Selected work

Representative Papers

GitSkills: A Dataset of Agent Skills on GitHub

Aug 11, 2026

This study addresses the lack of systematic empirical research on the authoring, reuse, and maintenance of agent skills for large language models in GitHub repositories. To bridge this gap, we introduce GitSkills, a novel dataset comprising 3,797,117 SKILL.md files collected from 282,200 public repositories. After deduplication via content hashing, the dataset contains 1,877,981 unique skills, each preserving its full textual content, YAML front matter, directory structure, and associated repository metadata. GitSkills is the first large-scale, systematically curated collection of agent skills written in natural language, released in SQLite format to support reproducible research. It establishes a foundational resource for empirical studies at the intersection of software engineering and AI agents, enabling multidimensional analyses of skill adoption, reuse patterns, evolution, provenance, and security.

0 citationsRead paper

Fisher's ideas and the design of field experiments in agronomy and plant breeding

May 28, 2026

This study addresses the inefficiency and insufficient data reliability in field trial design within agricultural and plant breeding contexts by integrating Fisher’s classical experimental design principles with contemporary breeding requirements. It systematically synthesizes methodologies including systematic designs, row–column designs, augmented row–column designs, multi-environment trials, and partially replicated designs to develop an optimized spatial allocation strategy tailored to specific target population of environments. The proposed framework substantially enhances the statistical efficiency and generalizability of field trial outcomes, offering a unified and robust theoretical foundation along with practical implementation guidelines for international crop breeding networks.

0 citationsRead paper

Optimal allocation of trials to sub-regions in crop variety testing with multiple years and correlated genotype effects

Apr 30, 2026

This study addresses the challenge of optimally allocating limited resources in multi-environment genotype trials to accurately predict regional performance under large-scale heterogeneity. It proposes a novel approach that integrates pedigree-based relatedness into multi-location trial design by constructing a structured variance–covariance matrix for genotype-by-environment effects, enabling information sharing across regions through best linear unbiased prediction (BLUP). By combining analytical derivations with numerical optimization, the method determines the optimal allocation of sub-regional testing efforts under a fixed budget. The framework scales efficiently to hundreds of genotypes, significantly enhancing resource use efficiency while maintaining high prediction accuracy, and is well-suited for large-scale breeding programs and major crop variety testing systems.

0 citationsRead paper

Operationalizing Fairness in Text-to-Image Models: A Survey of Bias, Fairness Audits and Mitigation Strategies

Apr 15, 2026

Text-to-image (T2I) generative models commonly exhibit social stereotypes, yet the lack of operational definitions for “bias” and “fairness” hinders effective evaluation and mitigation. This study addresses this gap through a systematic literature review, establishing a taxonomy of bias types and fairness concepts in T2I generation, and clarifying the distinction between target fairness and threshold fairness. It further evaluates a range of mitigation strategies—from prompt engineering to interventions in the diffusion process—and proposes an actionable fairness framework. By shifting the evaluation paradigm from descriptive metrics toward goal-oriented, rigorous testing, this work constructs a structured knowledge base for T2I fairness research, offering both theoretical grounding and practical pathways for the responsible development of generative AI systems.

0 citationsRead paper

Local Performance vs. Out-of-Distribution Generalization: An Empirical Analysis of Personalized Federated Learning in Heterogeneous Data Environments

Oct 28, 2025

To address the client drift and generalization imbalance in federated learning under non-independent and identically distributed (Non-IID) data, this paper identifies a critical limitation of existing personalization methods: their excessive focus on local accuracy while neglecting out-of-distribution (OOD) generalization—a fundamental pillar of FedAvg’s robustness. We propose a unified evaluation paradigm that jointly optimizes local accuracy and OOD generalization, and design FLIU, an adaptive personalization update mechanism. Within the FedAvg framework, FLIU introduces learnable, client-specific scaling factors to dynamically balance global consistency and local adaptability. Extensive experiments across MNIST and CIFAR-10 under IID, pathological Non-IID, and Dirichlet Non-IID settings demonstrate that FLIU achieves high local accuracy while significantly improving OOD generalization—outperforming state-of-the-art personalized federated learning methods.

0 citationsRead paper

Recent publications

Latest Papers

GitSkills: A Dataset of Agent Skills on GitHub

Aug 11, 2026

This study addresses the lack of systematic empirical research on the authoring, reuse, and maintenance of agent skills for large language models in GitHub repositories. To bridge this gap, we introduce GitSkills, a novel dataset comprising 3,797,117 SKILL.md files collected from 282,200 public repositories. After deduplication via content hashing, the dataset contains 1,877,981 unique skills, each preserving its full textual content, YAML front matter, directory structure, and associated repository metadata. GitSkills is the first large-scale, systematically curated collection of agent skills written in natural language, released in SQLite format to support reproducible research. It establishes a foundational resource for empirical studies at the intersection of software engineering and AI agents, enabling multidimensional analyses of skill adoption, reuse patterns, evolution, provenance, and security.

0 citationsRead paper

Fisher's ideas and the design of field experiments in agronomy and plant breeding

May 28, 2026

This study addresses the inefficiency and insufficient data reliability in field trial design within agricultural and plant breeding contexts by integrating Fisher’s classical experimental design principles with contemporary breeding requirements. It systematically synthesizes methodologies including systematic designs, row–column designs, augmented row–column designs, multi-environment trials, and partially replicated designs to develop an optimized spatial allocation strategy tailored to specific target population of environments. The proposed framework substantially enhances the statistical efficiency and generalizability of field trial outcomes, offering a unified and robust theoretical foundation along with practical implementation guidelines for international crop breeding networks.

0 citationsRead paper

Optimal allocation of trials to sub-regions in crop variety testing with multiple years and correlated genotype effects

Apr 30, 2026

This study addresses the challenge of optimally allocating limited resources in multi-environment genotype trials to accurately predict regional performance under large-scale heterogeneity. It proposes a novel approach that integrates pedigree-based relatedness into multi-location trial design by constructing a structured variance–covariance matrix for genotype-by-environment effects, enabling information sharing across regions through best linear unbiased prediction (BLUP). By combining analytical derivations with numerical optimization, the method determines the optimal allocation of sub-regional testing efforts under a fixed budget. The framework scales efficiently to hundreds of genotypes, significantly enhancing resource use efficiency while maintaining high prediction accuracy, and is well-suited for large-scale breeding programs and major crop variety testing systems.

0 citationsRead paper

Operationalizing Fairness in Text-to-Image Models: A Survey of Bias, Fairness Audits and Mitigation Strategies

Apr 15, 2026

Text-to-image (T2I) generative models commonly exhibit social stereotypes, yet the lack of operational definitions for “bias” and “fairness” hinders effective evaluation and mitigation. This study addresses this gap through a systematic literature review, establishing a taxonomy of bias types and fairness concepts in T2I generation, and clarifying the distinction between target fairness and threshold fairness. It further evaluates a range of mitigation strategies—from prompt engineering to interventions in the diffusion process—and proposes an actionable fairness framework. By shifting the evaluation paradigm from descriptive metrics toward goal-oriented, rigorous testing, this work constructs a structured knowledge base for T2I fairness research, offering both theoretical grounding and practical pathways for the responsible development of generative AI systems.

0 citationsRead paper

Local Performance vs. Out-of-Distribution Generalization: An Empirical Analysis of Personalized Federated Learning in Heterogeneous Data Environments

Oct 28, 2025

To address the client drift and generalization imbalance in federated learning under non-independent and identically distributed (Non-IID) data, this paper identifies a critical limitation of existing personalization methods: their excessive focus on local accuracy while neglecting out-of-distribution (OOD) generalization—a fundamental pillar of FedAvg’s robustness. We propose a unified evaluation paradigm that jointly optimizes local accuracy and OOD generalization, and design FLIU, an adaptive personalization update mechanism. Within the FedAvg framework, FLIU introduces learnable, client-specific scaling factors to dynamically balance global consistency and local adaptability. Extensive experiments across MNIST and CIFAR-10 under IID, pathological Non-IID, and Dirichlet Non-IID settings demonstrate that FLIU achieves high local accuracy while significantly improving OOD generalization—outperforming state-of-the-art personalized federated learning methods.

0 citationsRead paper