Virtual reality (VR) can foster self-directed learning (SDL), yet traditional feedback in self-directed VR (SDVR) environments often fails to provide timely and individualized support, particularly for average and low achievers. Recent advancements in GPT-based assistants may offer adaptive, real-time feedback to address these limitations. This study employed a stratified randomized controlled design with 83 undergraduates (experimental n = 42; control n = 41) enrolled in an embedded-AI VR course. Learners completed SDVR units on embedded-AI hardware and block-based programming, followed by two hands-on tasks (EAI assembly and EAI programming). Primary outcome measures included SDL abilities, learning motivation, cognitive levels, and hands-on performance. Compared with traditional feedback, GPT-based feedback was associated with higher post-test SDL, motivation, cognitive level, and hands-on scores, with the most substantial gains observed among average and low achievers (LAs). The GPT-based feedback showed significant main effects across all outcomes (p < .001), explaining a substantial portion of the variance in cognitive level ( ω _p^2 = 0.397), SDL abilities ( ω _p^2 = 0.299), and motivation ( ω _p^2 = 0.397), representing large effect sizes. These findings demonstrate that integrating GPT-based adaptive feedback into SDVR environments provides timely and personalized support that enhances SDL, motivation, and higher-level cognitive and practical performance, particularly for learners with weaker prior achievement. The study highlights the potential of generative AI to support more personalized and equitable learning in immersive VR settings.
Classifying code snippet-based questions is essential for teaching, preparing assessment materials, and supporting intelligent learning systems in programming education. Traditional frequency-based encodings, such as TF-IDF, often fail to capture the contextual semantics within code-related questions. This study employs contextualized embeddings generated by the large language model Text-Embedding-3-Large (TE3L) to evaluate their effectiveness in classifying code-related questions. It further investigates which classifier architecture best complements the TE3L representation. Using a small-scale dataset of 171 SQL certification-style questions representative of course-level repositories, we analyze the classification complexity reduced by the TE3L scheme compared to TF-IDF. Then, we investigate classification performance under various classifier architectures with TE3L embeddings, including single models, boosting, and stacking ensembles. Results demonstrate that the TE3L scheme significantly reduces classification complexity and improves performance compared to the TF-IDF. Single classifiers, particularly the support vector machine with a linear kernel and the stochastic gradient descent classifiers, performed the best with the TE3L scheme and achieved an 11-percentage-point relative improvement over the benchmark in the weighted macro-average F1 score. The boosting and stacking techniques did not enhance performance, reflecting the challenges of ensemble learning under small-sample, imbalanced conditions. This work highlights the practical value of using LLM-based embeddings to automate question classification in low-resource educational contexts, supporting teachers in building intelligent assessment tools without requiring deep expertise in NLP or machine learning.
Elongation at fracture is one of the most sensitive yet difficult-to-predict mechanical properties in selective laser melting (SLM), due to complex and nonlinear interactions among process parameters and defect formation mechanisms. In this study, a curated dataset of more than 400 experimentally reported data points was systematically constructed from the literature, linking key SLM parameters—including laser power, scan speed, hatch distance, layer thickness, and spot size—to elongation. Statistical analysis revealed that individual parameters exhibit only weak linear correlations with elongation, highlighting the limitations of traditional regression-based approaches. To address this challenge, a unified machine learning framework was developed to benchmark seven predictive models spanning linear, kernel-based, distance-based, ensemble, and neural network paradigms. The results demonstrate that nonlinear and ensemble models significantly outperform linear approaches, with Gradient Boosting achieving the highest predictive accuracy (R² = 0.871), followed by Artificial Neural Networks (R² = 0.815) and Random Forest (R² = 0.699). In contrast, linear models explain less than 20
Abstract Background Studies have demonstrated the benefits of therapeutic music listening in reducing anxiety for patients undergoing breast-related surgical procedures. However, specific factors contributing to the effects remain unclear, nor do the patient experience with such interventions. Purpose This study aimed to explore factors associated with peri-operative music listening that influence anxiety and patient satisfaction. Methods This study employed a prospective, cross-sectional observational design with a descriptive focus. Forty patients undergoing breast-related surgical procedures were enrolled at Guangzhou Concord Cancer Center, with 39 completing the survey. Recruitment and data collection commenced after surgical procedures that integrated peri-operative music listening. Considering effects of the intervention on psycho-physiological indicators during the peri-operative period, patient perspectives regarding their experience was collected via electronic surveys for analysis. Results Results indicated significant reductions in peri-operative anxiety (p < 0.001), pre-operative heart rate (p = 0.23), pre-operative respiration rate (p = 0.008), post-operative heart rate (p = 0.003), while post-operative blood pressure increased (systolic: p = 0.020; diastolic: p = 0.027). Music listening duration showed no significant impact on changes in anxiety (r = -0.108, p = 0.513), procedural satisfaction (r = -0.026, p = 0.876), or music selection satisfaction (r = 0.015, p = 0.933). Music selection satisfaction positively correlated with procedural satisfaction (r = 0.766, p < 0.001). Over half of the participants had no preference between self-selected and interventionist-selected music and emphasized the need for trained interventionists. Conclusion Findings suggest that improving music selection satisfaction may enhance procedural experiences. Future considerations for peri-operative music-based interventions include trained interventionists, music therapist-curated selections, purpose-tailored music, and interventions suitable for patients' peri-operative states. Limitations include a small sample size and no control group. Future studies should investigate the effects of listening duration on pre- and post-operative anxiety, refine the music selection, and develop training guidelines for medical staff to implement effective and patient-need-based interventions.
Accurate detection of junction lines is critical in automated footwear manufacturing, as the junction line determines the processing path for roughening and adhesive application during sole–upper assembly. However, material flexibility, geometric variability among shoe styles, and deformation during pressing make reliable detection difficult in industrial environments. This study proposes an adaptive junction line detection method for robotic footwear assembly based on contour sensing and shoe-last-based geometric matching. The method captures assembly contours using a contour sensor and matches them with a reference shoe-last model to identify junction line positions without relying on visual appearance features. Experimental results demonstrate that, under the nominal local-normal-aligned sensing condition, the proposed approach achieves a maximum detection error below 0.3 mm, with most junction line recognition processes completed within 10 s and all detections finished within 20 s. In addition, one shoe-last model can be used for multiple shoe styles formed on the same last and size, thereby reducing modeling cost and improving adaptability in industrial footwear manufacturing.
