Selected work

Representative Papers

SoftBoard: A Multi-Agent Tool for the Creation and Evaluation of Low-Fidelity Prototypes

Jul 14, 2026

This work addresses the challenges faced by resource-constrained software startup teams with limited user experience (UX) expertise in efficiently creating and evaluating low-fidelity prototypes. To this end, we propose SoftBoard, a web-based multi-agent system that integrates large language model–driven intelligent agents into the prototyping workflow for the first time, enabling an end-to-end pipeline from requirement elicitation to automated generation of low-fidelity prototypes. The system incorporates an embedded evaluation mechanism based on usability heuristic rules and unifies prototype editing, team collaboration, and AI-assisted functionalities within a single platform, substantially reducing reliance on specialized UX knowledge. Preliminary feasibility studies demonstrate that SoftBoard effectively standardizes and streamlines the minimum viable product (MVP) development process.

0 citationsRead paper

Usable but Conventional: An Empirical Study on the UX of AI-Generated Interface Prototypes

May 14, 2026

This study presents the first systematic empirical investigation into the dual performance of user interfaces prototyped by generative artificial intelligence (GenAI), specifically examining the trade-off between utility and creativity. Through a blind experiment involving 92 participants, the research employed the UEQ-S questionnaire to evaluate GenAI-generated prototypes against human-designed counterparts across dimensions of pragmatic quality (e.g., usability, efficiency) and hedonic quality (e.g., originality, novelty). Findings indicate that AI-generated prototypes received positive assessments on pragmatic attributes but scored neutrally or negatively on hedonic aspects. These results highlight current limitations of GenAI in balancing functional effectiveness with creative expression in interface design, offering empirical insights to inform future human-AI collaborative approaches in user experience development.

0 citationsRead paper

Floquet Codes from Derived Semi-Regular Hyperbolic Tessellations on Orientable and Non-Orientable Surfaces

Mar 31, 2026

This work addresses the construction of high-performance quantum Floquet codes on compact orientable and non-orientable surfaces. The authors represent such surfaces via hyperbolic polygons and, for the first time, extend the framework of semi-regular hyperbolic tessellations to the non-orientable setting, enabling a systematic design of encoding structures suitable for high-genus and non-orientable topologies. By integrating tools from hyperbolic geometry, surface topology classification, semi-regular tiling theory, and quantum error-correcting code design, they generate several new families of Floquet codes. Performance analysis and asymptotic studies demonstrate their superior error-correction capabilities on higher-dimensional topological surfaces, significantly generalizing the existing construction framework for high-genus Floquet codes.

0 citationsRead paper

A derivative-free trust-region approach for Low Order-Value Optimization problems

Nov 25, 2025

This paper addresses the constrained low-order value optimization (LOVO) problem: minimizing the pointwise minimum of a finite number of continuously differentiable, yet derivative-free, black-box functions over a nonempty closed convex set. To tackle this nonlinear, derivative-free, and constrained LOVO problem, we propose— for the first time—a derivative-free trust-region algorithm based on linear interpolation surrogate models. We establish global convergence to weak critical points under mild assumptions and derive a worst-case iteration complexity bound. The algorithm relies solely on function evaluations, making it naturally suitable for robust parameter estimation, protein alignment, and portfolio optimization. An open-source implementation and comprehensive numerical experiments demonstrate that our method consistently outperforms existing surrogate-based alternatives in both convergence reliability and computational efficiency.

0 citationsRead paper

Nonlinear Rank Scaling and Hidden Structure in NHS Expenditure Transparency Data

Jun 24, 2025

The UK government’s 2010 mandatory disclosure policy for expenditures ≥£25,000—intended to enhance fiscal transparency—systematically omits over 90% of high-frequency, low-value transactions, creating a critical oversight gap. This study leverages transparency data from England’s National Health Service (NHS), including NHS England (NHSE) and Integrated Care Boards (ICBs), to conduct the first application of nonlinear rank-scaling analysis in public expenditure research. Methodologically, we employ rank–frequency distribution modeling, multi-segment power-law fitting, and multifractal scaling identification across supplier, expense-type, and category dimensions. Results reveal pronounced multiscale self-similarity in spending patterns—structurally analogous to Zipfian word-frequency and urban-size distributions. This challenges the policy’s binary disclosure threshold, exposing its structural neglect of dense, small-scale fiscal activity. The findings establish a novel analytical paradigm for optimizing transparency thresholds and advance understanding of organizational self-organization in public finance.

0 citationsRead paper

Recent publications

Latest Papers

SoftBoard: A Multi-Agent Tool for the Creation and Evaluation of Low-Fidelity Prototypes

Jul 14, 2026

This work addresses the challenges faced by resource-constrained software startup teams with limited user experience (UX) expertise in efficiently creating and evaluating low-fidelity prototypes. To this end, we propose SoftBoard, a web-based multi-agent system that integrates large language model–driven intelligent agents into the prototyping workflow for the first time, enabling an end-to-end pipeline from requirement elicitation to automated generation of low-fidelity prototypes. The system incorporates an embedded evaluation mechanism based on usability heuristic rules and unifies prototype editing, team collaboration, and AI-assisted functionalities within a single platform, substantially reducing reliance on specialized UX knowledge. Preliminary feasibility studies demonstrate that SoftBoard effectively standardizes and streamlines the minimum viable product (MVP) development process.

0 citationsRead paper

Usable but Conventional: An Empirical Study on the UX of AI-Generated Interface Prototypes

May 14, 2026

This study presents the first systematic empirical investigation into the dual performance of user interfaces prototyped by generative artificial intelligence (GenAI), specifically examining the trade-off between utility and creativity. Through a blind experiment involving 92 participants, the research employed the UEQ-S questionnaire to evaluate GenAI-generated prototypes against human-designed counterparts across dimensions of pragmatic quality (e.g., usability, efficiency) and hedonic quality (e.g., originality, novelty). Findings indicate that AI-generated prototypes received positive assessments on pragmatic attributes but scored neutrally or negatively on hedonic aspects. These results highlight current limitations of GenAI in balancing functional effectiveness with creative expression in interface design, offering empirical insights to inform future human-AI collaborative approaches in user experience development.

0 citationsRead paper

Floquet Codes from Derived Semi-Regular Hyperbolic Tessellations on Orientable and Non-Orientable Surfaces

Mar 31, 2026

This work addresses the construction of high-performance quantum Floquet codes on compact orientable and non-orientable surfaces. The authors represent such surfaces via hyperbolic polygons and, for the first time, extend the framework of semi-regular hyperbolic tessellations to the non-orientable setting, enabling a systematic design of encoding structures suitable for high-genus and non-orientable topologies. By integrating tools from hyperbolic geometry, surface topology classification, semi-regular tiling theory, and quantum error-correcting code design, they generate several new families of Floquet codes. Performance analysis and asymptotic studies demonstrate their superior error-correction capabilities on higher-dimensional topological surfaces, significantly generalizing the existing construction framework for high-genus Floquet codes.

0 citationsRead paper

A derivative-free trust-region approach for Low Order-Value Optimization problems

Nov 25, 2025

This paper addresses the constrained low-order value optimization (LOVO) problem: minimizing the pointwise minimum of a finite number of continuously differentiable, yet derivative-free, black-box functions over a nonempty closed convex set. To tackle this nonlinear, derivative-free, and constrained LOVO problem, we propose— for the first time—a derivative-free trust-region algorithm based on linear interpolation surrogate models. We establish global convergence to weak critical points under mild assumptions and derive a worst-case iteration complexity bound. The algorithm relies solely on function evaluations, making it naturally suitable for robust parameter estimation, protein alignment, and portfolio optimization. An open-source implementation and comprehensive numerical experiments demonstrate that our method consistently outperforms existing surrogate-based alternatives in both convergence reliability and computational efficiency.

0 citationsRead paper

Nonlinear Rank Scaling and Hidden Structure in NHS Expenditure Transparency Data

Jun 24, 2025

The UK government’s 2010 mandatory disclosure policy for expenditures ≥£25,000—intended to enhance fiscal transparency—systematically omits over 90% of high-frequency, low-value transactions, creating a critical oversight gap. This study leverages transparency data from England’s National Health Service (NHS), including NHS England (NHSE) and Integrated Care Boards (ICBs), to conduct the first application of nonlinear rank-scaling analysis in public expenditure research. Methodologically, we employ rank–frequency distribution modeling, multi-segment power-law fitting, and multifractal scaling identification across supplier, expense-type, and category dimensions. Results reveal pronounced multiscale self-similarity in spending patterns—structurally analogous to Zipfian word-frequency and urban-size distributions. This challenges the policy’s binary disclosure threshold, exposing its structural neglect of dense, small-scale fiscal activity. The findings establish a novel analytical paradigm for optimizing transparency thresholds and advance understanding of organizational self-organization in public finance.

0 citationsRead paper