Detecting Flow of Sensitive Data in Mini-Programs with Static Taint Analysis

An Empirical Study of Date and Time Bugs in Open-Source Python Software.

Quantifying prompt quality using information theory: entropy and mutual information analysis of 1,800 LLM generations

Large-scale empirical study analyzing 1.7M developer-written Java tests to characterize test scope, fixtures, assertions, inputs, and mocking patterns.

An empirical study evaluating the correctness of LLM-generated Python code for introductory programming tasks.

Official AI ROI Dataset (N=200). Longitudinal analysis of B2B AI deployments (2022-2025). Verified 159.8% median ROI. Peer-reviewed methodology for SMEs/ETIs and Fractional CAIOs.

Machine learning framework for detecting flaky tests and improving CI/CD pipeline reliability.

This repository contains the LaTeX source code and additional resources for a research paper that was accepted for publication at the 2021 Mining Software Repositories Conference.

Replication package for "Do LLMs Generate Adequate REST API Tests?" — ASE 2026 NIER

Replication package for "Who Tests Better When They Built It?" — ICSE 2027 NIER (under review)

Estudo empírico que avalia a assertividade lógica, desempenho (tempo/memória) e qualidade estrutural (Complexidade Ciclomática e Cognitiva) de código PHP gerado via Zero-Shot Prompting por Claude Sonnet 5, Gemini 3.5 Flash e GPT-5.5 Instant, usando 20 problemas do LeetCode.

📊 Explore how Shannon entropy and mutual information can quantify prompt quality in generative AI systems across various temperature settings.

📡 Visualize and understand amateur radio antenna patterns and physics through an interactive 3D platform designed for Ham Radio enthusiasts.

An empirical study of bug-fixing patterns in open-source Python repositories.

Do AI coding assistants change code quality? An ITS + DiD mining study over 272 treated / 299 control OSS repositories.

对 Karpathy「不要问 LLM 你怎么想」建议的实证批评。2天单人实验证明：Karpathy 说对了机制原文，但说错了使用建议。附量化测量框架和可复现方法。

Empirical study of linear, exact-kernel, and approximate-kernel classifiers on nonlinear binary classification tasks.

In the volatile telecommunications sector of Pakistan, projects often face extreme pressures. This study explores how Project Internal Social Capital (PISC)—the relational and cognitive bonds within a team—allows individuals to maintain performance. The study demonstrates that this is not a simple direct link, but a complex serial mediation.

A comparative empirical evaluation assessing the capabilities of AI coding assistants (GitHub Copilot and Windsurf) for generating Infrastructure-as-Code with Terraform.