinking. The purpose is to enhance symbiosis in a pedagogical framework for individual exercises with a common outcomes-based assessment; internalization of each skill set is followed by nuance exploration-"learning moments." Descriptive analyses for (1) a symbiotic network of patient-based, student-led demonstrations of thinking and judgment with expanded opportunities for critical thinking, (2) a common outcomes-based assessment for patient-based student-led demonstrations of thinking and judgment, and (3) learning moments for each skill set: periodontics, operative dentistry (caries), geriatrics, ethics, pediatrics, TMD, and technology decision-making. Two factors were seen to contribute to interactions/sharing/symbiosis of learning guides among different departments: using a common emulation-learning model and using a common outcomes-based assessment. The use of a learning guide based on a designated thought process, combined with the use of a common outcomes-based assessment were also associated with subsequent rich and varied learning moments distinctive for each exercise. With little literature on learning outcomes for critical thinking, the development of an emulation learning model based on the thought processes of the master clinician led to the development initially of multiple critical thinking skill sets, followed by interactions of exercises. The result was a symbiotic network with rigorous demonstration of a skill set followed by a rich exploration of nuances.

Algorithmic thinking refers to students’ abilities and skills in understanding problems, formulating strategies, and creating algorithms. This study explores the integration of large language model-driven AI programming assistants into the cultivation of university students’ algorithmic thinking, and examines the effect of AI assistants on university students’ algorithmic thinking and self-efficacy. The results indicate that AI assistants, equipped with functions such as timely feedback, personalized support, emotional companionship, and cognitive scaffolding, exert a significantly positive effect on students’ algorithmic thinking and self-efficacy. This effect shows no differences across genders or learning styles. While AI assistants demonstrated clear advantages in areas such as real-time feedback, problem diagnosis, and generating diverse solutions, students perceived that teachers retained a crucial and complementary role in guiding goals and values, constructing knowledge frameworks, providing emotional support, and offering holistic learning guidance. Based on these perceptions, we propose that teachers and AI assistants can collaborate to form a collaborative model for cultivating algorithmic thinking. This study provides practical guidance for applying AI assistants in programming or algorithm courses to enhance students’ algorithmic thinking, and offers valuable insights for designing human-machine collaborative teaching methods and optimizing AI assistants.

The Role and Challenge of Knee Cartilage Magnetic Resonance Imaging in Early Diagnosis of Knee Osteoarthritis: A Literature Review.

Have UK radiologists embraced the peer learning guidance issued by the Royal College of Radiologists?

Conventional teacher-centered learning approaches result in superficial conceptual understanding and suboptimal development of students' practical skills. This study employed a qualitative approach with library research method on primary sources including theoretical works of Piaget, Vygotsky, and Woolfolk, as well as secondary sources from scientific journals related to constructivist learning. This study presents a new synthesis on the adaptation of constructivist learning within the context of Islamic values through: (1) epistemological dimension that integrates knowledge construction with tawhid and moral values, (2) methodological dimension that connects constructivist methods with munazarah, halaqah, and tadabbur traditions, (3) axiological dimension that directs learning toward character and spiritual formation, and (4) pedagogical dimension that positions teachers as murabbi and uswatun hasanah. The implementation of constructivist learning requires transformation of teachers' roles from information centers to facilitators and learning guides with formative evaluation that holistically assesses learning processes and outcomes.

This study aims to examine the use of historical sites in the city of Medan as a source of learning local history for students of the History Education Study Program, the Islamic University of North Sumatra (UISU). The research uses a descriptive qualitative approach with the type of field research. The research subjects consist of students and lecturers teaching local history courses. Data collection was carried out through observation, in-depth interviews, and documentation. The data was analyzed using an interactive model that included data reduction, data presentation, and conclusion drawn, and was tested for validity through triangulation of sources and techniques. The results of the study show that the use of historical sites in the city of Medan in learning local history has been carried out, but it has not been systematically integrated in lecture planning. Sites such as Maimun Palace, Al-Mashun Grand Mosque, Tjong A Fie Mansion, and the Kesawan area are used as contextual learning resources that can increase students' interest in learning, active engagement, and understanding of local history. Historical site-based learning also has a positive impact on the development of students' historical thinking skills and fosters awareness of the importance of preserving local cultural heritage. However, this study found several obstacles, including limited lecture time, the lack of optimal integration of field activities in the Semester Learning Plan, and the lack of academic guidance for historical site-based learning. This study recommends the need to develop an integrated and sustainable historical site-based local history learning model to improve the quality of history learning in higher education.

Learning problems among students at school remain a challenge in the educational process, such as difficulties in understanding learning materials, low learning motivation, and a lack of learning independence. These conditions require the presence of guidance services that are able to assist students in overcoming learning obstacles in a systematic manner. This study aims to identify effective forms of learning guidance and analyze the role of guidance organizations in schools in supporting students’ learning success. The research method used in this study is a literature review by examining various relevant sources, including books and scientific journal articles related to learning guidance and guidance organizations in schools. The results of the study indicate that effective learning guidance includes individual guidance, group guidance, learning assistance, as well as remedial and enrichment services. In addition, the success of learning guidance is strongly influenced by the role of well-coordinated guidance organizations involving guidance and counseling teachers, homeroom teachers, subject teachers, school principals, and parents. Synergy among these elements enhances the effectiveness of learning guidance services, improves students’ academic achievement, and creates a supportive learning environment. Therefore, well-planned learning guidance supported by a strong guidance organization plays an important role in improving the quality of learning processes and outcomes in schools

Designing a high-entropy alloy is a promising approach for stabilizing the Li anode, but selecting appropriate elemental compositions and developing scalable fabrication methods are key to optimizing its performance and promoting practical applications. Here, we adopt machine learning and density functional theory calculations to screen out a set of optimized alloy compositions of Fe, Co, Ni, Cu, and Zn, which demonstrate high structural strength, low strain coefficient, and strong Li adsorption. More importantly, we develop a universal and mild chemical metathesis method to construct a lithiophilic Fe-Co-Ni-Cu-Zn high-entropy alloy in situ onto Li metal. This alloy exhibits an amorphous structure with a controllable morphology and thickness. The modified Li electrode shows a homogeneous, isotropic, and dense interphase, which bears low interfacial impedance and high electrochemical stability, enabling fast charge transfer and uniform Li plating/stripping, thereby suppressing Li dendrites and side reactions. As a result, symmetric cells of the modified Li electrode achieve 900 h of stable cycling at a high current density of 10 mA cm-2, and asymmetric cells coupled with the high-loading LiFePO4, LiNi0.8Co0.1Mn0.1O2, or S/C cathodes exhibit significantly improved cell performance, especially long-term cycle stability. This study provides efficient machine learning guidance and scalable fabrication technology for developing high-entropy alloy-stabilized Li-metal anodes.

Antimicrobial resistance (AMR) is a major global health threat, with a disproportionate impact on low- and middle-income countries. The TEACH PROA-ECHO (Telementoring, Equity & Advocacy Collaboration for Health through Antimicrobial Stewardship) initiative aimed to strengthen antimicrobial stewardship programs (ASPs) in Latin America using the ECHO telementoring model. From March to December 2024, 80 healthcare institutions from 10 Latin American countries participated in biweekly telementoring sessions focused on case-based learning guidance on stewardship interventions and collaborative problem-solving related to antimicrobial prescribing and surveillance. A validated self-assessment tool was used to categorize baseline ASP development and monitor progress over time. Learning activities were evaluated using Moore's framework, and participating teams identified local barriers and enablers for stewardship activities. Of the 80 institutions, 73 (91%) completed the program. A total of 322 professionals were registered in the project, accounting for 2166 attendances. Moore's indicators showed high satisfaction (Net Promoter Score 79) and significant knowledge improvement after sessions (97.3% vs 81.4%; P < .0001). Mean ASP self-assessment scores increased from 53.1 to 64.0 (P < .0001). The proportion of institutions with intermediate or advanced ASP development rose from 56.2% to 83.6% (P < .001). Higher baseline scores were associated with for-profit institutions, ASP implementation longer than 5 years, and program continuity (lack of ASP interruptions in the last 5 years). Telementoring is an effective and scalable approach to strengthening antimicrobial stewardship in Latin America. The TEACH PROA-ECHO model represents a valuable strategy to support national AMR action plans in resource-limited settings.

In the era of artificial intelligence, teachers role transformation is a core issue in teachers professional development under the background of educational digitalization, which is directly related to the quality of education and teachers own practical adaptation. Currently, teachers roles are shifting from mere knowledge transmitters to learning guides and technological collaborative constructors. Meanwhile, teachers are confronted with multiple challenges such as insufficient technical integration capabilities, vague role cognition, and lagging external support systems. To address these issues, at the individual level, teachers should enhance their digital skills and ethical awareness to improve their transformation competence; at the teaching practice level, efforts should be made to integrate technology empowerment with emotional education to construct a new teaching model; at the external support level, it is necessary to improve the training system and evaluation mechanism, and optimize the institutional environment for transformation. These measures aim to help teachers adapt to the new requirements of intelligent education and provide practical references for the digital transformation of education.

This paper presents a technology-enhanced pedagogy for power-electronics labs. It combines self-regulated learning (SRL) and error-based learning (EBL). The approach guides students through three phases: forethought, performance, and reflection. Each phase uses formative checks, scaffolded feedback, and opportunities to retry. We implemented two modules: AC–AC and DC–AC converters. Both are run on microcontroller-based boards with real instruments. Oscilloscopes and Fast Fourier Transform (FFT) and Total Harmonic Distortion (THD) analyze errors in real time. We collected three Likert-scale surveys: one on the method and two on the lab modules. The students reported high satisfaction (≈4.5/5). They would recommend this method and felt more confident in facing mistakes. The results also show the role of difficulty and logistics in the task. These factors suggest more scaffolding for complex work. We conclude with design principles for SRL-oriented EBL in engineering laboratories. The principles focus on clear feedback loops, visible error signatures, and deliberate practice. The study offers practical guidance for postsecondary technology-enhanced learning.

This study is motivated by the low reading ability of elementary school students, which has an impact on their difficulties in understanding learning materials and on the development of early literacy skills. Among 20 first-grade students at SDI Bandar Kidul, Kediri City, 10 students are not yet fluent in reading, 7 students experience difficulties in reading, and 3 students are unable to read at all. This condition highlights the need for learning guidance (tutoring) as one of the possible solutions. The research problems addressed in this study are how learning guidance influences the reading development of first-grade students at SDI Bandar Kidul, Kediri City, and to what extent this influence contributes to improving students’ reading abilities. Accordingly, this study aims to examine the effect of learning guidance on the reading development of first-grade students at SDI Bandar Kidul, Kediri City, and to measure the extent of its contribution in helping students improve letter recognition, syllable decoding, reading words and simple sentences, as well as enhancing students’ self-confidence.

ABSTRACT While mobile applications are increasingly popular and effective for early childhood Chinese learning, the existing tools often suffer from fixed interaction flow, inflexible learning content, and a lack of personalized feedback. This severely limit the development of children’s Chinese literacy and reading skills. To address these limitations, this study integrates Generative Artificial Intelligence (GAI) with Augmented Reality (AR) to improve children’s literacy and reading skills, where AR environment can be used to provide children with a multi-modal interactive learning experience, whilst GAI technology enables the provision of customized learning guidance and dynamically generated learning resources. Additionally, we designed and implemented the Chinese learning application (GAR-CL) based on this approach, and a quasi-experimental study with 80 children and 10 parents was conducted to evaluate its effectiveness. Experimental results indicate that the experimental group (40 children) using GAR-CL demonstrated significantly superior learning performance compared to the control group (40 children) using AR applications without GAI, while also experience lower cognitive load. Additionally, questionnaire and interview results reveal that both children and parents expressed high satisfaction and strong willingness to use GAR-CL application.

In the era of intelligent media, the interaction between teachers and students in higher education is undergoing a profound transformation. The model has shifted from one-way transmission to multi-agent, two-way collaboration involving “teacher-student-AI (artificial intelligence)”. Interaction depth moves from surface Q&amp;A to deep thought engagement, supported by instant, precise feedback and a blended virtual-physical space. New forms such as data-driven personalized interaction and immersive collaborative learning have emerged. However, this evolution brings significant challenges: over-reliance on technology may weaken cognitive autonomy; virtual interaction risks emotional detachment and trust erosion; ethical concerns like algorithmic bias and data privacy arise; teachers’ roles become blurred; and evaluation systems lag behind technological advances. Future pathways should position AI as a supportive tool while upholding human centrality. Strengthening emotional connection through online-offline blending, reforming assessment to value process and growth, and empowering teachers as digitally literate “learning guides” and “emotional connectors” are key to building a healthy, sustainable interactive ecosystem.

Abstract Legged robots have demonstrated remarkable potential for dynamic locomotion and terrain adaptability, making them a prominent focus of research. However, achieving robust and agile bipedal running remains challenging due to the complex dynamics of legged locomotion. In this paper, we propose a reinforcement learning framework for robust bipedal running, incorporating a simple reference trajectory generator and an asymmetric actor-critic architecture. The reference generator, based on kinematics, provides diverse trajectory references while preserving key gait characteristics, facilitating efficient policy exploration. To mitigate the simulation-to-reality gap, we extract latent variables encoding environmental and motion information from dual historical observations. Our method simplifies the trajectory generation process while maintaining effective guidance for learning. Extensive simulation and physical experiments demonstrate that, compared to model-based and learning-based baselines, our approach achieves higher agility, more accurate velocity tracking, and stronger disturbance rejection while preserving gait stability. The resulting controller exhibits spring–mass running dynamics that remain robust on both flat and uneven terrains.

Recent works have shown the potential of diffusion models in computer vision and natural language processing. Apart from the classical supervised learning fields, diffusion models have also shown strong competitiveness in reinforcement learning (RL) by formulating decision-making as sequential generation. However, incorporating temporal information of sequential data and utilizing it to guide diffusion models to perform better generation is still an open challenge. In this paper, we take one step forward to investigate controllable generation with temporal conditions that are refined from temporal information. We observe the importance of temporal conditions in sequential generation in sufficient scenarios and provide a comprehensive discussion and comparison of different temporal conditions. Based on the observations, we propose an effective temporally-conditional diffusion model coined Temporally-Composable Diffuser (TCD), which extracts temporal information from interaction sequences and explicitly guides generation with temporal conditions. Specifically, we separate the sequences into three parts according to time expansion and identify historical, immediate, and prospective conditions accordingly. Each condition preserves non-overlapping temporal information of sequences, enabling more controllable generation when we jointly use them to guide the diffuser. Finally, we conduct extensive experiments and analysis to reveal the favorable applicability of TCD in offline RL tasks, where our method reaches or matches the best performance compared with prior SOTA baselines. Codes are available at https://github.com/JF-Hu/Temporally-Composable-Diffuser.