Selected work

Representative Papers

The Capability Ladder: A Curriculum-Modernization Framework for Workforce Readiness in the AI Era

Aug 07, 2026

This study addresses the misalignment between rapidly evolving skill demands driven by artificial intelligence and the current computing education landscape. It proposes a competency-assured curriculum modernization framework that reconceptualizes AI-induced task transformation not as replacement but as redistribution. The framework introduces a five-tier “competency ladder”—spanning Trigger, Automation, Workflow, AI Agent, and Agent Team—to delineate levels of human autonomy and oversight in human-AI collaboration. Guided by this model, the work outlines strategies for curriculum redesign, assessment development, and stackable credentialing. Feasibility is demonstrated through a structured narrative review and a two-semester interdisciplinary pilot course integrating labor market analytics, empirical software engineering, and no-code collaborative platforms, offering computing and business students a targeted pathway to thrive in AI-augmented workplaces and facilitating a smooth educational transition toward human-AI collaborative paradigms.

0 citationsRead paper

VetClaw: An Edge-Cloud Multimodal Agentic System for Veterinary Disease Screening

Jul 28, 2026

This work addresses the limitations of static-image reliance, insufficient multimodal integration, and lack of systematic workflows in early veterinary disease screening by proposing an edge-cloud collaborative multimodal agent system. The system captures images via edge devices and fuses them with symptom-related text, leveraging a cloud-based vision-language model for zero-shot disease classification. It innovatively decouples agent interaction from workflow orchestration, enabling dynamic tool invocation, enforcement of safety protocols, failure recovery, and escalation of uncertain cases to human experts. Experimental results demonstrate that multimodal inputs significantly outperform unimodal image-based classification, and the overall system exhibits practical diagnostic assistance capabilities suitable for real-world deployment.

0 citationsRead paper

Artificial Intelligence in Elementary STEM Education: A Systematic Review of Current Applications and Future Challenges

Oct 30, 2025

This study addresses eight critical gaps in AI integration within elementary STEM education: fragmented empirical evidence, insufficient interdisciplinary integration, geographic imbalance, and deficiencies in privacy protection and equity. A systematic review of 258 empirical studies published between 2020 and 2025 identifies key trends: 65% focus on upper elementary grades; 38% concentrate exclusively on mathematics; only 15% achieve authentic interdisciplinary STEM integration; and merely 34% report standardized effect sizes—while geographic representation remains heavily skewed toward North America, East Asia, and Europe. To bridge these gaps, the study proposes a novel four-dimensional technical framework—centered on *teacher agency*, *developmentally appropriate design*, *privacy-by-design*, and *STEM disciplinary integration*. It synthesizes seven AI-enabled pedagogical pathways: intelligent tutoring systems, learning analytics, computer vision, educational robotics, multimodal sensing, AI-augmented extended reality (XR), and adaptive content generation—advancing AI’s role in education from isolated tool deployment to holistic, ecosystem-level reconstruction.

0 citationsRead paper

A Graph-Attentive LSTM Model for Malicious URL Detection

Oct 12, 2025

Malicious URL detection faces challenges in identifying novel and obfuscated URLs, while traditional blacklist-based approaches suffer from poor generalization. To address this, we propose an end-to-end deep learning model that jointly integrates Graph Neural Networks (GNNs), Graph Attention Networks (GATs), and Long Short-Term Memory (LSTM) networks. Specifically, URLs are modeled as character-level graphs; GATs capture salient local topological patterns, while LSTMs encode long-range sequential dependencies—enabling, for the first time, unified representation of both structural and sequential URL characteristics within a single framework. Robustness is further enhanced via one-hot encoding, domain-informed feature engineering, and class-balanced sampling. Evaluated on a large-scale dataset of 650,000 URLs, the model achieves an accuracy of 0.9806 and a weighted F1-score of 0.9804, demonstrating significantly improved generalization to previously unseen and morphed malicious URLs.

0 citationsRead paper

Multiple Regression Analysis of Unmeasured Confounding

Aug 11, 2025

This paper addresses bias in causal effect identification arising from unmeasured confounding in observational data. We propose a quantitative sensitivity analysis method grounded in a multiple regression framework. Our key innovation extends the confounding interval approach—previously limited to single-regression settings—to multivariate regression, leveraging observed covariates and domain knowledge (particularly the coefficient of determination, $R^2$) to derive theoretical bounds on omitted-variable bias and thereby achieve partial identification of causal effects. The method supports $R^2$-based bound sensitivity analysis, enabling quantification of estimation uncertainty induced by unmeasured confounders, and is accompanied by an open-source implementation. Simulation studies and empirical applications demonstrate its robustness even under natural stochasticity, offering an interpretable and actionable tool for uncertainty assessment in causal inference.

0 citationsRead paper

Recent publications

Latest Papers

The Capability Ladder: A Curriculum-Modernization Framework for Workforce Readiness in the AI Era

Aug 07, 2026

This study addresses the misalignment between rapidly evolving skill demands driven by artificial intelligence and the current computing education landscape. It proposes a competency-assured curriculum modernization framework that reconceptualizes AI-induced task transformation not as replacement but as redistribution. The framework introduces a five-tier “competency ladder”—spanning Trigger, Automation, Workflow, AI Agent, and Agent Team—to delineate levels of human autonomy and oversight in human-AI collaboration. Guided by this model, the work outlines strategies for curriculum redesign, assessment development, and stackable credentialing. Feasibility is demonstrated through a structured narrative review and a two-semester interdisciplinary pilot course integrating labor market analytics, empirical software engineering, and no-code collaborative platforms, offering computing and business students a targeted pathway to thrive in AI-augmented workplaces and facilitating a smooth educational transition toward human-AI collaborative paradigms.

0 citationsRead paper

VetClaw: An Edge-Cloud Multimodal Agentic System for Veterinary Disease Screening

Jul 28, 2026

This work addresses the limitations of static-image reliance, insufficient multimodal integration, and lack of systematic workflows in early veterinary disease screening by proposing an edge-cloud collaborative multimodal agent system. The system captures images via edge devices and fuses them with symptom-related text, leveraging a cloud-based vision-language model for zero-shot disease classification. It innovatively decouples agent interaction from workflow orchestration, enabling dynamic tool invocation, enforcement of safety protocols, failure recovery, and escalation of uncertain cases to human experts. Experimental results demonstrate that multimodal inputs significantly outperform unimodal image-based classification, and the overall system exhibits practical diagnostic assistance capabilities suitable for real-world deployment.

0 citationsRead paper

Artificial Intelligence in Elementary STEM Education: A Systematic Review of Current Applications and Future Challenges

Oct 30, 2025

This study addresses eight critical gaps in AI integration within elementary STEM education: fragmented empirical evidence, insufficient interdisciplinary integration, geographic imbalance, and deficiencies in privacy protection and equity. A systematic review of 258 empirical studies published between 2020 and 2025 identifies key trends: 65% focus on upper elementary grades; 38% concentrate exclusively on mathematics; only 15% achieve authentic interdisciplinary STEM integration; and merely 34% report standardized effect sizes—while geographic representation remains heavily skewed toward North America, East Asia, and Europe. To bridge these gaps, the study proposes a novel four-dimensional technical framework—centered on *teacher agency*, *developmentally appropriate design*, *privacy-by-design*, and *STEM disciplinary integration*. It synthesizes seven AI-enabled pedagogical pathways: intelligent tutoring systems, learning analytics, computer vision, educational robotics, multimodal sensing, AI-augmented extended reality (XR), and adaptive content generation—advancing AI’s role in education from isolated tool deployment to holistic, ecosystem-level reconstruction.

0 citationsRead paper

A Graph-Attentive LSTM Model for Malicious URL Detection

Oct 12, 2025

Malicious URL detection faces challenges in identifying novel and obfuscated URLs, while traditional blacklist-based approaches suffer from poor generalization. To address this, we propose an end-to-end deep learning model that jointly integrates Graph Neural Networks (GNNs), Graph Attention Networks (GATs), and Long Short-Term Memory (LSTM) networks. Specifically, URLs are modeled as character-level graphs; GATs capture salient local topological patterns, while LSTMs encode long-range sequential dependencies—enabling, for the first time, unified representation of both structural and sequential URL characteristics within a single framework. Robustness is further enhanced via one-hot encoding, domain-informed feature engineering, and class-balanced sampling. Evaluated on a large-scale dataset of 650,000 URLs, the model achieves an accuracy of 0.9806 and a weighted F1-score of 0.9804, demonstrating significantly improved generalization to previously unseen and morphed malicious URLs.

0 citationsRead paper

Multiple Regression Analysis of Unmeasured Confounding

Aug 11, 2025

This paper addresses bias in causal effect identification arising from unmeasured confounding in observational data. We propose a quantitative sensitivity analysis method grounded in a multiple regression framework. Our key innovation extends the confounding interval approach—previously limited to single-regression settings—to multivariate regression, leveraging observed covariates and domain knowledge (particularly the coefficient of determination, $R^2$) to derive theoretical bounds on omitted-variable bias and thereby achieve partial identification of causal effects. The method supports $R^2$-based bound sensitivity analysis, enabling quantification of estimation uncertainty induced by unmeasured confounders, and is accompanied by an open-source implementation. Simulation studies and empirical applications demonstrate its robustness even under natural stochasticity, offering an interpretable and actionable tool for uncertainty assessment in causal inference.

0 citationsRead paper